summaryrefslogtreecommitdiff
path: root/Master/texmf-dist/source/generic/mfpic/grafbase.dtx
blob: a309f242f6eab55d35a8b586b92c0ac3f54656fe (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
% \iffalse
% File: grafbase.dtx
% A part of mfpic 1.06 2011/03/08
%
% -------------------------------------------------------------------
%
% Copyright 2002--2011, Daniel H. Luecking
%
% Mfpic may be distributed and/or modified under the conditions of the
% LaTeX Project Public License, either version 1.3b of this license or (at
% your option) any later version. The latest version of this license is in
%    <http://www.latex-project.org/lppl.txt>
% and version 1.3b or later is part of all distributions of LaTeX version
% 2003/12/01 or later.
%
% Mfpic has maintenance status "author-maintained". The Current Maintainer
% is Daniel H. Luecking. There are several Base Interpreters: plain TeX, LaTeX,
% plain Metafont and plain MetaPost.
%
%<*driver>
\ProvidesFile{grafbase.dtx}
  [2011/03/08 v1.06. Metafont/post macros to interface with mfpic.]%
\documentclass{ltxdoc}
\usepackage{docmfp}

\addtolength{\textwidth}{.5878pt}

\def\mytt{\upshape\mdseries\ttfamily}
\renewcommand\marg[1]{{\mytt \{#1\}}}
\renewcommand\oarg[1]{{\mytt  [#1]}}
\renewcommand\parg[1]{{\mytt  (#1)}}
\renewcommand{\meta}[1]{{$\langle$\rmfamily\itshape#1$\rangle$}}
\DeclareRobustCommand\cs[1]{{\mytt\char`\\#1}}
\def\prog#1{{\mdseries\scshape #1}}
\def\grafbase{\prog{grafbase}}
\def\Grafbase{\prog{Grafbase}}
\def\mfpic{\prog{mfpic}}
\def\Mfpic{\prog{Mfpic}}
\def\MF{\prog{meta\-font}}
\def\MP{\prog{meta\-post}}
\def\PS{\prog{Post\-Script}}
\def\CMF{\prog{Meta\-font}}
\def\CMP{\prog{Meta\-post}}
\def\opt#1{{\sffamily\upshape#1}}
\def\mfc#1{{\mytt#1}}
\let\env\mfc
\let\file\mfc
\let\gbc\mfc
\renewcommand\{{{\mytt\char`\{}}
\renewcommand\}{{\mytt\char`\}}}
\renewcommand\|{${}\mathrel{|}{}$}

\makeatletter
\newcommand\bsl{{\mytt\@backslashchar}}
% Stupid lists!
\def\@listi{\leftmargin\leftmargini
  \parsep \z@ \@plus\p@ \@minus\z@
  \topsep 4\p@ \@plus\p@ \@minus2\p@
  \itemsep\parsep}
\let\@listI\@listi \@listi
\renewcommand\labelitemi{\normalfont\bfseries \textendash}
\renewcommand\labelitemii{\textasteriskcentered}
\renewcommand\labelitemiii{\textperiodcentered}
\leftmargini\parindent
% Stupid index!
\def\usage#1{\textrm{#1}}
\def\index@prologue{\section*{Index}\markboth{Index}{Index}}
\def\IndexParms{%
  \parindent \z@ \columnsep 15pt
  \parskip 0pt plus 1pt
  \rightskip 5pt plus2em \mathsurround \z@
  \parfillskip=-5pt \small
  % less hanging:
  \def\@idxitem{\par\hangindent 20pt}%
  \def\subitem{\@idxitem\hspace*{15pt}}%
  \def\subsubitem{\@idxitem\hspace*{25pt}}%
  \def\indexspace{\par\vspace{10pt plus 2pt minus 3pt}}}
\renewcommand\routinestring{}
\renewcommand\variablestring{\space(var.)}
% Why does every command have to be indexed twice?
\renewcommand\SpecialMfpIndex[3]{\@bsphack
  \index{%
    \string#1\actualchar
    \string\verb\quotechar*\verbatimchar\string#1\verbatimchar
    #2 \encapchar usage}%
  \@esphack}
\def\close@crossref{\SpecialEscapechar{:}}
\makeatother
\def\VariableIndex#1{\SpecialMfpIndex{#1}{\variablestring}{}}
\def\RoutineIndex #1{\SpecialMfpIndex{#1}{}{}}

\def\pdfTeX{\textrm{pdf\kern.04em\TeX}}
\def\pdfLaTeX{\textrm{pdf\kern.06em\LaTeX}}
\def\ConTeXt{\textrm{Con\kern-.16em\TeX\kern-0.06em t}}
\def\PiCTeX{\textrm{P\kern-.13em\lower.3ex\hbox{I}C\TeX}}

\title{The \grafbase{} macros\thanks{This file has version number
        \fileversion, last revised \filedate. The code described here
        was developed by several people, notably Thomas Leathrum,
        Geoffrey Tobin and Dan Luecking. Dan wrote this documentation.}}
\author{Dan Luecking}
\date{\filedate}
\SpecialEscapechar{:}
\def\bslash{:}
\DisableCrossrefs
\CodelineIndex
\AlsoImplementation

\begin{document}
  \DeleteShortVerb{\|}
  \DocInput{grafbase.dtx}
\end{document}
%</driver>
%\fi
%
% \CheckSum{1473}
% \CharacterTable
%  {Upper-case    \A\B\C\D\E\F\G\H\I\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z
%   Lower-case    \a\b\c\d\e\f\g\h\i\j\k\l\m\n\o\p\q\r\s\t\u\v\w\x\y\z
%   Digits        \0\1\2\3\4\5\6\7\8\9
%   Exclamation   \!     Double quote  \"     Hash (number) \#
%   Dollar        \$     Percent       \%     Ampersand     \&
%   Acute accent  \'     Left paren    \(     Right paren   \)
%   Asterisk      \*     Plus          \+     Comma         \,
%   Minus         \-     Point         \.     Solidus       \/
%   Colon         \:     Semicolon     \;     Less than     \<
%   Equals        \=     Greater than  \>     Question mark \?
%   Commercial at \@     Left bracket  \[     Backslash     \\
%   Right bracket \]     Circumflex    \^     Underscore    \_
%   Grave accent  \`     Left brace    \{     Vertical bar  \|
%   Right brace   \}     Tilde         \~}
%
% \catcode`\_=12
% \GetFileInfo{grafbase.dtx}
% \maketitle
%
% \begin{abstract}
% Macros are defined for use with \mfpic{}. The latter is a set of \TeX{}
% macros which allows a \file{.tex} file to write a \file{.mf} or
% \file{.mp} file that, with the help of these macros and \MF{} (or \MP),
% can be used to create pictures in the document, especially mathematical
% pictures. There are two versions of \grafbase, one for \MF{} and one for
% \MP{}. As they are more alike than different (95\% of the code is
% identical), we document both here.
%
% This file documents the \grafbase{} source code. The user manual for
% \mfpic{} is distributed as \file{mfpic-doc.pdf}, produced from
% \file{mfpic-doc.tex}. An introductory guide to \mfpic{} is available
% in \file{mfpguide.pdf}, produced from \file{mfpguide.pdf}
% \end{abstract}
%
% \StopEventually{\PrintIndex}
% \tableofcontents
%
% \section{Introduction}\label{intro}
%
% \subsection{Identification and checks}\label{checks}
%
% \DescribeVariable{grafbaseversion} We use \mfc{grafbaseversion} to
% check if \prog{grafbase} has been previously loaded, later we use it
% to check a mismatch with the version of \mfpic{} (if used).
% \gbc{grafbase} was used in previous versions. If either is known, we
% bail out. The \gbc{grafbase} boolean is really never needed, but it
% has been around since I took over. It is possible to write files that
% change behavior when \grafbase{} is loaded, however if they say
% ``\gbc{if grafbase:}, they can only be called with \gbc{grafbase} a
% known boolean.
%
% \VariableIndex{fileversion} \gbc{fileversion} and
% \VariableIndex{filedate} \gbc{filedate} provide identifying information.
%
% \DescribeRoutine{GBmsg}
% These are used fairly consistently and identify the source of the
% message delivered as being `\gbc{Grafbase}'.
% \DescribeRoutine{GBwarn}
% Warnings are delivered by \gbc{GBwarn}. The macro \gbc{GBerrmsg} takes
% care of both the
% \DescribeRoutine{GBerrmsg}
% error message and the \mfc{errhelp} string.
%    \begin{macrocode}
%<*MF|MP>
if (known grafbaseversion) or (known grafbase):
  message "Grafbase (" & jobname & "): You have loaded grafbase more "
    & "than once! Please make sure that it is loaded only once.";
  endinput;
fi
boolean grafbase; grafbase := true;

string fileversion, filedate;
fileversion := "1.06";  filedate := "2011/02/25";

message " Loading grafbase macros, version " & fileversion & ", " &
  filedate & ".";
message " ";

def GBmsg expr s =  message "Grafbase (" & jobname & "): " & s;  enddef;
def GBwarn expr s =  GBmsg "Warning, " & s;  enddef;
def GBerrmsg (expr s) expr t =  errhelp t;
  errmessage "Grafbase (" & jobname & "): " & s;  errhelp "";
enddef;

%    \end{macrocode}
%
% \DescribeVariable{MFPIC}
% The \gbc{MFPIC} variable is not yet used at all.
% It might be possible (at some point) to optimize things for \mfpic{} use
% whenever \gbc{mfpicversion} is defined, but so far we don't do anything
% except test the version and set this boolean.
%
% \DescribeVariable{mfpicversion}
% The output file written by the \mfpic{} macros includes a test that
% \gbc{mfpicversion} and \mfc{grafbaseversion} are the same,
% but that would fail to catch a new \grafbase{} with an old \mfpic. So
% we also put a test here with
% \DescribeRoutine{checkversions}
% \gbc{checkversions}, though it alone would fail to catch the use
% of older versions of \grafbase{} with current versions of \mfpic. Newer
% versions of \mfpic{} signal their version before inputting
% \file{grafbase}. Unfortunately (for error checking), \grafbase{} can
% also be used without \mfpic, so if \gbc{mfpicversion} is unknown, we
% merely write a message.
%    \begin{macrocode}
boolean MFPIC;  MFPIC := false;

def checkversions (expr g)=
  numeric grafbaseversion;  grafbaseversion := g;
  if unknown mfpicversion:  % no mfpic, or  < 0.63
    GBmsg "Recent mfpic not detected.";
  elseif g = mfpicversion:
    MFPIC := true;
  else:
    message "";
    GBwarn "Version mismatch: "
         & "mfpic and grafbase versions do not match.";
    message "";
  fi
enddef;

checkversions (106);

%    \end{macrocode}
%
% We try to make sure that the macros of \file{plain.mf} or
% \file{plain.mp} (where \mfc{base_name} is defined to be \mfc{"plain"})
% are available.
%    \begin{macrocode}
if unknown base_name       :  input plain;
elseif not string base_name:  input plain;
elseif base_name <> "plain":  input plain;
fi

%    \end{macrocode}
%
% \DescribeRoutine{GBdebug}
% The \gbc{debug} flag is for developers, who should set it before
% inputing \file{grafbase}.
% \DescribeRoutine{GBenddebug}
% These two routines start and end debug messages.
%
% \DescribeRoutine{mftitle}
% The \gbc{mftitle} macro is  useful when debugging.
% It will put its argument, which should be a string, as a TFM comment,
% and also print it to the terminal and log file.
%    \begin{macrocode}
if not boolean debug:  boolean debug;  debug := false;  fi

def GBdebug =
  begingroup
    save >>;  def >> =  message  "  " &  enddef;
    message "Grafbase DEBUG";
enddef;
def GBenddebug =
    message "End DEBUG";
    >> "";
  endgroup
enddef;

vardef mftitle expr t =
  if string t:  t;  message t;  fi
enddef;

%    \end{macrocode}
%
% \DescribeVariable{METAPOST}
% We try to determine which of \MF{} or \MP{} is using these macros.
% Perhaps one day we'll merge both versions of \grafbase{} in one file and
% use the following boolean. For now, we only use it to catch cases where
% the \grafbase{} file is being used by the wrong compiler.
%
% Of course, \MP{} natively knows about colors but \MF{} doesn't, so we
% use that to set a boolean \gbc{METAPOST}. We don't simply check if
% `\mfc{blue}' is \mfc{known} because `\mfc{blue}' is certainly a legal
% variable name in \MF. Instead we check `\mfc{known color X}' for
% some unlikely \gbc{X}. In \MP, `\gbc{color X}' is either true or
% false (\gbc{X} is a color or it isn't) and therefore always known, so
% `\gbc{known color X}' is always true.
%
% In \MF{} `\mfc{color X}' is an identifier (presumably unknown) with the
% base name \mfc{color} and suffix \mfc{X}.
%    \begin{macrocode}
boolean METAPOST;
METAPOST := known color Geamparalele din Babadag;

%<*MF>
if METAPOST:
  GBerrmsg ("wrong compiler.")
    "This file is for Metafont. For Metapost use grafbase.mp.";
fi
%</MF>

%<*MP>
if not METAPOST:
  GBerrmsg ("wrong compiler.")
    "This file is for Metapost. For Metafont use grafbase.mf.";
fi

%    \end{macrocode}
%
% \MP{} now exists in a couple of slightly incompatible versions.
% Versions 1.000 and later (beta versions 0.900 also) have native support for
% \texttt{CMYK} colors with a \mfc{cmykcolor} data type. It also
% supports grayscale colors (i.e., \mfc{withcolor} will accept a numeric
% expression), and has the alias \mfc{rgbcolor} for \mfc{color}.
% It also has a means to set the  name of the output file: the
% \mfc{filenametemplate} command. In versions 1.2 and later, this is
% deprecated in favor of setting the internal string variable
% \mfc{outputtemplate}.
%    \begin{macrocode}
numeric metapostversion;
boolean has_cmyk;
boolean has_filenametemplate; % e.g., filenametemplate "%j.%n;
boolean has_outputtemplate; % e.g., outputtemplate := "%j.%n;
if unknown mpversion:
  % prior to introduction of cmykcolor and output filename templates.
  % Latest such version was 0.641.
  metapostversion := 0.641;
  has_cmyk       := false;
  has_filenametemplate := false;
  has_outputtemplate := false;
else:
  metapostversion := scantokens (mpversion);
  has_cmyk := true;
  has_filenametemplate := true;
  if metapostversion < 1.200:
    has_outputtemplate := false;
  else:
    has_outputtemplate := true;
  fi
fi

%</MP>
%    \end{macrocode}
%
% \subsection{Setting up the font, \MF{} only}\label{font}
%
% Font-related housekeeping is only for \MF{}. \MF{} only produces
% fonts, so we have to define the variables it thinks are needed for
% fonts.
%
% \DescribeVariable{GBgeneric}
% We intercept the \mfc{mode} variable before \mfc{mode_setup} can set
% \mfc{proof} mode. We used to set \mfc{mode := cx} (and later
% \mfc{ljfour}) if it was unknown. For a while we just issued an error
% message. In this version we define a 600dpi mode called \mfc{GBgeneric}
% as a fallback (neither \gbc{mode} nor \gbc{localfont} known).
%
% The font identifier and coding scheme are just for information and end
% up as comments in the \file{.tfm} file (in all capitals). The design
% size just needs to be rather large for graphics, and \mfc{128pt\#} is
% anyway the default if we didn't set it ourselves.
%    \begin{macrocode}
%<*MF>
if unknown mode:
  GBerrmsg ("Metafont mode is unknown.")
    "Set mode to a known mode, perhaps ljfour. "
  & "If you proceed, localfont will be tried. "
  & "If that is unknown, a generic mode will be tried.";
  if known localfont:  mode := localfont;
  else:
    if unknown GBresolution:  GBresolution := 600  fi;
    mode_def GBgeneric =
      mode_param (pixels_per_inch, GBresolution);
      mode_param (blacker, 0);
      mode_param (fillin, 0);
      mode_param (o_correction, 1);
      mode_common_setup_;
    enddef;
    mode := GBgeneric;
  fi
fi
mode_setup;
if debug:
  GBdebug;
    >> "pixels_per_inch = " & decimal pixels_per_inch;
  GBenddebug;
fi

font_identifier := "MFpic graphics";
font_coding_scheme := "Arbitrary";
interim designsize := 128pt#;

%</MF>
%    \end{macrocode}
%
% \Mfpic-generated files make reference to \mfc{aspect_ratio} and
% \mfc{pt\#}, while \MP{} has no need for them. Rather than make
% \mfpic{} write different things, and to make the files intended for
% \MF{} also work with \MP, we define them in the obvious way. We also
% add a definition of \mfc{hppp} and \gbc{t_} to simplify maintenance of
% two versions of the \file{grafbase} files. Then we define
% \gbc{currenttransform} for \MP{} sake.
%    \begin{macrocode}
%<MP>pt# := pt;  bp# := bp;
%<MP>def t_ =  transformed currenttransform  enddef;
if unknown aspect_ratio:  aspect_ratio := 1;  fi
if unknown hppp:  hppp := 1  fi;
if unknown currenttransform:
  transform currenttransform;
  currenttransform := identity yscaled aspect_ratio;
fi

%    \end{macrocode}
%
% Don't complain when variables get too large. For \MF{} this \emph{must}
% be after \mfc{mode_setup}. Also don't complain if a clockwise path is
% filled (only \MF{} worries about this).
%    \begin{macrocode}
interim warningcheck := 0;
%<MF>interim turningcheck := 0;

%    \end{macrocode}
%
% \subsection{Initializations}\label{init}
%
% \VariableIndex{unitlen}
% \VariableIndex{xscale}
% \VariableIndex{yscale}
% \VariableIndex{xneg}
% \VariableIndex{xpos}
% \VariableIndex{yneg}
% \VariableIndex{ypos}
% The following are the various variables determining the extent of a
% picture. These variables would normally be set by a user for each
% picture, or by \mfpic, but we give them default values anyway. They
% give a nominal picture size of one inch with a graph unit corresponding
% to $1/10$ inch.
%    \begin{macrocode}
numeric unitlen, xscale, yscale, xneg, xpos, yneg, ypos;

unitlen := 1 bp#;
xscale := 7.2;
yscale := 7.2;
xneg := 0;  xpos := 10;
yneg := 0;  ypos := 10;

%    \end{macrocode}
%
% \DescribeVariable{deg}
% \VariableIndex{degree}
% We support both degrees and radians for angles.  In \MF, one degree is
% the unit of angle.
% \DescribeVariable{radian}
% One radian is $180/\pi$ degrees. We also define \gbc{pi} so a user can say
% \gbc{90} or \gbc{90deg} or \gbc{pi/2*radian}
% \DescribeVariable{pi}
% for the same effect. Actually, not quite: because of \MF{}'s precision
% limits, the latter is about 90.00025 degrees. \MF{}'s precision is 16
% binary places, or slightly under 5 decimals. The accuracy of \gbc{pi}
% and \gbc{radian} is the maximum possible. If we \emph{define}
% \gbc{radian} to be \gbc{90/(pi/2)} or \gbc{180/pi} the value of
% \gbc{pi/2*radian} is even less accurate.
%    \begin{macrocode}
newinternal deg, pi, radian;
deg := 1;  pi := 3.14159;
radian := 57.29578;
numeric degree;  degree := deg;

%    \end{macrocode}
%
% \VariableIndex{drawpen}
% \VariableIndex{penwd}
% \DescribeRoutine{resizedrawpen}
% Since we need to do this frequently, we define a macro that changes the
% pen width for subsequent drawing. This enables the file written by
% \mfpic{} to be less cluttered. At least that was the original reason.
% Now it gives us the opportunity to localize changes to \mfc{currentpen}
% and \gbc{drawpen}. (We already had this for different
% \gbc{beginmfpic}, since that reinitializes drawpen, but now it is local
% to other groups as well.)
%
% \VariableIndex{hatchpen}
% We could do this for the hatching pen, but it doesn't seem to change as
% often. The \mfc{pickup} command performs \mfc{yscaled aspect_ratio}, but
% so does the \gbc{shpath}, the only other place pens are required. In
% fact, we wouldn't need to \mfc{pickup} the pen at all, except power
% users may want to rely on \gbc{drawpen} always being the current pen. We
% make its diameter \mfc{.5pt} for backward compatibility.  But many
% journal publisher (e.g., AMS) recommend no smaller than \mfc{.5bp} for
% author-supplied drawings.
%
% \VariableIndex{hatchwd}
% The default \gbc{hatchwd} used to be larger, but it seemed ugly to me.
% (Backward compatibility? What's that?).
%    \begin{macrocode}
newinternal penwd;  penwd := 0.5pt;
pen drawpen;

def resizedrawpen (expr s) =
  interim penwd := s;
  setvariable (pen) (drawpen) pencircle scaled penwd;
  save currentpen;  pen currentpen;  pickup drawpen;
enddef;

numeric hatchwd;  hatchwd := 0.5bp;
pen hatchpen;  hatchpen := pencircle scaled hatchwd;

%    \end{macrocode}
%
% \DescribeVariable{clipall}
% We have two booleans related to clipping. One, \gbc{clipall} is meant to
% be turned on just once (per picture), and it causes the \gbc{endmfpic}
% code to clip the current picture to the boundaries defined by the
% picture size variables. The other,
% \DescribeVariable{ClipOn}
% \gbc{ClipOn}, is meant to be turned on and off. While on, most drawing
% macros (all?) will clip their result to the current \emph{clipping path
% array}.
% \DescribeVariable{ClipPath}
% The clipping path array is an array of paths: \gbc{ClipPath[\,]} together
% with a numeric \gbc{ClipPath}. The numeric variable contains the number
% of clipping paths; the paths are \gbc{ClipPath[1]} through
% \gbc{ClipPath[ClipPath]}. A macro later on is defined to loop through
% the array, clipping the current picture to the union of their interiors.
%
% \DescribeVariable{truebbox}
% The \gbc{truebbox} boolean sets the bounding box of the picture to its
% natural size in \MP. The default behavior of \MP{} is to output a
% bounding box that is the natural size of the graphic. The \grafbase{}
% default is to override this default, setting \gbc{truebbox} to
% \mfc{false}. \CMF's default behavior is to force the user to specify the
% bounding box, and provides no natural way to obtain any information
% about the actual extent of the ink. So, for now, this boolean is only
% for \MP.
%
% \DescribeRoutine{DoClip}
% This is for the frequent conditional code to implement \gbc{ClipOn}.
% The command \gbc{clipsto} is defined later.
%
% \DescribeRoutine{noclip}
% For debugging we sometimes want to make sure something is drawn
% without clipping being applied. For this we have \gbc{noclip}.
%    \begin{macrocode}
boolean clipall;  clipall := false;
boolean ClipOn;  ClipOn := false;
path ClipPath[];  numeric ClipPath;  ClipPath = 0;
boolean truebbox;  truebbox := false;

def DoClip (suffix v) =
  if ClipOn and (ClipPath > 0):  clipsto (v, ClipPath);  fi
enddef;

def noclip (text t) =
  hide ( setboolean (ClipOn) false;  t)
enddef;

%    \end{macrocode}
%
% \DescribeVariable{showbbox}
% The boolean \gbc{showbbox} is for debugging the \gbc{*bbox} macros.
%    \begin{macrocode}
boolean showbbox;  showbbox := false;

%    \end{macrocode}
%
% \subsubsection{Colors}\label{colors}
%
% Of course colors are only recognized by \MP. The colors \mfc{black},
% \mfc{white}, \mfc{red}, \mfc{green} and \mfc{blue} are part of
% \file{plain.mp}. We define other standard colors to get all eight
% colors where the coordinates are 0 or 1.
%
% \DescribeRoutine{color}
% We define \MF{} replacements for some of the \MP{} color variables and
% macros. Our point of view will be: make each color variable a numeric in
% \MF{}. Each will lie between $0$ and $1$ representing shades of gray.
% For \emph{drawing} commands we will only distinguish between nonwhite
% (black, ${}<1$) and white (${}\ge1$). For filling commands we will allow
% levels in between, and fill with an approximation using a version of
% the \gbc{shade} macro.
%    \begin{macrocode}
%<*MF>
let color = numeric;  color black, white;
let rgbcolor = numeric;
let cmykcolor = numeric;
black := 0;  white := 1;
def withcolor text t =  enddef;
%</MF>
%<MP>if not has_cmyk:  let rgbcolor = color;  let cmykcolor = color;  fi
def _wc_ =  withcolor  enddef;

%    \end{macrocode}
%
% \VariableIndex{currentcolor}
% \VariableIndex{drawcolor}
% \VariableIndex{fillcolor}
% \VariableIndex{hatchcolor}
% \VariableIndex{headcolor}
% \VariableIndex{pointcolor}
% \VariableIndex{tlabelcolor}
% We also define some color variables whose names reflect their use.
% Thus, \gbc{fillcolor} is used for filling, etc. The color
% \gbc{currentcolor} isn't used anywhere yet. The color
% \mfc{background} is used in \MP{} for unfilling a region.
%    \begin{macrocode}
color currentcolor, drawcolor, fillcolor, hatchcolor,
  headcolor, pointcolor, tlabelcolor, background;
currentcolor := fillcolor := drawcolor := hatchcolor :=
  headcolor := pointcolor := tlabelcolor := black;
background := white;

%    \end{macrocode}
%
% \DescribeRoutine{snapto}
% The \gbc{snapto} macro truncates numerics to the $[0,1]$ range, but also
% returns a value ($0$) for unknown input. It used to do the same for
% nonnumeric input, but that should be an error.  It would have made at
% least one of our bugs easier to find if it had produced an error message
% back then.
%    \begin{macrocode}
vardef snapto expr t =
  if numeric t:
    if unknown t:  0
    elseif t < 0:  0
    elseif t > 1:  1
    else:          t
    fi
  else:
    GBerrmsg ("Improper expression type.")
      "The argument to `snapto' must be a numeric.";
  fi

enddef;

%    \end{macrocode}
%
% \DescribeRoutine{cmykgray}
% \DescribeRoutine{rgbgray}
% Now we deal with all the color functions and utilities that enable
% \mfpic{} users to use colors without knowing what they are doing.
% Since colors now come in three flavors, we start with gray levels in
% the three models. In early \MP{}, the \gbc{cmyk} function will return
% an \opt{rgb} color, so there is will be no  difference between these
% two. In \MF{} \gbc{white} is a numeric and \gbc{cmyk} returns a
% numeric, so these can be used with \MF{}, and both produce the same
% result.
%
% \DescribeRoutine{grayscalegray}
% The grayscale version should return a numeric in recent \MP{}, so
% it needs a different definition for early \MP{}. Thus, it occurs
% in the conditional code. Oddly, its definition is the same for \MF{}
% and recent \MP{}.
%    \begin{macrocode}
vardef rgbgray  (expr g) =  (snapto g) * white  enddef;
vardef cmykgray (expr g) =  cmyk(0,0,0,1 - snapto g)  enddef;

%    \end{macrocode}
%
% In recent \MP{}, all the color functions are essentially no-ops. In
% early \MP{}, they all return an \opt{rgb} color expression. In \MF{}
% they all return a numeric. It is easiest if we simply separate the three
% cases and write the code for each, rather than load all the functions
% with three-way booleans (often containing nested booleans).
%
% For all three engines we require a definition of the color functions
% \gbc{gray(g)}, \gbc{rgb(r,g,b)}, and \gbc{cmyk(c,m,y,k)}, as well as
% conversion functions \gbc{makegray(x)}, \gbc{makergb(x)}, and
% \gbc{makecmyk(x)}, and the boolean \gbc{iscolor clr}. The first three
% have to return numerics for \MF{}, colors for early \MP{}, and the
% associated color type for recent \MP{}.
%    \begin{macrocode}
%<*MP>
if has_cmyk :
  vardef grayscalegray (expr g) =  snapto g  enddef;
  vardef gray (expr g) =  grayscalegray (g)  enddef;
  vardef cmyk (expr c, m, y, k) =
    (snapto c, snapto m, snapto y, snapto k)
  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{colorchoice}
% The \gbc{colorchoice} function (like \cs{mathchoice} in \TeX{}, after
% which it was named) returns one of four bits of code: \gbc{D} (default)
% if the first argument is unknown or not one of the recognized color
% models, \gbc{N} if it is numeric, \gbc{R} if it is \mfc{rgbcolor}, and
% \gbc{C} if it is \mfc{cmykcolor}. These arguments have to be \mfc{text}:
% if they were `\mfc{expr}' \MP{} would try to evaluate them, with
% possible errors since some of them apply functions that are not relevant
% to the other types.
%    \begin{macrocode}
  def colorchoice (expr clr) (text D)(text N)(text R)(text C) =
    if unknown clr:        D
    elseif numeric clr:    N
    elseif rgbcolor clr:   R
    elseif cmykcolor clr:  C
    else:                  D
    fi
  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{makecmyk}
% \DescribeRoutine{makergb}
% \DescribeRoutine{makegray}
% In \gbc{makecmyk} and all the other `\gbc{make}' conversions, the
% default is to return black in the appropriate model, numerics produce
% gray, and cmyk or rgb is either retained unchanged or converted to the
% appropriate model.
%
% \DescribeRoutine{iscolor}
% A variable is taken to be a color if it can validly follow
% \mfc{withcolor}. This includes boolean, though we hope no one tries to
% use that.
%    \begin{macrocode}
  vardef makecmyk primary clr =
    colorchoice (clr)(cmykblack)(cmykgray(clr))
    (rgbtocmyk(redpart clr,greenpart clr,bluepart clr))
    (clr)
  enddef;
  vardef makergb primary clr =
    colorchoice (clr)(rgbblack)(rgbgray(clr))(clr)
    (cmyktorgb(cyanpart clr, magentapart clr, yellowpart clr,
        blackpart clr))
  enddef;
  vardef makegray primary clr =
    colorchoice (clr)(grayscaleblack)(grayscalegray(clr))
    (rgbtogray (redpart clr, greenpart clr, bluepart clr))
    (cmyktogray(cyanpart clr, magentapart clr, yellowpart clr,
        blackpart clr))
  enddef;
  vardef iscolor expr clr =
    (rgbcolor clr) or (cmykcolor clr) or (numeric clr) or (boolean clr)
  enddef;
else:
%    \end{macrocode}
%
% \DescribeRoutine{colorchoice}
% In early \MP{} \gbc{colorchoice} is a three-way choice, since
% \mfc{cmykcolor} is not an available data type, but numeric can still be
% interpreted as a gray.
%
% \DescribeRoutine{makecmyk}
% \DescribeRoutine{makergb}
% \DescribeRoutine{makegray}
% The \gbc{make*} functions are simpler in early \MP{}, though not as
% simple as in \MF{}. Ditto
% \DescribeRoutine{iscolor}
% \gbc{iscolor}. The parentheses in \gbc{iscolor} are necessary to force
% \MP{} to see this \mfc{color} as the start of a boolean expression and
% not a type declaration.
%    \begin{macrocode}
  vardef gray (expr g) =  rgbgray(g)  enddef;
  vardef grayscalegray (expr g) =  rgbgray(g)  enddef;
  vardef cmyk (expr c, m, y, k) =  rgb (1-c-k, 1-m-k, 1-y-k)  enddef;
  def colorchoice (expr clr) (text D)(text N)(text R) =
    if unknown clr:       D
    elseif numeric clr:   N
    elseif rgbcolor clr:  R
    else:                 D
    fi
  enddef;
  vardef makergb  primary clr =
    colorchoice (clr)(rgbblack)(rgbgray(clr))(clr)
  enddef;
  vardef makegray primary clr =
    colorchoice (clr)(rgbblack)(rgbgray(clr))
      (rgbtogray (redpart clr, greenpart clr, bluepart clr))
  enddef;
  def makecmyk =  makergb  enddef;
  vardef iscolor expr clr =  (color clr)  enddef;
fi

%</MP>
%    \end{macrocode}
%
% \DescribeMacro{knowncolor}
% Once we have \gbc{iscolor} all we need to do is add a test for
% \mfc{known} to get this boolean test.
%    \begin{macrocode}
vardef knowncolor expr clr =  (known clr) and (iscolor clr)  enddef;

%    \end{macrocode}
%
% These are the \MF{} versions. Everything pretty much returns its
% numeric argument or $0$ (black).
%    \begin{macrocode}
%<*MF>
vardef grayscalegray (expr g) =  snapto g  enddef;
vardef gray (expr g) =  grayscalegray (g)  enddef;
vardef cmyk (expr c, m, y, k) =  rgb (1-c-k, 1-m-k, 1-y-k) enddef;
vardef makegray  primary clr =
  if knowncolor clr:  clr  else:  black  fi
enddef;
def makergb  =  makegray  enddef;
def makecmyk =  makegray  enddef;
vardef iscolor expr clr =  (color clr)  enddef;

%</MF>
%    \end{macrocode}
%
% \DescribeRoutine{forcecolor}
% \DescribeRoutine{named}
% This is only used in the \gbc{named} function to force a color. In
% \MF{} the tests are all `\mfc{if numeric}'. In early \MP{} `\mfc{if
% cmykcolor}' is the same as `\mfc{if rgbcolor}'.
%    \begin{macrocode}
vardef forceclr (expr c) =
  if unknown c :
    if numeric c:  grayscaleblack
    elseif rgbcolor c:   rgbblack
    elseif cmykcolor c: cmykblack
    else: black
    fi
  elseif numeric c:  gray (c)
  elseif iscolor c:  c
  else:  black
  fi
enddef;
vardef named (suffix c) =  forceclr (c)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{togray}
% \DescribeRoutine{rgbtogray}
% \DescribeRoutine{cmyktogray}
% \DescribeRoutine{cmyktorgb}
% \DescribeRoutine{rgbtocmyk}
% These are used for the conversions. Strictly speaking they do not
% `convert' as they all take multiple numeric arguments rather than any
% sort of color. As \mfc{rgbcolor} exists in both early and recent
% \MP{} as the same data type, we need only distinguish \MF{} from \MP{}
% \DescribeRoutine{rgb}
% in the function \gbc{rgb},
%    \begin{macrocode}
vardef togray (expr r, g, b) =
  gray (sqrt((2r*r + 4g*g + b*b)/7))
enddef;

vardef rgbtogray (expr r, g, b) =
  togray(snapto r, snapto g, snapto b)
enddef;
vardef cmyktogray (expr c, m, y, k) =
  rgbtogray (1-c-k,1-m-k,1-y-k)
enddef;

vardef cmyktorgb (expr c,m,y,k) =
  rgb(1-c-k,1-m-k,1-y-k)
enddef;
vardef rgbtocmyk (expr r,g,b) =
  cmyk(1-r,1-g,1-b,0)
enddef;

vardef rgb (expr r, g, b) =
%<MF>  togray (snapto r, snapto g, snapto b)
%<MP>  (snapto r, snapto g, snapto b)
enddef;
vardef RGB (expr R, G, B) =
  rgb (R/255, G/255, B/255)
enddef;

%    \end{macrocode}
%
% \subsection{Arrays}\label{arrays}
%
% \gbc{ClipPath} is a typical example of an array. Arrays are based on the
% fact that a variable can be of a different type from (and can be almost
% completely unrelated to) the variables formed by putting numeric
% suffixes on it.
%
% \DescribeRoutine{list}
% The \gbc{list} macro is essentially due to Frank Michielsen, and assigns
% a \emph{list} (i.e., a comma separated sequence of expressions) to an
% array. Note that the items in the list have to be the same type, and the
% same type as \mfc{v[\,]}. But \mfc{v} itself must be numeric.
%
% \DescribeRoutine{map}
% The \gbc{map} macro takes two text parameters. The first is any
% procedure, the second is a list of expressions. The procedure is applied
% to each expression and the resulting new expressions are separated by
% commas, that is, a new list is generated (for use in \mfc{for} loops).
% This is full of possibilities for errors. One reared its head because
% the original version started with a comma indicating an empty starting
% expression (normally it would be ignored and that turn through the loop
% skipped). However, it managed to produce an error in a reasonable
% but unforeseen usage (which I've since forgotten) and so I added the
% \gbc{_map} variable that skips the comma on the first time through the
% loop. This routine is currently only used in the code that \mfpic's \
% \cs{plr} writes.
%    \begin{macrocode}
def list (suffix v) (text lst) =
  v := 0;  for _itm = lst:  v[incr v] := _itm;  endfor
  if v = 0:
    GBerrmsg ("No list to process!")
      "An attempt was made to produce an array from a "
    & "list of expressions having no valid entries.";
  fi
enddef;

def map (text proc) (text lst) =
  hide (_map := 0;)
  for _a = lst:
    if _map = 0:  hide (_map := 1;)  else:  ,  fi
    proc (_a)
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{knownnumericarray}
% Checks if a suffix is the name of an array. Requires \gbc{arr} to be a
% known positive integer, and all the variables \gbc{arr[n]} to be known
% for \gbc{n} from 1 to \gbc{arr}. Since we so far only need it for
% numeric arrays, we also check if each entry is numeric.
%    \begin{macrocode}
vardef knownnumericarray suffix arr =
  setboolean (_kna) (known arr) and (numeric arr);
  if _kna :
    _kna := (arr = floor arr) and (arr >= 1);
    for _idx = 1 upto arr :
      exitif not _kna;
      _kna := (known arr[_idx]) and (numeric arr[_idx]);
    endfor
  fi
  _kna
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{copyarray}
% This makes some code much more readable. It simply steps through an
% array and copies the values into another array. It is only used for
% numeric arrays so far, but could be used for any kind.
%    \begin{macrocode}
def copyarray (suffix src, dest) =
  for _idx = 1 upto src:  dest[_idx] := src[_idx];  endfor
  dest := src;
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{maparr}
% The \gbc{maparr} macro applies a procedure \gbc{proc} to each member of
% array \gbc{p[\,]} with \gbc{p} members. It returns nothing. It is currently
% unused, although it was once used for things like \gbc{maxpair}.
%    \begin{macrocode}
def maparr (text proc) (suffix p) =
  for _idx = 1 upto p:  proc (p[_idx]);  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{textpairs}
% This macro takes a suffix (name of an array to be constructed) and a
% list of pairs, and assigns them to the array. It is normally called from
% another macro, which does any necessary \mfc{save}-ing of the variable
% used for the array name. We used to include \mfc{save} in this macro,
% but ran into a problem once when the argument had a suffix. You can't
% apply \mfc{save} to a variable with a suffix. Moreover, `\mfc{save p}'
% also renders \mfc{p.x} unknown, so I judged it best to let whoever calls
% this macro decide what to save. Actually, now it expands to the more
% general command \gbc{gsetarray} with type \gbc{pair}. That command
% then reads the suffix argument that should follow.
%
% Since the above change was made, macros evolved so that \emph{all} uses
% of \gbc{textpairs} are now preceeded by \gbc{save}. Thus, I have now
% replaced them all with calls to \gbc{setpairs} (it calls \gbc{setarray},
% which \emph{does} \gbc{save} the variable). In all those cases, the
% `\gbc{saved}' variable is a temporary local array.
%
% \DescribeRoutine{setuniquepairs}
% This does the same but omits any pair if it is identical to the previous
% one. It \mfc{save}\,s the variable, since all its uses are internal
% and require that.
%    \begin{macrocode}
def textpairs =  gsetarray (pair)  enddef;

def setuniquepairs (suffix p) (text t) =
  save p;  pair p[];
  setpairs (_up) (t);
  if _up > 0:
    p := 1;  p1 := _up1;
    for _i = 2 upto _up:
      if _up[_i] <> p[p]:  p[incr p] := _up[_i];  fi
    endfor
  else:
    p := 0;
  fi
enddef;

%    \end{macrocode}
%
% \subsection{Utilities}\label{utilities}
%
% \DescribeRoutine{chpair}
% This applies a procedure \gbc{proc} (which maps numeric to numeric) to
% each part of pair \gbc{p}, and returns the resultant pair. I've decided
% not to use it (for efficiency), but to leave it defined for backward
% compatibility.\\
% \DescribeRoutine{floorpair}
% \gbc{floorpair} applies \mfc{floor} to both parts of a pair.\\
% \DescribeRoutine{ceilingpair}
% \gbc{ceilingpair} does the same with \gbc{ceiling}.\\
% \DescribeRoutine{hroundpair}
% \gbc{hroundpair} does the same with \gbc{hround}.
%
% All three could use \gbc{chpair} with \gbc{proc} equal to \mfc{floor},
% \mfc{ceiling} and \mfc{hround}, but I now code them directly.
%
% \DescribeRoutine{goodpair}
% This last one is used (only in \MF{}) to adjust pairs to the pixel grid.
% It is the only place \gbc{hroundpair} is used. None of these is used in
% the \MP{} version.
%    \begin{macrocode}
vardef chpair (text proc) (expr p) =
  (proc (xpart p), proc (ypart p))
enddef;

vardef floorpair (expr p) =  (floor (xpart p), floor (ypart p))  enddef;
vardef ceilingpair (expr p) =
  (ceiling (xpart p), ceiling (ypart p))
enddef;

%<*MF>
def hroundpair (expr p) =  (hround (xpart p), hround (ypart p))  enddef;
vardef goodpair (expr p) =  hroundpair(p.t_)  enddef;

%</MF>
%    \end{macrocode}
%
% \DescribeRoutine{emin}
% The macro \gbc{emin} differs from \prog{plain}'s \mfc{min} in that it
% allows only two values. It can therefore be coded simply, without the
% overhead of a \mfc{for}-loop.
% \DescribeRoutine{emax}
% \gbc{emax} is analogous. Both are needed so often that it is possible a
% significant amount of time is saved with these versions.
%
% \DescribeRoutine{pairmin}
% The macro \gbc{pairmin} operates on two pairs, returning a pair having
% the smaller of the two xparts and the smaller of the two yparts. Of
% course
% \DescribeRoutine{pairmax}
% \gbc{pairmax} is analogous, producing the maximum.
%
% \DescribeRoutine{minpair}
% The \gbc{minpair} macro returns the pair comprising the minimum $x$ and
% minimum $y$ coordinates of all pairs in the array \gbc{p[\,]}.
% \DescribeRoutine{maxpair}
% \gbc{maxpair} is analogous. Somehow, both of them have disappeared from
% \grafbase. They were formerly used only in the \gbc{*bbox} macros.
% That code used a loop to build an array of control points and these
% routines would \emph{each} loop through that. The current code uses one
% loop (instead of three) through the control points, updating both the
% maximum and minimum at each one.
%    \begin{macrocode}
vardef emin (expr a, b) =  if a < b:  a  else:  b  fi  enddef;
vardef emax (expr a, b) =  if a > b:  a  else:  b  fi  enddef;

vardef pairmin (expr z, w) =
    ( emin (xpart z,  xpart w), emin (ypart z, ypart w ) )
enddef;
vardef pairmax (expr z, w) =
    ( emax (xpart z,  xpart w), emax (ypart z, ypart w ) )
enddef;

vardef minpair (suffix p) =  setpair (_mp) p1;
  for _idx = 2 upto p - 1:  _mp := pairmin (_mp, p[_idx]);  endfor
  pairmin (_mp, p[p])
enddef;
vardef maxpair (suffix p) =  setpair (_mp) p1;
  for _idx = 2 upto p - 1:  _mp := pairmax (_mp, p[_idx]);  endfor
  pairmax (_mp, p[p])
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xprod}
% A binary operation between pairs $z\sb1$ and $z\sb2$ that returns the
% cross product $x\sb1 y\sb2 - x\sb2 y\sb1$. This gives, among other
% things, twice the area of the triangle with two sides $z\sb1$ and
% $z\sb2$. It is used only in \gbc{mkconvex}.
%    \begin{macrocode}
primarydef Z xprod W =  (xpart Z * ypart W - xpart W * ypart Z)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{force_initial}
% The command \gbc{force_initial} modifies a path so that it has all the
% same points and controls as before, except its first point is replaced
% with \mfc{p}.
% \DescribeRoutine{force_terminal}
% The command \gbc{force_terminal} replaces the last point. This is for
% cases where, theoretically, paths \gbc{f} and \gbc{g} should meet at an
% endpoint, but do not due to finite precision. Instead of doing
% \mfc{f..g}, which adds a random tiny segment, we adjust the endpoints to
% exactly match the other and do \mfc{f\&g}, producing a join without an
% additional segment.
%
% \DescribeRoutine{force_equal_ends}
% The command \gbc{force_equal_ends} forces the last point of the first
% path and the first point of the second to equal the average of their
% original values. It is the only one of these four actually used anywhere
% else in \grafbase.
% \DescribeRoutine{replace_ends_of_cycle}
% The command \gbc{replace_ends_of_cycle} applies something similar to a
% cycle.
%    \begin{macrocode}
def force_initial (expr p) (suffix f) =
  hide( setnumeric (_n) length f;
  f := p
    if _n = 0:
      {0,0}
    else:
      ..controls post0 (f) and pre 1 (f).. subpath (1,_n) of f
    fi;)
enddef;

def force_terminal (expr p) (suffix f) =
  hide(setpath (_f) reverse f;
  force_initial (p) (_f);
  f := reverse _f;)
enddef;

def force_equal_ends (suffix f, g) =
  hide(save _p;  pair _p;
  _p := .5[pnt[length f] (f), pnt0(g)];
  force_terminal (_p) (f);  force_initial (_p) (g);)
enddef;

def replace_ends_of_cycle (expr p) (suffix f) =
hide(
if cycle f:
  save _n;  _n := length f;
  f := p
    if _n = 0:  &cycle
    else:  .. controls post0 (f) and pre 1 (f) ..
      if _n = 1:  cycle
      else:  subpath (1, _n - 1) of f ..
        controls post[_n - 1](f) and pre[_n](f) .. cycle
      fi
    fi;
fi)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{intersects}
% A binary relation, with the precedence level (almost) that of other
% relations, produces \mfc{true} if \MF{} determines that the paths
% intersect, false otherwise. It also
% \DescribeVariable{thetimes}
% sets the pair variable \gbc{thetimes} and its parts \gbc{_Xtime} and
% \gbc{_Ytime}. Then
% \DescribeRoutine{misses}
% \gbc{misses} is the opposite relation, used when the intersection point
% is not needed. It only occurs in the (unused) code of \gbc{tightbbox}.
%    \begin{macrocode}
pair thetimes;
numeric _Xtime, _Ytime;
tertiarydef a intersects b =
  begingroup
    thetimes := a intersectiontimes b;
    _Xtime := xpart thetimes;
    _Ytime := ypart thetimes;
    (_Xtime > -1)
  endgroup
enddef;

tertiarydef a misses b =  ((a intersectiontimes b) < origin)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{makepicture}
% The \gbc{makepicture} command takes any expression and does what it can
% to make a picture from it.
%
% \DescribeRoutine{onepointpath}
% The \gbc{onepointpath} command takes a point and forces it to be a path.
% If a vardef takes a list of points and it \emph{must} return a path that
% perhaps \emph{must} be cyclic, it can use this as a fallback. If an
% \mfpic{} command such as \cs{arc} receives an invalid optional
% parameter, it won't know what command to write to the output file. It
% can use
% \DescribeRoutine{fallbackpath}
% \gbc{fallbackpath} as long as the first parameter is a point.
%
% \DescribeRoutine{even}
% \DescribeRoutine{divides}
% Of course \gbc{even} means \gbc{not odd}. The relation \gbc{divides}
% is true if the right side is an integer multiple of the left.
%    \begin{macrocode}
vardef makepicture (expr s) =
  if picture s:    s
%<MP>  elseif string s:  s infont defaultfont scaled defaultscale
  elseif path s:   picpath (s)
  else:            nullpicture
  fi
enddef;

vardef onepointpath (expr cyclic, q) =
  q if cyclic:  &cycle  else:  {0,0}  fi
enddef;

vardef fallbackpath (expr cyclic, p) (text t) =
  onepointpath (cyclic, p)
enddef;

def even =  not odd  enddef;

primarydef a divides b =
  ((b mod a) = 0)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{image}
% The \mfc{image} macro exists in \file{plain.mp} but not \file{plain.mf}.
% The purpose is to just use the \file{plain} \MF{} and \grafbase{} macros
% as you normally would, but wrap the whole thing in parentheses preceded
% by \gbc{X := image} to get all those things drawn on the picture
% variable \gbc{X}.
%
% \DescribeRoutine{beginimage}
% Instead of making lengthy drawing code a parameter, one might prefer an
% environment-like syntax, writing \gbc{X := beginimage } at the start
% and
% \DescribeRoutine{endimage}
% \gbc{endimage} at the end.
%
% \DescribeRoutine{makeimage}
% This is for the \mfpic{} command \cs{mfpimage}. It takes a suffix
% parameter (the name of the picture variable) and a coordinate pair (in
% graph coordinates). The drawing commands, up to the following
% \gbc{endimage}, draw on this picture variable with the given pair as the
% reference point.
%    \begin{macrocode}
%<*MF>
vardef image (text t) =
  newpicture (currentpicture);
  t;
  currentpicture
enddef;

%</MF>
def beginimage =
  begingroup
    newpicture (currentpicture);
enddef;
def endimage =
  ;  currentpicture
  endgroup
enddef;

def makeimage (suffix name) (expr refpt) =
  setpair (_image_reference_point) zconv(refpt);
  setpicture (name) beginimage
enddef;
def concludeimage  =
  endimage shifted
%<MF>    -goodpair (_image_reference_point)
%<MP>    -_image_reference_point
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{setvariable}
% This is are mainly to save space in \mfpic-generated files. In \grafbase{}
% itself the \mfc{save} is often inconvenient, but it turns out there are
% many cases where it \emph{is} used; enough so that we have abbreviations
% \RoutineIndex{setnumeric}\gbc{setnumeric},
% \RoutineIndex{setboolean}\gbc{setboolean},
% \RoutineIndex{setpair}\gbc{setpair},
% \RoutineIndex{setpath}\gbc{setpath},
% \RoutineIndex{setpicture}\gbc{setpicture},
% and \RoutineIndex{setstring}\gbc{setstring}, together with the
% common uses \RoutineIndex{newpicture}\gbc{newpicture} and
% \RoutineIndex{convertpath}\gbc{convertpath}. There is also a
% \gbc{setcolor}, but that has such a different definition that we reserve
% it for later.
%
% For completeness, we also include the remaining two abbreviations,
% \RoutineIndex{setpen}\gbc{setpen} and
% \RoutineIndex{settransform}\gbc{settransform}, even though they are not
% used anywhere in \grafbase{}.
%
% \DescribeRoutine{gsetvariable}
% The macro \gbc{gsetvariable} is the global version. It has no
% abbreviations, but it is occasionally needed for \mfpic{}. The only
% difference between it and the local version is the lack of a
% \gbc{save}. None of these commands take the value as a parameter. That
% should follow, and is picked up by the ending \mfc{:=}.
%
% \DescribeRoutine{setarray}
% Then \gbc{setarray} is the array version. It takes the same parameters
% as \gbc{setvariable}, but what should follow is a list of expressions in
% parentheses. It calls \gbc{list} to read each item into
% \gbc{name1}, \gbc{name2}, etc. There is also has a global version
% \DescribeRoutine{gsetarray}
% \gbc{gsetarray}.
% \DescribeRoutine{setpairs}
% \gbc{setpairs} is an abbreviation for arrays
% of pairs. Historically, it came first.
%    \begin{macrocode}
def setvariable (text kind) (suffix name) =
  save name;  kind name;  name :=
enddef;
def gsetvariable (text kind) (suffix name) =  kind name;  name :=
enddef;

def setnumeric (suffix name) =  save name;  name :=  enddef;
def setboolean   =  setvariable (boolean)  enddef;
def setpair      =  setvariable (pair)  enddef;
def setpath      =  setvariable (path)  enddef;
def setpicture   =  setvariable (picture)  enddef;
def setstring    =  setvariable (string)  enddef;
def settransform =  setvariable (transform)  enddef;
def setpen       =  setvariable (pen)  enddef;
def settension (suffix tn) expr tens =
  setnumeric (tn) if tens > 0:  tens  else:  default_tension  fi;
enddef;
def fixtension (suffix tn) =  if tn < .75:  tn := .75;  fi  enddef;

def newpicture (suffix pic) =  setpicture (pic) nullpicture;  enddef;
def convertpath (suffix g) expr f =  setpath (g) zconv (f);  enddef;

def setarray (text kind) (suffix name) =
  save name;  kind name[];  list (name)
enddef;
def setpairs =  setarray (pair)  enddef;
def gsetarray (text kind) (suffix name) =
  numeric name;  kind name[];  list (name)
enddef;

%    \end{macrocode}
% The next are slightly different, but seem to belong here.

% \DescribeRoutine{setbbox}
% In \gbc{setbbox} we save and initialize \emph{two} pair variables and
% set them to the bounding box of a path that should follow.
%
% \DescribeRoutine{setsplit}
% There are a couple of routines that modify a variable to make sure it is
% positive and integral. In a couple of places two routine \emph{must} use
% the same value. Here we isolate the code that does the modification, and
% then both routines call \gbc{setsplit}.
%    \begin{macrocode}
def setbbox (suffix ll, ur) =
  save ll, ur;  pair ll, ur;  getbbox (ll, ur)
enddef;

def setsplit (suffix s) expr ss =
  setnumeric (s) emax (1, ceiling ss);
enddef;

%<*MP>
if has_cmyk:
  def setrgbcolor  =  setvariable (rgbcolor)  enddef;
  def setcmykcolor =  setvariable (cmykcolor)  enddef;
  def setcolor (suffix name) expr val =
    if     boolean   val :  setboolean
    elseif numeric   val :  setnumeric
    elseif rgbcolor  val :  setrgbcolor
    elseif cmykcolor val :  setcmykcolor
    % this should give a suitable error message:
    else:  setvariable (color)
    fi  (name) val;
  enddef;
  def gsetcolor (suffix name) expr val =
    if     boolean   val :  boolean name;
    elseif numeric   val :  numeric name;
    elseif rgbcolor  val :  rgbcolor name;
    elseif cmykcolor val :  cmykcolor name;
    else:  color name;
    fi  name := val;
  enddef;
else:
  def setrgbcolor  =  setcolor  enddef;
  def setcmykcolor =  setcolor  enddef;
  def setcolor = setvariable (color) enddef;
  def gsetcolor = gsetvariable (color) enddef;
fi
%</MP>
%<*MF>
def setrgbcolor  =  setcolor  enddef;
def setcmykcolor =  setcolor  enddef;
def setcolor = setvariable (color) enddef;
def gsetcolor = gsetvariable (color) enddef;
%</MF>

%    \end{macrocode}
%
% And then the standard colors. Using the color functions ensures that
% they are defined in \MF{} as well as all versions of \MP{}. In early
% \MP{} they are all \mfc{rgbcolor}, in \MF{} they are all numeric. In
% recent \MP{}, they have the type correspondimg to the name of the
% color function, with \gbc{gray()} being numeric.
%    \begin{macrocode}
setcolor(rgbblack) rgb(0,0,0);
setcolor(red)      rgb(1,0,0);
setcolor(green)    rgb(0,1,0);
setcolor(blue)     rgb(0,0,1);
setcolor(rgbwhite) rgb(1,1,1);
setcolor(cmykwhite) cmyk(0,0,0,0);
setcolor(cyan)      cmyk(1,0,0,0); % Maybe these should
setcolor(magenta)   cmyk(0,1,0,0); % be rbg for backward
setcolor(yellow)    cmyk(0,0,1,0); % compatibility?
setcolor(cmykblack) cmyk(0,0,0,1);
setcolor(grayscaleblack) gray(0);
setcolor(grayscalewhite) gray(1);

%<*MP>
if has_outputtemplate:
  def setoutputtemplate =  outputtemplate :=  enddef;
elseif has_filenametemplate:
  def setoutputtemplate =  filenametemplate  enddef;
else:
  def setoutputtemplate text garbage =  enddef;
fi
%</MP>
%<MF>def setoutputtemplate text garbage = enddef;

%    \end{macrocode}
%
% \DescribeRoutine{GBromannumeral}
% We will append roman numerals to the ends of a variable name to
% emulate an array. This will be needed when our `array' consists of
% colors with  different types. \MP{} doesn't permit true arrays to
% contain different types. We use `\gbc{GB}' in the name because a
% package exists that defines \mfc{romannumeral} differently
%
% \DescribeRoutine{GBromandigit}
% Roman numerals can conveniently be computed one digit at a time. The
% algorithm is the same for each digit, differing only in the letters
% used. Thus we define \gbc{GBromandigit} and call it three times with
% different sets of letters.
%
% \DescribeRoutine{strrepeat}
% The helper macro \gbc{strrepeat} creates a new string by concatenating
% \mfc{rep} copies of the string \mfc{str}.
%    \begin{macrocode}
vardef GBromannumeral (expr X) =
  save Y, _tmp, U; string U;

  Y.m := X div 1000;        % thousands digit
    _tmp := X - 1000Y.m;    % hundreds digits and lower
  Y.c := _tmp div 100;      % hundreds
    _tmp := _tmp - 100Y.c;  % tens and units
  Y.x := _tmp div 10;       % tens
  Y.i := _tmp - 10Y.x;      % units

  strrepeat("m", Y.m) &
  GBromandigit("c", "d", "m", Y.c) &
  GBromandigit("x", "l", "c", Y.x) &
  GBromandigit("i", "v", "x", Y.i)
enddef;

vardef GBromandigit (expr bot, mid, top, n) =
  if n > 9 :    top & strrepeat(bot, n-10)   % shouldn't happen
  elseif n > 8 :
    bot & top                                % "ix"
  elseif n > 4 :  mid & strrepeat (bot, n-5) % "v"--"viii"
  elseif n > 3 :
    bot & mid                                % "iv"
  else:         strrepeat (bot, n)           % ""--"iii" for 0--3
  fi
enddef;

vardef strrepeat (expr st, rep) =
  "" for i = 1 upto rep:  & st  endfor
enddef;

%    \end{macrocode}
%
% \section{The \grafbase{} Coordinate System}\label{coordinate}
%
% We need to make a distinction between graph units, sharped units, and
% device units. In \MF, a device unit is 1 pixel. On a LaserJet IV, one
% inch is 600 pixels. When constructing a character, \MF{} uses the pixel
% as its unit. Since this differs from one printing device to another,
% \file{plain.mf} arranges for \emph{sharped} units (the name comes from the
% convention that they are written using a name that ends in \mfc{\#}). The
% dimension \mfc{1pt\#} in \MF{} is arbitrarily set to 1, and other
% units defined by conversion factors (\mfc{in\#=72.27}; neither \MF{}
% nor \MP{} makes a distinction between distances and numbers: \mfc{2pt}
% just means \mfc{2} times the value of \mfc{pt}).  When one needs to
% draw something actually \emph{one point long}, then \mfc{1pt} is used.
% It is defined to equal \mfc{pt\#*hppp}, where \mfc{hppp} stands for
% ``horizontal pixels per point'' and its value is usually set by
% \mfc{mode_setup}. So \mfc{1pt} is $600/72.27$ (pixels) if
% \mfc{mode} is \mfc{ljfour}.
%
% Often, when we want numbers not to become too large, we do calculations,
% define paths, etc., in sharped units, then draw by scaling to device
% units. In \grafbase{} we take this one step further: a horizontal graph
% unit (i.e., the difference between the graph points $(0,0)$ and $(1,0)$)
% represents \gbc{unitlen*xscale} sharped units, and
% \gbc{unitlen*xscale*hppp} actual pixels. The \grafbase{} macros do much
% of the calculations in graph units.
%
% In \MP, there is no difference between device and sharped units.
% The \emph{postscript point} or \emph{big point} (1/72 inches) is the
% unit in \MP: \mfc{bp = 1}.
%
% Some things need to be in graph units (for example, positions within a
% graph defined by the user) or independent of units (standard shapes)
% that scale appropriately when scales change. Other things (thickness of
% lines) are a design decision that either should be independent of scale
% or should scale in a nonobvious way. The diameter of the drawing pen is
% one of the latter things, so the default pen width is in device units.
% Also for the hatching pen.
%
% When drawing a path we want to use device coordinates. When defining
% paths, we typically want to use graph coordinates. The macros that do
% the drawing, therefore, need to convert from one to the other. In
% addition, for inclusion of the picture in a \TeX{} document, we normally
% want the lower left corner of the graph space to have device coordinates
% $(0,0)$.
%
% \subsection{The main transforms}\label{ztr}
%
% \DescribeVariable{vtr}
% We therefore have two transforms: \gbc{vtr} is the \emph{vector} or
% linear transform for pair quantities that remain invariant under shifts,
% and
% \DescribeVariable{ztr}
% \gbc{ztr} is a \emph{point} or affine transformation for pair quantities
% that change appropriately under shifts.
%
% The quantities \gbc{xneg}, \gbc{xpos}, \gbc{yneg}, and  \gbc{ypos} are
% in \emph{graph} coordinates. Shifting by \gbc{(-xneg, -yneg)} transforms
% the lower left corner to $(0,0)$.  Multiplication by \gbc{xscale} and
% \gbc{yscale} converts to multiples of \gbc{unitlen} and multiplication
% by \gbc{unitlen} gets us sharped coordinates. For \MF{},
% multiplication by \mfc{hppp} converts to device coordinates, while for
% \MP{} sharped and device are the same (the printer's PostScript
% rasterizing engine---\prog{GhostScript} perhaps---does the final
% conversion to actual pixels).
%
% In \MF{}, \mfc{currenttransform} (via the macro \mfc{.t_}, defined by
% \mfc{mode_setup}) takes care of the aspect ratio. In \MP{} the final
% rasterizer should do this.
%
% \gbc{charwd} and \gbc{charht} are sharped coordinates defined by the
% startup code \gbc{beginmfpic}, while \gbc{w_} and \gbc{h_} are the
% corresponding device (pixel) coordinates
%
% \DescribeRoutine{setztr}
% This macro does the defining of \gbc{ztr} and \gbc{vtr}. It is called
% by \gbc{beginmfpic}, at which time all the necessary quantities should be
% known.
%    \begin{macrocode}
transform ztr, vtr;
def setztr =
  if debug:
    GBdebug;
%<*MF>
      >> "charwd = " & decimal charwd & "pt#";
      >> "charht = " & decimal charht & "pt#";
      >> "w_ = " & decimal w_ & " pixels";
      >> "h_ = " & decimal h_ & " pixels";
      >> "unitlen = " & decimal unitlen & "pt#";
      >> "hppp = " & decimal hppp;
%</MF>
%<*MP>
      >> "w_ = " & decimal w_ & "bp";
      >> "h_ = " & decimal h_ & "bp";
      >> "unitlen = " & decimal unitlen & "bp";
%</MP>
      >> "xneg = " & decimal xneg;
      >> "xpos = " & decimal xpos;
      >> "yneg = " & decimal yneg;
      >> "ypos = " & decimal ypos;
      >> "xscale = " & decimal xscale;
      >> "yscale = " & decimal yscale;
    GBenddebug;
  fi
  save ztr, vtr;
  transform ztr, vtr;
  vtr := identity xscaled xscale yscaled yscale scaled (unitlen*hppp);
  ztr := identity shifted (-xneg, -yneg) transformed vtr;
  if debug:
    GBdebug;
      >> "ztr is";
      show ztr;
      >> "vtr is";
      show vtr;
    GBenddebug;
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{zconv}
% The macro \gbc{zconv} converts a variety of expressions from graph to
% device coordinates. The expressions include pairs, paths, and transforms.
% This is an affine transform. The inverse,
% \DescribeRoutine{invzconv}
% \gbc{invzconv}, converts a variety of expressions from device to graph
% coordinates.
%
% \DescribeRoutine{vconv}
% The vector version, \gbc{vconv}, converts a vector \gbc{v} from graph to
% device coordinates. This is a linear (ie, vector) transform. Also,
% \DescribeRoutine{invvconv}
% \gbc{invvconv} converts a vector from device to graph coordinates.
%    \begin{macrocode}
vardef zconv (expr a) =  a transformed ztr  enddef;
vardef invzconv (expr a) =  a transformed (inverse ztr)  enddef;
vardef vconv (expr v) =  v transformed vtr  enddef;
vardef invvconv (expr v) =  v transformed (inverse vtr)  enddef;

%    \end{macrocode}
%
% \subsection{The \gbc{mfpic} environment}\label{mfpic}
%
% \DescribeRoutine{active_plane}
% \gbc{active_plane} is the active drawing plane. \mfc{currentpicture} is
% unknown at this stage (because it's set in \gbc{beginmfpic}). We use a
% \mfc{def}, and not a picture assignment, partly for this reason but also
% because we can achieve special effects by redefining it (see the
% \gbc{tile} macro).
%    \begin{macrocode}
def active_plane =  currentpicture  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{initpic}
% \gbc{initpic} is called by \gbc{beginmfpic} after \gbc{w_} and
% \gbc{h_} are defined. At this point \gbc{xneg}, \gbc{xscale}, etc.,
% have known values and \gbc{setztr} can define the transforms that are
% based on them. Also, the default \gbc{drawpen} is initialized and the
% boundary of the graph space is assigned to the clipping array.
%
% If \gbc{underlaylabels} is true, we try to make them part of the
% background, adding them to the picture variable \gbc{background_labels}.
% Just before shipout, the picture is placed on top.
%
% If \gbc{overlaylabels} is \gbc{true}, we try to make labels in \MP{}
% behave the same as labels in \TeX{} (for \mfpic) by adding the labels
% on last. We do this by adding them to the picture variable
% \gbc{foreground_labels} as they occur, then add that picture onto
% \gbc{active_plane} just before shipout. For backward compatibility,
% the default for \gbc{overlaylabels} is \gbc{false}.
%
% We initialize \gbc{foreground_labels} and \gbc{background_labels} here. The
% pair variables \gbc{labelbb.ll} and \gbc{labelbb.ur} keep track of the
% bounding box of added labels in case \gbc{overlaylabels},
% \gbc{truebbox}, and \gbc{clipall} are all \gbc{false}.
%    \begin{macrocode}
%<*MP>
boolean overlaylabels, underlaylabels, havebackground;
overlaylabels := false;
underlaylabels := false;
havebackground := false;

%</MP>
def initpic =
  setztr;
  resizedrawpen (penwd);
  if ClipOn:  ClipPath := 1;
    ClipPath1 := rect (origin, (w_, h_));
  fi
  if debug:
    GBdebug;
      >> "Drawing nominal bounding box around picture";
    GBenddebug;
    noclip ( safedraw rect (origin, (w_, h_)) );
  fi
%<*MP>
  newpicture (foreground_labels);
  newpicture (background_labels);
  havebackground := false;
  save labelbb;  pair labelbb.ll, labelbb.ur;
  labelbb.ll := labelbb.ur := origin;
%</MP>
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mfpicenv}
% We define a \gbc{mfpicenv} environment for compatibility with older
% \file{graphbase.mf} (mainly for \prog{fig2dev}'s \file{genmf.c}).
% \DescribeRoutine{endmfpicenv}
% Actually, I have no idea if \prog{fig2dev} even works with the current
% \mfpic.
%
% \DescribeRoutine{bounds}
% This used to be for compatibility also, but I decided it was a
% convenient abbreviation, so \mfpic{} uses it now.
%    \begin{macrocode}
def mfpicenv =  enddef;
def endmfpicenv =  enddef;
def bounds (expr a, b, c, d) =
  xneg := a;  xpos := b;
  yneg := c;  ypos := d;
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{beginmfpic}
% This is the figure wrapper.  \mfpic{} used to begin with figure 1 and
% progressively increment the number. The current value of \gbc{gcode} was
% always equal to the current figure number. Now, \mfpic{} explicitly
% writes the figure number, so we assign \gbc{gcode} to that number in
% case any old files made use of the current number through the
% \gbc{gcode} variable.
%
% Originally, \gbc{beginmfpic} defined \mfc{w}, \mfc{h} and \mfc{d}, but
% that caused problems if an \mfpic{} user tried to store a path in a
% variable named \gbc{h}, etc. So now we use the less obvious names ending
% in underscore. Apart from this, the code below is a clone of
% \file{plain.mf}'s \mfc{beginchar} (for \MF). In fact, it used to invoke
% \mfc{beginchar}. For \MP, we invoke \mfc{beginfig} explicitly. This does
% the \mfc{clear...} actions and \mfc{charcode} assignment.
%
% The `\mfc{extra_...mfpic}' strings provide a compiler-independent way
% to add to the extra beginning and ending tokens.
%    \begin{macrocode}
string extra_beginmfpic;  extra_beginmfpic := "";
string   extra_endmfpic;   extra_endmfpic := "";

def beginmfpic (expr ch) =
%<MP>  beginfig (ch);
%<MF>  begingroup
    gcode := ch;
    save w_, h_, d_;
    charwd := (xpos-xneg)*xscale*unitlen;
    charht := (ypos-yneg)*yscale*unitlen;
    chardp := 0;
%<*MF>
    charcode := if known ch:  byte ch  else:  0 fi;
    w_ := hround (charwd*hppp);
    h_ := vround (charht*hppp);
    d_ := vround (chardp*hppp);
    charic := 0;  clearxy;  clearit;  clearpen;
    scantokens extra_beginchar;
%</MF>
%<*MP>
    w_ := charwd;
    h_ := charht;
    d_ := chardp;
%</MP>
    initpic;
    scantokens extra_beginmfpic;
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{endmfpic}
% For \MF, we again clone \file{plain.mf}'s \mfc{endchar}, adding support
% for the \gbc{clipall} (clip to the graph rectangle), and \gbc{ClipOn}
% (clip to some user specified array of paths), and \gbc{showbbox} (draw
% the boundary of the graph for debugging purposes).
%    \begin{macrocode}
def endmfpic =
  scantokens extra_endmfpic;
  if debug:
    GBdebug;
%<MF>      >> "TFM charwd = " & decimal charwd & "pt#";
%<MF>      >> "TFM charht = " & decimal charht & "pt#";
%<MP>      >> "width = " & decimal w_ & "bp";
%<MP>      >> "height = " & decimal h_ & "bp";
    GBenddebug;
  fi
  DoClip (active_plane);
  if clipall:  clipto (active_plane) rect (origin, (w_, h_));  fi
  if showbbox:  noclip ( safedraw rect (origin, (w_, h_)) );  fi
%<*MF>
  scantokens extra_endchar;
  if proofing > 0:  makebox (proofrule);  fi
  chardx := w_;     % desired width of character in pixels
  shipit;
  if displaying > 0:  makebox (screenrule);  showit;  fi
  endgroup
%</MF>
%    \end{macrocode}
%
% \MP's code is more involved due to the possibility to put typeset text
% in a picture. In addition to the \gbc{clipall}, \gbc{ClipOn} and
% \gbc{showbbox} support, we have support for labels and \gbc{truebbox}.
%    \begin{macrocode}
%<*MP>
  save _ll, _ur;
  pair _ll, _ur;
  if truebbox:
    _ll := llcorner active_plane;
    _ur := urcorner active_plane;
%    \end{macrocode}
% We try to let the bbox include labels, even when they extend beyond the
% nominal picture boundaries. However, they will have been clipped off if
% \gbc{clipall} is set. In that case, we just set the bounding box to the
% coordinates determined by \gbc{w_} and \gbc{h_}, otherwise we expand
% them to the \gbc{labelbb} values.
%    \begin{macrocode}
  elseif clipall:
    _ll := origin;
    _ur := (w_,h_);
  else:             % expand to accomodate labels
    _ll := pairmin ((0,  0 ), labelbb.ll);
    _ur := pairmax ((w_, h_), labelbb.ur);
  fi
%    \end{macrocode}
% A bounding box in the output PostScript code can have a side with
% length 0 (e.g., a picture drawn with \mfpic{} that contains only
% text placed by \TeX). This can cause division by 0 errors in some
% cases. That's why we don't just let \MP{} determine the bounding box,
% but force the upper and lower coordinates to differ.
%    \begin{macrocode}
  _ur := pairmax (_ur, _ll + eps*(1, 1));
  setbounds active_plane to rect (_ll, _ur);
%    \end{macrocode}
% Finally, if \gbc{overlaylabels} or \gbc{underlaylabels} was true during
% a \gbc{newgblabel} command, then the label was not added to
% \mfc{currentpicture} but rather to \gbc{foreground_labels} or
% \gbc{background_labels}. We add those pictures now, the former on top of
% \mfc{currentpicture}, the latter underneath. This might extend the bbox
% calculated above, but that is one of the effects we \emph{want} to
% achieve. Picture variables can consume a lot of memory, so we clear
% each one after we have added it. Unfortunately, we will temporarily
% have two copies of the current picture in memory for background text, so
% we perform this operation only if \gbc{havebackground} is true.
%    \begin{macrocode}
    if havebackground:
      addto background_labels also active_plane;
      active_plane := background_labels;
      background_labels := nullpicture;
    fi
    addto active_plane also foreground_labels;
    foreground_labels := nullpicture;
  endfig;
%</MP>
enddef;

%    \end{macrocode}
%
%
% \section{Text}\label{text}
%
% In the \MP{} version, \gbc{label_adjust}, \gbc{label_sep} and
% \gbc{labelpath_sep} are the equivalent of \mfpic's \cs{tlabeloffset},
% \cs{tlpointsep} and \cs{tlpathsep}. In the \MF{} version they are still
% needed (in \gbc{textrect}, etc.) to place the paths that are to surround
% the text that \TeX{} places.
%
% \gbc{label_adjust} is a vector displacement applied to all labels,
% while \gbc{label_sep} is the distance from the label to
% the point of placement, when that point is on the edges of the label's
% bounding box. Both are in device coordinates (e.g., \mfc{3bp}).
% Finally, \gbc{labelpath_sep} is the separation of a surrounding path
% from the text.
%    \begin{macrocode}
pair label_adjust;
label_adjust := origin;
numeric label_sep, labelpath_sep ;
label_sep :=  0;  labelpath_sep := 0;
%    \end{macrocode}
%
% Another aspect of trying to make \mfpic's \file{.mp} and \file{.mf}
% the same, we here define a version of \mfc{verbatimtex} for \MF. This
% works only if \mfc{etex} is followed by a semicolon, and no semicolons
% appear in the \TeX{} material. (There may be other forbidden things, and
% certainly any parentheses have to be in matching pairs. Not so obvious
% is that \cs{begingroup} and \cs{endgroup} have to be balanced: \MF{}
% sees \cs{begingroup} as `\verb$\$' plus \mfc{begingroup}.)  We would
% like the output of \mfpic{} under the \opt{metapost} option to be usable
% in \MF{} with minimal changes.
%    \begin{macrocode}
%<MF>def verbatimtex text t =  enddef;

%    \end{macrocode}
%
% \subsection{Placement of text, \MP{} only}\label{placement}
%
% \DescribeRoutine{newgblabel}
% This is how \mfpic{} places labels when \opt{mplabels} is in effect.
% Since labels will typically be \mfc{btex...etex}, which are picture
% expressions, it will actually place any picture, \gbc{s}. If you feed it
% a string or path, it will convert it to a picture (with the \mfc{infont}
% operator or the \gbc{picpath} macro).
%
% The macro \gbc{newgblabel} takes 6 parameters. The first three
% parameters could easily be condensed into two if \mfpic{} support were
% all that was required, however I thought it best to make it general.
% The parameters \gbc{hf} and \gbc{vf} are numeric, with \gbc{hf}
% representing the fraction of the text that lies left of the point where
% the text is placed and \gbc{vf} represents the fraction of
% text that lies below that point. However, if the third parameter is
% \mfc{true}, then \gbc{vf} is relative to the baseline (i.e., the depth
% is ignored). In \mfpic{} this is only used with \gbc{vf = 0} to get
% placement on the baseline.
%
% These three parameters correspond to the optional parameter of
% \cs{tlabel} in \mfpic{} as follows:
% \begin{itemize}
% \item \gbc{hf} determines horizontal position: $0=\mathtt{l}$,
%   $.5=\mathtt{c}$, and  $1 = \mathtt{r}$.
% \item \gbc{vf} and \gbc{BL} determine vertical position. For placement
%   option \texttt{B}, $\mathtt{vf} = 0$ and \gbc{BL} is \mfc{true}. For the
%   rest, \gbc{BL} is \mfc{false} and \gbc{vf} corresponds as follows:
%   $0 = \mathtt{b}$, $.5 = \mathtt{c}$ and $1 = \mathtt{t}$.
% \end{itemize}
% The remaining parameters have the following meanings:
% \begin{itemize}
% \item \gbc{r} is degrees of rotation about the specified point.
% \item \gbc{s} is a string or picture expression (typically
% \mfc{btex ... etex} code)
% \item \gbc{pts} is a list of pairs in graph coordinates.
% \end{itemize}
% First the bounding box of the picture is determined using
% \gbc{pathdims}. (Why \texttt{\textit{path}dims}? Because it was written
% for the paths that surround text, and was then incorporated into text
% placement when \gbc{newgblabel} replace \gbc{gblabel}.) Then
% \gbc{readjustdims} extends that box by \gbc{label_sep}, a new
% reference point for the picture is calculated using
% \DescribeRoutine{ref_shift}
% \gbc{ref_shift}, and then \gbc{thegblabel} rotates it around the
% reference point and adds the \gbc{label_adjust}. Finally, for each
% \gbc{_itm} in \gbc{pts}, the result is shifted by \gbc{_itm}. If
% \gbc{overlaylabels} is true, the label is placed on the picture
% \gbc{foreground_labels} and added to \gbc{active_plane} at
% \gbc{endmfpic}. If \gbc{underlaylabels} is true, it is placed in picture
% \gbc{background_labels} and \gbc{active_plane} is placed on top of it.
% Otherwise, it is added directly to \gbc{active_plane} and the
% \gbc{labelbb} variables are adjusted.
%
% We also use \gbc{ref_shift} in \MF{} since the curves that surround text
% require it.
%
% \DescribeRoutine{gblabel}
% We keep \gbc{gblabel} for backward compatibility with old \mfpic{}
% files, but it merely calls \gbc{newgblabel}. While the old \gbc{gblabel}
% had the same flexibility as \gbc{newgblabel}, this one assumes that the
% parameters are only those that \mfpic{} would write.
%
% We provide a null definition of newgblabel for \MF{} to allow \mfpic's
% \file{.mp} files to be somewhat usable with minimal changes. It
% requires a text parameter, since \MF{} would be unable to evaluate
% \mfc{btex} expressions.
%    \begin{macrocode}
%<*MP>
vardef newgblabel (expr hf, vf, BL, r) (expr s) (text pts) =
  save _lab, _ll, _ur;  picture _lab;  pair _ll, _ur;
  _lab := makepicture (s);
  pathdims (origin, _lab) (_ll, _ur);
  readjustdims (_ll, _ur) (label_sep);
  _lab := thegblabel (ref_shift (hf, vf, BL, _ll, _ur), r, _lab);
  save _b;  pair _b;
  for _itm = pts:
    _b := zconv (_itm);
    if overlaylabels:
      addto foreground_labels also _lab shifted _b _wc_ tlabelcolor;
    elseif underlaylabels:
      addto background_labels also _lab shifted _b _wc_ tlabelcolor;
      havebackground := true;
    else:
      addto active_plane also _lab shifted _b _wc_ tlabelcolor;
      labelbb.ll := pairmin (_b + llcorner _lab, labelbb.ll);
      labelbb.ur := pairmax (_b + urcorner _lab, labelbb.ur);
    fi
  endfor
%</MP>
%<MF>vardef newgblabel (expr hf, vf, BL, r) (text s) (text pts) =
enddef;

% Assumes a+b=1 and either c+d=1 or c=d=0:
%<MP>vardef gblabel (expr a, b, c, d, r) (expr s) (text t) =
%<MF>vardef gblabel (expr a, b, c, d, r) (text s) (text t) =
  newgblabel (b, d, (c = 0) and (d = 0), r) (s) (t);
enddef;

vardef ref_shift (expr hf, vf, BL, ll, ur) =
  - ( (hf)[xpart ll, xpart ur],
        (vf)[if BL:  0  else:  (ypart ll) fi, ypart ur] )
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{thegblabel}
% When \gbc{thegblabel} is called by the above, \gbc{p} is a text picture,
% but it is also called by the \gbc{textrect}, etc., in which case \gbc{p}
% is a path. This is why it is needed in the \MF{} version.
%    \begin{macrocode}
vardef thegblabel (expr z, r, p) =
  ((p shifted z) rotated r) shifted label_adjust
enddef;

%    \end{macrocode}
%
% \subsection{Decorating the text, \MF{} or \MP{}}\label{decorating}
%
% The three macros \gbc{textrect}, \gbc{textoval} and \gbc{textellipse}
% are designed to surround a bit of text with some curve. These macros
% return the path in graph coordinates. In
% \DescribeRoutine{textrect}
% \gbc{textrect}, the path is a rectangle with optionally rounded corners.
% The second parameter, \gbc{rad}, is the radius of quarter circles at the
% corners (in device units). In the other two cases, the path is an
% ellipse. They differ in the meaning of the second parameter.
%
% \DescribeRoutine{textoval}
% In \gbc{textoval}, the second parameter \emph{multiplies} the ratio of
% width to height of the text to produce the ratio for the ellipse. Thus,
% with \gbc{mult}=1, the ratio will be the same as that of the text. In
% \DescribeRoutine{textellipse}
% \gbc{textellipse}, the second parameter \gbc{rat} is the actual value of
% the ratio of width to height of the ellipse and a value of 1 produces a
% circle. In either macro, if that parameter is 0, we draw a rectangle.
%
% The size of each path is determined so that, when the text is placed and
% the path drawn, it passes through the four corners of the following
% rectangle: the rectangle which just encloses the text plus the amount of
% space on all sides determined by \gbc{labelpath_sep}. Note that this means
% a rectangle with rounded corners will have larger height and width than
% one without. These versions always center the surrounding path on the
% the point \gbc{loc}. The extended versions (below) have the same
% flexibility of placement as the commands that place the label being
% surrounded.
%
% The first parameter \gbc{lbl} is either a pair representing the
% height and width of the text (only possibility in \MF) or the actual
% text. These macros are being kept for backward compatibity, but now they
% call the extended versions that allow the path to follow arbitrary
% text placement. The parameters \gbc{(.5,.5,false,0)} were those
% assumed in the past version: centered at the point, with no rotation.
%
% The extended versions of \gbc{textoval} and \gbc{textellipse} are both
% now implemented in a single command \gbc{xellipse}, with a boolean to
% specify whether the aspect ratio of the text is used to calculate the
% aspect of the ellipse.
%    \begin{macrocode}
vardef textrect (expr lbl, rad, loc) =
  textrectx (.5, .5, false, 0) (origin, lbl, rad, loc)
enddef;
vardef textoval (expr lbl, mult, loc) =
  xellipse (true, .5, .5, false, 0) (origin, lbl, mult, loc)
enddef;
vardef textellipse (expr lbl, rat, loc) =
  xellipse (false, .5, .5, false, 0) (origin, lbl, rat, loc)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{textrectx}
% Macro \gbc{textrectx} is the extended version of \gbc{textrect} which
% allows the same adjustments to the rectangle that we can apply to the
% text it surrounds (via \gbc{newgblabel}). In fact, it calculates the
% position in exactly the same manner as that macro, and the first 4
% parameters encode that position in the same way.
%
% The placement of each path is: shifted and rotated by the same amount
% as the text (by \gbc{ref_shift}) according to the first four parameters,
% then shifted to the point given in the third parameter \gbc{loc}, and
% finally shifted by the vector specified in \gbc{label_adjust}.
%
% \gbc{lbl} is either the upper right corner of the text or the label
% itself. In the first case \gbc{xy} is the lower left corner, in the
% second case it is a dummy parameter, the bounding box being obtained (in
% \gbc{pathdims}) by measuring the label. For these extended macros, the
% parameters \gbc{lbl}, \gbc{mult}, \gbc{rad}, and \gbc{loc} are as in
% the unextended versions.
%
% \DescribeVariable{roundends}
% The variable \gbc{roundends} is a boolean. We really only need it to be
% a type distinguishable from any numeric value. \Mfpic{} users can
% specify it rather than an explicit radius, and when the code of
% \gbc{textrectx} detects this, it uses the maximum radius for the corners
% (making the short side of the `rectangle' a semicircle).  That is, if
% \gbc{rad} is a boolean (and \mfc{true}) then the radius at the corners
% is so chosen. If \gbc{rad} is \mfc{false} the corners are not rounded at
% all.
%    \begin{macrocode}
boolean roundends;  roundends := true;
vardef textrectx (expr a, b, c, rot, xy, lbl, rad, loc) =
  save ll, ur, _r, f, zz;
  pair ll, ur, zz;  path f;
  pathdims (xy, lbl) (ll, ur);
  readjustdims (ll, ur) (labelpath_sep)
  _r := if numeric rad:  rad
    elseif not boolean rad:  0
    elseif rad:  emin (xpart(ur-ll), ypart (ur-ll))/sqrt(2)
    else:  0
    fi;
  if _r = 0:
    f :=  rect (ll, ur);
  else:
    save p, q;
    pair p[];  path q;
    p1 := ur - _r*dir(45);
    p3 := ll + _r*dir(45);
    p2 := (xpart p3, ypart p1);
    p4 := (xpart p1, ypart p3);
%    \end{macrocode}
% We allow the rounding radius to be negative and make the corners
% indented in that case. We no longer reverse the path in this case.
%    \begin{macrocode}
    q := if _r < 0:  reverse  fi  quartercircle scaled 2_r;
    f :=
        (q shifted p1)--(q rotated 90 shifted p2)
             --(q rotated 180 shifted p3)
             --(q rotated -90 shifted p4)--cycle;

  fi
  readjustdims (ll, ur) (label_sep - labelpath_sep);
  invvconv (thegblabel (ref_shift(a, b, c, ll, ur), rot, f)) shifted loc
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{textellipsex}
% The macro \gbc{textellipsex} is a simlar extension for
% \gbc{textellipse}. It and the related macro
% \DescribeRoutine{textovalx}
% \gbc{textovalx} now call a common macro with different values of a
% boolean parameter.
%    \begin{macrocode}
def textovalx =  xellipse (true)  enddef;
def textellipsex =  xellipse (false)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xellipse}
% In \gbc{xellipse}, \gbc{aa} and \gbc{bb} are the horizontal and
% vertical radii of the resulting ellipse, while \gbc{ww} and \gbc{hh}
% are half the width and height size of the text. If the boolean
% \gbc{aspect} is true, the aspect ratio of the ellipse (i.e., \gbc{aa/bb})
% equals \gbc{mult*ww/hh}, otherwise it equals \gbc{mult}.
%    \begin{macrocode}
vardef xellipse (expr aspect, a, b, c, r, xy, lbl, mult, loc) =
  if mult = 0:
    textrectx (a, b, c, r) (xy, lbl, 0, loc)
  else:
    save ll, ur, cc, ww, hh, f;
    pair ll, ur, cc;  path f;
    pathdims (xy, lbl) (ll, ur);
    readjustdims (ll, ur) (labelpath_sep)
    cc := .5[ll, ur];
    (ww, hh) = ur - cc;
    if (ww = 0) or (hh = 0):
      f = (ll--ur);
    else:
      save aa, bb;
%    \end{macrocode}
%
% The \gbc{aa} and \gbc{bb} are now calculated in a way that decreases the
% chance of overflow. As a side effect, negative \gbc{mult} no longer
% reverses the path.
%    \begin{macrocode}
      aa := ww ++ if aspect:  ww  else:  hh  fi  *mult;
      bb := hh ++ if aspect:  hh  else:  ww  fi  /mult;
      f := ellipse (cc, aa, bb, 0);
    fi
    readjustdims (ll, ur) (label_sep - labelpath_sep);
    invvconv (thegblabel (ref_shift(a, b, c, ll, ur), r, f)) shifted loc
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{pathdims}
% This has been changed to make the code of \mfpic{} a bit simpler and
% to aid in backward compatibility. It takes a couple of pairs (the actual
% or nominal label bounding box corners) or something visible (picture,
% string or path) and assigns suitable values to \gbc{ll} and \gbc{ur}.
%
% \DescribeRoutine{readjustdims}
% This is used to add the separations needed to implement the effects of
% \gbc{label_sep} and \gbc{labelpath_sep}.
%    \begin{macrocode}
def pathdims (expr xy, lbl) (suffix ll, ur) =
  if pair lbl:
    ll := xy;  ur := lbl;
  else:
%<MF>    ll := ur := origin;
%<*MP>
    setpicture (_lbl) makepicture (lbl);
    ll := llcorner _lbl;
    ur := urcorner _lbl;
%</MP>
  fi
enddef;

def readjustdims (suffix ll, ur) (expr s) =
  ll := ll - s*(1,1);
  ur := ur + s*(1,1);
enddef;

%    \end{macrocode}
%
%
% \section{Additional Functions}\label{functions}
%
% Complex variable functions are provided, which interpret a pair $(x, y)$
% as the complex number $z = x + iy$. We also provide for the use of
% radians, add the standard exponential and logarithms, and add the
% hyperbolic functions and their inverses.
%
% Normally \mfc{infinity = 2**12 - epsilon} is the largest number allowed
% (as a value involved in actual drawing in \MF). Since we set
% \mfc{warningcheck=0}, values not assigned to a variable and not
% written to the \file{.tfm} file (and any value in \MP) can be as high as
% \mfc{2**15 - epsilon}, which is a speck smaller than \mfc{1/(2epsilon)}.
% So \gbc{reallysmall} is the smallest number whose reciprocal is a
% usable number. (\mfc{epsilon} is the smallest possible positive number
% in \MF.)
%
% The value \gbc{eps/2 + epsilon} is the smallest value with
% reciprocal less than \mfc{infinity}. I set \gbc{nottoosmall} a speck
% bigger to ensure that the same is true of \gbc{2*(nottoosmall/2)}.
%
% We set \gbc{secd x = 1/(cosd x)} unless \gbc{cosd x} is less than
% \gbc{reallysmall}, then we set it equal to \gbc{1/reallysmall}. We do a
% similar thing with \gbc{cscd}. (When such a substitution happens
% \DescribeRoutine{TruncateWarn}
% \gbc{TruncateWarn} prints a message that a truncation has taken place.)
%
% Why not just determine what number will produce arithmetic overflow and
% test for that?  Because I'm lazy: it would require a different number
% for each of the functions. Instead, since \MF{} has no `arithmetic
% underflow', I compute something that is guaranteed to work and occurs
% in the formula for the function as a reciprocal (e.g., $t = e^{-|x|}$
% for \gbc{cosh x}) and make sure the number is not too small to take its
% reciprocal.
%
% \DescribeRoutine{signof}
% This expands to a minus sign if its argument is negative, otherwise
% nothing.
%    \begin{macrocode}
newinternal reallysmall;  reallysmall := 3epsilon;
newinternal nottoosmall;  nottoosmall := eps/2 + 2epsilon;
def signof (expr X) =  if X < 0:  -  fi  enddef;
def TruncateWarn expr s =
  GBwarn s & " is too large or undefined, so it will be truncated.";
enddef;

%    \end{macrocode}
%
% In addition to \mfc{sind} and \mfc{cosd} which take angles in degrees,
% we define the remaining trig functions \gbc{tand}, \gbc{cotd},
% \gbc{secd}, and \gbc{cscd}.
%
% We define \RoutineIndex{secd}\gbc{secd}, one of the simplest, to include
% an out of range test (which also prevents division by 0). Then
% \RoutineIndex{tand}\gbc{tand} can make use of it without any division.
% We do the same with \RoutineIndex{cscd}\gbc{cscd} and
% \RoutineIndex{cotd}\gbc{cotd}.
%    \begin{macrocode}
vardef secd primary X =
  setnumeric (temp) cosd(X);
  if abs(temp) < reallysmall:
    TruncateWarn "Secant or Tangent";
    temp := signof (temp) reallysmall;
  fi
  1/temp
enddef;
vardef tand primary X =  sind(X)*secd(X)  enddef;

vardef cscd primary X =
  setnumeric (temp) sind(X);
  if abs(temp) < reallysmall:
    TruncateWarn "Cosecant or Cotangent";
    temp := signof(temp) reallysmall;
  fi
  1/temp
enddef;
vardef cotd primary X =  cosd(X)*cscd(X)  enddef;

%    \end{macrocode}
%
% These are the inverse functions, which return an angle in degrees:
% \RoutineIndex{acos}\gbc{acos}, \RoutineIndex{asin}\gbc{asin} and
% \RoutineIndex{atan}\gbc{atan}.
%    \begin{macrocode}
vardef acos primary X =
  if abs X > 1:
    TruncateWarn "Argument of arccosine";
    angle (signof(X) 1, 0)
  else:
    angle (X, 1 +-+ X)
  fi
enddef;
vardef asin primary X =
  if abs X > 1:
    TruncateWarn "Argument of arcsine";
    angle (0, signof(X) 1)
  else:
    angle (1 +-+ X, X)
  fi
enddef;
vardef atan primary X =  angle (1, X)  enddef;

%    \end{macrocode}
%
% Now the trig functions that take angles in radians:
% \RoutineIndex{sin}\gbc{sin}, \RoutineIndex{cos}\gbc{cos},
% \RoutineIndex{tan}\gbc{tan}, \RoutineIndex{cot}\gbc{cot},
% \RoutineIndex{sec}\gbc{sec} and \RoutineIndex{csc}\gbc{csc}.
%    \begin{macrocode}
vardef sin primary X =  sind (X*radian)  enddef;
vardef cos primary X =  cosd (X*radian)  enddef;
vardef tan primary X =  tand (X*radian)  enddef;
vardef cot primary X =  cotd (X*radian)  enddef;
vardef sec primary X =  secd (X*radian)  enddef;
vardef csc primary X =  cscd (X*radian)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{degrees}
% It is useful to have a command to convert from radians to degrees and
% one to
% \DescribeRoutine{radians}
% convert from degrees to radians. For example, \gbc{degrees(pi)} produces
% (approximately) $180$ and \gbc{radians(180)} is approximately $\pi$.
%    \begin{macrocode}
vardef degrees (expr t) =  t*radian  enddef;
vardef radians (expr t) =  t/radian  enddef;

%    \end{macrocode}
%
% And the inverses (\RoutineIndex{invsin}\gbc{invsin},
% \RoutineIndex{invcos}\gbc{invcos} and \RoutineIndex{invtan}\gbc{invtan})
% that return angles in radians.
%    \begin{macrocode}
vardef invcos primary X =  (acos X)/radian  enddef;
vardef invsin primary X =  (asin X)/radian  enddef;
vardef invtan primary X =  (atan X)/radian  enddef;

%    \end{macrocode}
%
% Here we define the standard exponential function
% \RoutineIndex{exp}\gbc{exp}. (The \MF{} function \mfc{mexp} has the
% unusual base $e^{1/256}$ to avoid overflow.) The inverse of \gbc{exp} is
% the natural logarithm (\RoutineIndex{ln}\gbc{ln} or
% \RoutineIndex{log}\gbc{log}). We also have the general base logarithm
% \RoutineIndex{logbase}\gbc{logbase} and its two special instances
% \RoutineIndex{logtwo} \gbc{logtwo} and \RoutineIndex{logten}\gbc{logten}.
%    \begin{macrocode}
vardef exp primary X =  mexp (256 * X)  enddef;
vardef ln  primary X =  (mlog X) / 256  enddef;
def log =  ln  enddef;
vardef logbase (expr B) primary X =  (mlog X)/(mlog B)  enddef;
def logtwo =  logbase( 2)  enddef;
def logten =  logbase(10)  enddef;

%    \end{macrocode}
%
% The hyperbolic functions: \RoutineIndex{cosh}\gbc{cosh}
% \RoutineIndex{sinh}\gbc{sinh}, \RoutineIndex{tanh}\gbc{tanh},
% \RoutineIndex{sech}\gbc{sech}, \RoutineIndex{csch}\gbc{csch} and
% \RoutineIndex{coth}\gbc{coth}.
%    \begin{macrocode}
vardef cosh primary X =
  setnumeric (temp) 2 exp (-abs(X));
  if temp < reallysmall:
    TruncateWarn "Cosh";
    temp := reallysmall;
  fi
  1/temp + temp/4
enddef;

vardef sinh primary X =
  setnumeric (temp) 2 exp (-abs(X));
  if temp < reallysmall:
    TruncateWarn "Sinh";
    temp := reallysmall;
  fi
  signof (X) (1/temp - temp/4)
enddef;

vardef sech primary X =
  setnumeric (temp) exp(-(abs (X)));
  2temp/(1 + temp*temp)
enddef;

vardef tanh primary X =
  setnumeric (temp) exp(-2(abs (X)));
  signof (X) (1 - temp)/(1 + temp)
enddef;

vardef csch primary X =
  save temp, tempa;  temp := exp(-(abs (X)));
  tempa := (1 - temp*temp)/2;
  if tempa < reallysmall:
    TruncateWarn "Csch";
    tempa := reallysmall;
  fi
  signof (X) temp / tempa
enddef;

vardef coth primary X =
  setnumeric (temp) tanh(X);
  if abs(temp) < reallysmall:
    TruncateWarn "Coth";
    temp := signof (X) reallysmall;
  fi
  1/temp
enddef;

%    \end{macrocode}
%
% The inverses of some of the hyperbolic functions:
% \RoutineIndex{acosh}\gbc{acosh}, \RoutineIndex{asinh}\gbc{asinh} and
% \RoutineIndex{atanh}\gbc{atanh}.
%    \begin{macrocode}
vardef acosh primary y =
  if y < 1:
    TruncateWarn "acosh";
    0
  else:
    ln (y + (y +-+ 1))
  fi
enddef;

vardef asinh primary y =  ln (y + (y ++ 1))  enddef;

vardef atanh primary y =
  if abs (y) < 1:
    (ln (1 + y) - ln (1 - y))/2
  else:
    TruncateWarn "atanh";
    signof (y) infinity
  fi
enddef;

%    \end{macrocode}
%
% \CMF's pair variables are a decent replacement for complex variables.
% These give some of the more basic functions of standard complex
% analysis: \RoutineIndex{Arg}\gbc{Arg}, \RoutineIndex{Log}\gbc{Log},
% \RoutineIndex{cis}\gbc{cis}, \RoutineIndex{zexp}\gbc{zexp},
% \RoutineIndex{sgn}\gbc{sgn}, and \RoutineIndex{conj}\gbc{conj}.
%    \begin{macrocode}
vardef Arg primary Z =  (angle Z)/radian  enddef;
vardef Log primary Z =  (ln (abs Z), Arg Z)  enddef;
vardef cis primary T =  dir (T*radian)  enddef;
vardef zexp primary Z =  (exp (xpart Z)) * cis (ypart Z)  enddef;
vardef sgn primary Z =  if not (Z = origin):  unitvector  fi  Z  enddef;
vardef conj primary Z = (xpart Z, -ypart Z) enddef;

%    \end{macrocode}
%
% \DescribeRoutine{Moebius}
% A less basic operation: the Moebius shift which takes the disk $|z| <
% 1$ onto itself. It is a hyperbolic geometry analog of shifting points
% in Euclidean geometry. Its mathematical definition (all variables are
% complex numbers):
% \[
%   M_a(z) = \frac{z + a}{1 - \bar az}
% \]
% \DescribeRoutine{pshdist}
% Related to \gbc{Moebius} is the pseudohyperbolic metric. The distance
% between $z$ and $w$ in this metric is $|z-w|/|1 - \bar wz|$.
%    \begin{macrocode}
vardef Moebius (expr A) primary Z =
  save _D; pair _D;
  _D := (1, 0) + (Z zscaled (conj A));
  (Z + A)/(abs _D) rotated (- angle _D)
enddef;
vardef pshdist (expr Z,W) =  abs(Moebius(-W)(Z))  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{polar}
% \gbc{polar} converts a polar coordinate pair $(r, \theta)$ to the
% corresponding rectangular coordinate pair.
% \DescribeRoutine{id}
% \gbc{id} returns its argument, which can be any expression of any type.
%    \begin{macrocode}
vardef polar primary p =  (xpart p) * dir (ypart p)  enddef;
def id (expr x) =  x  enddef;

%    \end{macrocode}
%
% The definition of powers (\gbc{x**y}) in \prog{plain} \MF{} and \MP{}
% could be more accurate. In particular \gbc{x=2**10} ought to be an
% integer (that is, satisfy \mfc{x=floor x}). Here we redefine
% \prog{plain}'s \mfc{**}, intercepting the case of a positive integer
% power of an integer.
%    \begin{macrocode}
primarydef x**y =
  if y=2:  x*x
  elseif (x = floor x) and (abs y = floor y):
    1 for n=1 upto y:  *x endfor
  else:  takepower y of x
  fi
enddef;
let ^ = **;

%    \end{macrocode}
%
% \section{Coordinate Systems and Transformations}\label{systems}
%
% \DescribeVariable{T_stack}
% We want to define a localization of the current transform. To do
% this we define a LIFO stack of transforms \gbc{T_stack[\,]}, and a pair of
% macros.
% \DescribeRoutine{T_push}
% \gbc{T_push} puts its argument (a transform) on the stack, and
% \DescribeRoutine{T_pop}
% \gbc{T_pop} pops it off into its argument (a transform variable name).
% We also define two localizing macros
% \DescribeRoutine{bcoords}
% \gbc{bcoords} that pushes our \gbc{ztr}
% on the stack, and
% \DescribeRoutine{ecoords}
% \gbc{ecoords} that pops it off. We no longer put \gbc{vtr} on the stack,
% since we can recalculate it whenever \gbc{ztr} is changed. \gbc{apply_t}
% always did this, now \gbc{ecoords} does so as well.
%    \begin{macrocode}
transform T_stack[];
numeric T_stack;  T_stack := 0;

def T_push (expr T) =  T_stack[incr T_stack] := T;  enddef;
def T_pop (suffix $) =
  if T_stack > 0:
    $ := T_stack[T_stack];
    T_stack := T_stack - 1;
  fi
enddef;

def bcoords =  hide ( T_push (ztr) )  enddef;
def ecoords =  hide ( T_pop (ztr);  vtr := vectorpart ztr )  enddef;

%    \end{macrocode}
%
% \subsection{Coordinate changes}\label{changes}
%
% \DescribeRoutine{apply_t}
% Here we define a mechanism for changing \gbc{ztr} and \gbc{vtr} by
% composing them with a new transform. Since a transform can be any affine
% transform, we get \gbc{ztr} by composing with the transform, but we
% calculate \gbc{vtr} from \gbc{ztr} by arranging that \mfc{origin
% transformed vtr} is \mfc{origin}. The syntax is \gbc{apply_t(rotated
% theta)} or \gbc{apply_t(transformed T)} if \mfc{T} is a variable or
% expression of type transform. Thus the argument of \gbc{apply_t} is a
% phrase which, were it to follow a path, would produce a transformed
% path. Knuth calls such a phrase a \emph{transformer}.
%    \begin{macrocode}
vardef vectorpart primary T =  T shifted -(origin transformed T)  enddef;

def apply_t (text Transformer) =
  ztr := identity Transformer transformed ztr;
  vtr := vectorpart ztr;
enddef;

%    \end{macrocode}
%
% And now we define some available transformers.
% \RoutineIndex{xslant}\gbc{xslant}, \RoutineIndex{yslant}\gbc{yslant},
% \RoutineIndex{zslant}\gbc{zslant}, \RoutineIndex{xyswap}\gbc{xyswap} and
% \RoutineIndex{boost}\gbc{boost}. The only two that need comment are
% \gbc{zslant} and \gbc{boost}. I know that boost comes from special
% relativity, but I have no idea why zslant is a `slant'.
%    \begin{macrocode}
def xslant =  slanted  enddef;  % (x+sy, y).
def yslant primary s =  % (x, y+sx).
  transformed
    begingroup
      save T;  transform T;
      origin transformed T = origin;
      (1, 0) transformed T = (1, s);
      (0, 1) transformed T = (0, 1);
      T
    endgroup
enddef;

def zslant primary p =  % (xu+yv, xv+yu), where p =  (u, v).
  transformed
    begingroup
      save T;  transform T;
      xpart T = ypart T = 0;
      xxpart T = yypart T = xpart p;
      xypart T = yxpart T = ypart p;
      T
    endgroup
enddef;

def xyswap =  zslant (0, 1)  enddef;
def boost primary X =  zslant (cosh X, sinh X)  enddef;

%    \end{macrocode}
%
% \subsection{Path transformation}\label{transformation}
%
% These are functions that accept and return a path in graph coordinates.
% For the most part they are named and defined to apply a similarly named
% transform to the path and return the result. There are two exceptions.
% When we draw things, we expect that rotated and reflected objects appear
% congruent to the originals. If we define a path in graph coordinates,
% and the $x$ and $y$ directions are scaled differently, then simply
% rotating the graph coordinates will distort angles. The same is true of
% reflection. Therefore, we apply \gbc{vtr} (so we are in drawing
% coordinates) then rotate or reflect, then apply \gbc{inverse vtr}. This
% may be a mistake, or perhaps we should do it for all of these. For now,
% I'm sticking with the scheme I inherited. One can always use
% \gbc{coords} and \gbc{apply_t} if one wants the difference in scales
% ignored.
%
% \DescribeRoutine{transformedpath}
% This is a vardef that reads an undelimited path expression and returns
% the path transformed by the text argument. All the others run this,
% allowing it to grab the path expression.
%
% \DescribeRoutine{rotatedpath}
% This returns the path rotated around point \gbc{p} by angle
% \gbc{th} in degrees.
%
% \DescribeRoutine{reflectedpath}
% This reflects the path through the line containing points \gbc{p} and
% \gbc{q}.
%
% \DescribeRoutine{scaledpath}
% This returns the path scaled so that distances from the point
% \gbc{p} are multiplied by \gbc{s}.
% \DescribeRoutine{xscaledpath}
% \gbc{xscaledpath} is similar, but only the horizontal distances from
% the line $x={}$\gbc{a} are multiplied by \gbc{s}. And with
% \DescribeRoutine{yscaledpath}
% \gbc{yscaledpath} the vertical distances from the line $y={}$\gbc{b} are
% multiplied by \gbc{s}.
%
% \DescribeRoutine{xslantedpath}
% The macro \gbc{xslantedpath} returns the path xslanted with line
% $y = {}$\gbc{b} being the pivot rather than the $x$-axis.
% \DescribeRoutine{slantedpath}
% The command \gbc{slantedpath} is just an alias for \gbc{xslantedpath},
% while
% \DescribeRoutine{yslantedpath}
% \gbc{yslantedpath} is the vertical version, yslanted with line $x =
% {}$\gbc{a} being the pivot rather than the $y$-axis.
%
% \DescribeRoutine{shiftedpath}
% This returns the path shifted by the vector (pair) \gbc{v}.
%
% \DescribeRoutine{xyswappedpath}
% The command \gbc{xyswappedpath} returns the path in which all points
% have had the coordinates exchanged $(a, b) \to (b, a)$. Note that this
% is not the same as \gbc{reflectedpath ((0,0), (1,1))}, as it performs
% the reflection in graph coordinates, as its name implies. If \gbc{vtr}
% has not been changed (by \gbc{apply_t}) then \gbc{xyswappedpath} will
% convert vertical lines to horizontal and vice versa. The
% \gbc{reflectedpath} version will not when $x$ and $y$ are scaled
% differently, for then the line \gbc{(0,0)--(1,1)} is not at a 45 degree
% angle in device coordinates where drawing takes place.
%
%    \begin{macrocode}
vardef transformedpath (text Transformer) expr f =  f Transformer
enddef;

def rotatedpath (expr p, th) =
  transformedpath (
    transformed vtr
    rotatedaround (p transformed vtr, th)
    transformed (inverse vtr)
  )
enddef;
def reflectedpath (expr p, q) =
  transformedpath (
    transformed vtr
    reflectedabout (p transformed vtr, q transformed vtr)
    transformed (inverse vtr)
  )
enddef;

def scaledpath (expr p, s) =
  transformedpath (shifted -p scaled s shifted p)
enddef;
def xscaledpath (expr a, s) =
  transformedpath (shifted (-a, 0) xscaled s shifted (a, 0))
enddef;
def yscaledpath (expr b, s) =
  transformedpath (shifted (0, -b) yscaled s shifted (0, b))
enddef;

def slantedpath =  xslantedpath  enddef;
def xslantedpath (expr b, s) =
  transformedpath (shifted (0, -b) slanted s shifted (0, b))
enddef;
def yslantedpath (expr a, s) =
  transformedpath (shifted (-a, 0) yslant s shifted (0, a))
enddef;

def shiftedpath (expr v) =  transformedpath (shifted v)  enddef;

def xyswappedpath =  transformedpath (xyswap)  enddef;

%    \end{macrocode}
%
% It seems odd, in retrospect, that we got by with a user interface that
% didn't include any subpath operations. But recently a user asked for the
% ability to add an arrowhead to the \emph{middle} of a path, and it
% seemed best to provide a subpath and use existing commands to add an
% arrowhead on its end.
%
% \DescribeRoutine{partialpath}
% The \gbc{partialpath} macro takes two fractions $\alpha$ and $\beta$
% between 0 and 1, and a path \gbc{f}, and returns the subpath from
% $\alpha * {} $\meta{length of \gbc{f}} to $\beta * {}$\meta{length of
% \gbc{f}} of \gbc{f}. Since the \gbc{gettime} routine was written to
% find the times for an increasing sequence of lengths, it was optimized
% to save the index of the previous length and begin from there. Thus it
% is more efficient to find the smaller of \gbc{a} and \gbc{b} first.
%
% Since running \gbc{gettime} would be a very inefficient way to get the
% first or last point of a path we skip that if either fraction is $0$ or
% $1$ (a common use is to get the first or last half of a path). We also
% skip finding the second time if \gbc{a = b} (an unlikely choice, but
% legal).
%
% \DescribeRoutine{gsubpath}
% \gbc{gsubpath} is the same as \MF's subpath primitive, but follows the
% prefix macro syntax of accepting a path expression (rather than a
% primary) and wrapping the result in a \mfc{vardef}.
%    \begin{macrocode}
vardef partialpath (expr a, b) expr f =
  save flag, flo, fhi, lo, hi, n;
  boolean flag;  flag = true;
  convertpath (g) f;
  n := length f;

  flo := snapto emin(a,b);
  if flo = 0:
    lo := 0;
  elseif flo < 1:
    setuplengtharray (cum, tot, idx) g;
    flag := false;
    lo := gettime (cum, idx) (flo*tot);
  else:
    lo := n;
  fi

  fhi := snapto emax (a,b);
  if flo = fhi:
    hi := lo;
  elseif fhi < 1:
    if flag:  setuplengtharray (cum, tot, idx) g;  fi
    hi := gettime (cum, idx) (fhi*tot);
  else:
    hi := n;
  fi

  if a > b:  reverse  fi  subpath (lo, hi) of f
enddef;

vardef gsubpath (expr a, b) expr f =  subpath (a, b) of f  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{setuplengtharray}
% This does the frequently repeated saving, rescaling and initializing
% for those commands that need to convert distance along a path to the
% corresponding time or point. A path variable should follow, but that
% is picked up by the \gbc{makelengtharry} at the end.
%    \begin{macrocode}
def setuplengtharray (suffix cum, tot, idx) =
  save cum, tot, idx;  idx := 0;  tot := makelengtharray (cum)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{pathtime}
% \gbc{pathtime} returns the time \mfc{t} such that \mfc{point t of p} is
% \gbc{frac} of the distance along \gbc{p} from the start, and
% \DescribeRoutine{pathpoint}
% \gbc{pathpoint} returns the point itself. Because the \gbc{gettime}
% routine requires it anyway, we truncate \gbc{frac} to the interval
% $[0,1]$ and avoid calling that rather lengthy function at $0$ and $1$.
%
% The path in pathtime should be in device coordinates, whereas the
% user-level command \gbc{pathpoint} expects it in graph coordinates.
% In fact, since \gbc{pathpoint} would most likely be used in the
% argument of some figure macro in \mfpic{}, it would require a
% previously stored path, so we make the path a suffix parameter.
%    \begin{macrocode}
vardef pathtime@# (suffix p) =
  if @# <= 0:  0
  elseif @# >= 1:  length p
  else:
    setuplengtharray (cum, tot, idx) p;
    gettime (cum, idx) (@#*tot)
  fi
enddef;

vardef pathpoint (expr frac) (suffix p) =
  convertpath (_pp) p;  pnt[pathtime[frac] (_pp)] (p)
enddef;

%    \end{macrocode}
%
%
% \section{Picture-level Operations}\label{picture}
%
% \subsection{Bitwise logical operations}\label{logical}
%
% None of these operations are available in \MP. Mostly these are used by
% higher level operations. Those higher level operations are available in
% \MP, but need to be defined differently.
%
% We have two types of operations. One type is a binary operator that
% takes two picture expressions and returns a picture, the other type
% returns nothing, but merely modifies a given picture variable. These
% take the name of a picture and a picture expression and modify the named
% one. The binary operators are not used elsewhere in graphbase except
% for \gbc{picsub}, which occurs only in \gbc{shadepic}. They are all
% rather wasteful of memory.
%
% \DescribeRoutine{mono}
% Here we define the bitwise logical operations: and, or, xor, and
% difference. These mostly only work if all pixels have values 0 or 1.
% Since \MF{} allows other integer values, we define a \gbc{mono} operator
% that converts all pixels with weight ${}\ge 1$ to 1 and all pixels
% with weight ${}\le 0$ to 0. It is important to note that we can apply
% \gbc{mono} only to the suffix parameter in such things as \gbc{orto}.
% The expression parameter needs to be prepared by the routine that calls
% these. The return result is culled, so it consists only of 0s and 1s.
%    \begin{macrocode}
%<*MF>
def mono (suffix u) =  cull u keeping (1, infinity);  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{andto, picand}
% The bitwise and: in the resulting picture, a pixel is \emph{on} if and
% only if it is \emph{on} in both \gbc{u} and \gbc{v}. \gbc{andto} is
% only used in \gbc{interior} and \gbc{interiors}, \gbc{picand} is not
% used at all.
%    \begin{macrocode}
def andto (suffix u) (expr v) =
  mono (u);  addto u also v;  cull u keeping (2, 2);
enddef;
primarydef u picand v =
  begingroup setpicture (t) u;  andto (t, v);  t endgroup
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{orto, picor}
% The inclusive or: in the result, a pixel is \emph{on} if and only if it
% is \emph{on} in \gbc{u} or \gbc{v} or both. I've written these so that
% it doesn't matter if the expression parameter is not mono. It
% \emph{is} required that it have only positive pixels. The command
% \gbc{orto} is only used three places: in \gbc{coloraddto}, which
% is itself never used, and in \gbc{patcharcs} and \gbc{patchrays}. This
% lack of use is because a less memory intensive version, \gbc{_orto}, is
% defined later, and that is what we use. Usually we build a picture
% in a variable \gbc{src} and add that onto another variable \gbc{dest}.
% If one used \gbc{orto (dest, src)}, then \MF{} would evaluate \gbc{src}
% and pass a \emph{copy} of it as the parameter of \gbc{orto}. This
% doubles the memory used, so mostly we use \gbc{_orto}, which passes both
% parameters as suffixes. \gbc{picor} is never used.
%    \begin{macrocode}
def orto (suffix u) (expr v) =
  mono (u);  addto u also v;  cull u keeping (1, infinity);
enddef;
primarydef u picor v =
  begingroup setpicture (t) u;  orto (t, v);  t endgroup
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xorto, picxor}
% The exclusive or, also called the symmetric difference:
% in the result, a pixel is \emph{on} if and only if it is \emph{on} in
% \gbc{u} or \gbc{v}, but not both. These are not used elsewhere in
% \grafbase.
%    \begin{macrocode}
def xorto (suffix u) (expr v) =
  mono (u);  addto u also v;  cull u keeping (1, 1);
enddef;
primarydef u picxor v =
  begingroup setpicture (t) u;  xorto (t, v);  t endgroup
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{subto}
% The nonsymmetric difference: in the result, a pixel is \emph{on} if
% and only if it is \emph{on} in \gbc{u} and off in \gbc{v}. It is
% unclear whether a \gbc{v} with negative weights will ever occur, but
% if so, subtracting negative pixels ought to be like adding positive
% ones, so I've changed \mfc{keeping (1,1)} to \gbc{keeping (1,infinity)}.
% With this understanding, it doesn't matter here whether \gbc{v} is
% not mono. As with \gbc{orto}, we have a more memory efficient
% \gbc{_subto} and now use that everywhere. \gbc{subto} is only used in
% \gbc{coloraddto}, which is not used anymore. The binop version
% \DescribeRoutine{picsub}\gbc{picsub}
% is used only in \gbc{shadepic}.
%    \begin{macrocode}
def subto (suffix u) (expr v) =
  mono (u);  addto u also -v;  cull u keeping (1, infinity);
enddef;
primarydef u picsub v =
  begingroup setpicture (t) u;  mono (t);  subto (t, v);  t endgroup
enddef;

%</MF>
%    \end{macrocode}
%
% \subsection{Producing and modifying pictures}\label{pictures}
%
% Here we define some slightly higher level commands that make use (in \MF)
% of the previous bitmap operations. In \MP, they mostly need different
% definitions, but we have merged most of them by providing a \MP{}
% alternative for the most frequently used bitmap operation in the
% previous section, \gbc{orto}. These operations either return a picture
% or modify a picture variable. They do not draw anything unless
% \gbc{active_plane} is the modified picture. All curves, points,
% dimension, etc., are in device coordinates.
%
% \DescribeRoutine{coloraddto}
% This was once a useful abbreviation. In \MF{} it adds when the color
% is not white, subtracts when it is. Grays are handles in \MF{} by
% appropriate preparation of \gbc{u} and \gbc{v}. See, for example, the
% code of \gbc{colorsafefill}. In \MP{} it is an abbreviation for the
% basic \mfc{addto} operation. It was defined only so that \MP{} and \MF{}
% can share the same higher level code.
%
% When the last parameter \gbc{v} is the name of picture we can save
% memory if we pass the name rather than the value. Problems with picture
% memory turned up in the shading macros for \MF{} and the dashing macros
% for \MP{}.
% \DescribeRoutine{coloraddon}
% The macro \gbc{coloraddon} applies this memory-saving trick and has
% completely replaced \gbc{coloraddto} in \grafbase{} code. Since
% \gbc{coloraddto} turned out to be used only with \gbc{u} equal to
% \gbc{active_plane}, we have eliminated that parameter from
% \gbc{coloraddon}.
%
% \DescribeRoutine{_orto}
% This version of \gbc{orto} saves memory by passing \emph{both}
% parameters by name. This also allows the application of \gbc{mono} to
% both parameters. In addition to \gbc{coloraddon}, it is used in
% \gbc{shade} and \gbc{tess}.
% \DescribeRoutine{_subto}
% We also have \gbc{_subto}, an analogous version of \gbc{subto}.
%    \begin{macrocode}
def coloraddto (expr clr) (suffix u) (expr v) =
%<*MF>
  if clr < white:
    orto (u, v);
  else:
    subto (u, v);
  fi;
%</MF>
%<MP>  addto u also v _wc_ clr;
enddef;

%<MP>def  orto (suffix u) (expr v) =  addto u also v;  enddef;
%<MP>
def coloraddon (expr clr) (suffix v) =
%<*MF>
  if clr < white:
    _orto (active_plane, v);
  else:
    _subto (active_plane, v);
  fi;
%</MF>
%<MP>  addto active_plane also v _wc_ clr;
enddef;

def _orto (suffix u, v) =
%<MF>  mono (u);  mono (v);
  addto u also v;
%<MF>cull u keeping (1, 2);
enddef;

%<*MF>
def _subto (suffix u, v) =
  mono (u);  mono (v);  addto u also -v;  cull u keeping (1, 1);
enddef;

%</MF>
%    \end{macrocode}
%
% \DescribeRoutine{interior}
% This takes the following expresion, \gbc{c}, which must be a
% closed path, and returns the picture expression which is that path
% filled. The cull command (\MF{} only) retains negative pixels
% (converting them to positive). This way, clockwise contours are filled
% also. \gbc{interior} is one of the most used commands throughout the
% rest of \grafbase.
%
% We ignore color (new behavior with \mfpic{} version 0.7), since the
% higher level commands now implement the coloring operations.
%    \begin{macrocode}
vardef interior expr c =
  newpicture (v);
  addto v contour (c.t_);
%<MF>  cull v dropping (0,0);
  v
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{interiors}
% This is followed by the name of an array of closed paths and
% returns the picture of the interiors of those closed paths. It builds
% the returned picture from \mfc{nullpicture} by successively adding
% the result of \gbc{interior} applied to each path in the array. This is
% only used once by \grafbase, in \gbc{clipsto}, which might be a better
% place to put the \mfc{for}-loop and not use this at all.
%    \begin{macrocode}
vardef interiors suffix cc =
  newpicture (_ints);
  for _idx = 1 upto cc:
    addto _ints also interior cc[_idx]);
  endfor
%<MF>  mono (_ints);
  _ints
enddef;

%    \end{macrocode}
%
% \subsection{Clipping}\label{basicclipping}
%
% \DescribeRoutine{clipto}
% \gbc{clipto} takes the name of a picture \gbc{vt} and a closed path
% \gbc{c} and modifies the picture leaving only the part inside the path.
% In \MP{} we just invoke the \mfc{clip} primitive.
%
% \DescribeRoutine{clipsto}
% This is similar, except it takes an array of paths \gbc{cc} and
% leaves what is interior to any of the paths. This is one case where
% \MP{} requires a substantially different point of view. In \MF, we
% create the interiors and `and' the result to the named picture. In \MP,
% we have to create the picture which is \gbc{vt} clipped to each separate
% path, and combine the results. \Grafbase{} only uses this in the
% \gbc{DoClip} command.
%    \begin{macrocode}
def clipto (suffix vt) expr c =
  if path c:
%<MF>  andto (vt, interior c);
%<MP>  clip vt to c;
  fi
enddef;
def clipsto (suffix vt, cc) =
%<MF>  andto (vt, interiors cc);
%<*MP>
  begingroup
    save _cl, _cl_;  picture _cl, _cl_;  _cl_ := nullpicture;
    for _idx = 1 upto cc:
      _cl := vt;  clip _cl to cc[_idx];  addto _cl_ also _cl;
    endfor
    vt := _cl_;
  endgroup
%</MP>
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{Clipped}
% Here, rather than modify a given picture, \gbc{Clipped} is a vardef
% returning the picture which is the result of clipping the given picture
% to the path. This is not used elsewhere in \grafbase{} nor \mfpic.
%
% Having found out that \mfc{clipped} is a \MP{} primitive, I've
% changed the name to the uppercase version.
%    \begin{macrocode}
vardef Clipped (suffix vt) expr c =
  setpicture (_Cl) vt;  clipto (_Cl) c;  _Cl
enddef;
%<MF>def clip =  Clipped  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{picneg}
% The reverse video is easy in \MF, where \gbc{picneg} takes a picture
% name and a closed path, and returns the part of the picture inside the
% path, but with pixels reversed. In \MP{} we can only approximate this:
% we clip the given picture and add that (using  color \gbc{background})
% on top of the \gbc{interior} of the curve colored \gbc{fillcolor}. This
% is not used elsewhere in \file{grafbase.mp} so it may not be really
% important whether \gbc{fillcolor} and \mfc{background} are the right
% choices.
%    \begin{macrocode}
vardef picneg (suffix vt) expr c =
%<*MF>
  setpicture (_pn) interior c;
  _subto (_pn, vt);
%</MF>
%<*MP>
  setpicture (_cl) vt;  clip _cl to c;
  newpicture (_pn);
  addto _pn also (interior c ) _wc_ fillcolor;
  addto _pn also _cl _wc_ background;
%</MP>
  _pn
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{shpath}
% \gbc{shpath} does most of the work of drawing curves in \grafbase. It is
% called by \gbc{safedraw} which is used by almost all the commands that
% somehow draw a curve. It takes the name of a picture, a pen expression
% and a path expression. It draws the path on the picture with the pen.
% Since we use this (ultimately) for almost all drawing of paths, we
% automatically have the aspect ratio taken care of by the \mfc{.t_}
% macro.
%
% \DescribeRoutine{picpath}
% \gbc{picpath} accepts a path expression and returns a picture, which is
% either \gbc{nullpicture} (\gbc{penwd} too small) or the path drawn with
% \gbc{drawpen}. This is mostly how \gbc{shpath} gets used: curve drawing
% commands produce a picture with \gbc{picpath} and that gets used.
%
%    \begin{macrocode}
def shpath (suffix v) (expr q, f) =
  addto v doublepath (f.t_) withpen (q.t_);
enddef;

numeric minpenwd;
%<MF>minpenwd := 1;      % 1 pixel
%<MP>minpenwd := .05bp;  % 1 pixel at 1440dpi
vardef picpath expr d =
  newpicture (v);
  if penwd >= minpenwd:
    shpath (v, drawpen) (d);
%<MF>    mono (v);
  fi
  v
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{picdot}
% This places a specified picture expression (\gbc{w}) at a specified
% location (\gbc{p}) in a specified picture variable (\gbc{v}). It is used
% a number of places. It's \MF{} version takes care of the aspect ratio
% via \mfc{.t_}. This is how we draw points and symbols and dots along a
% curve: make the symbol into a picture \gbc{w} and add that picture with
% \gbc{picdot}.
%    \begin{macrocode}
def picdot (suffix v) (expr w, p) =
  addto v also
%<MP>    (w shifted p);
%<MF>    (w shifted goodpair (p));
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{setdot}
% \gbc{setdot} is named for its use rather than what it does. It takes a
% path and a scale (numeric expression) and returns a picture which is a
% drawing of the filled interior of the path (if it is a cycle) or the
% path itself (not a cycle). In \MF, we ensure that the scale is at least
% one pixel (assumes that the \gbc{apath} has dimension about 1 and
% \gbc{minpenwd} is 1). This usually assures that something is drawn. In
% \MP, \gbc{minpenwd} has the same purpose (though it is probably not
% necessary). This routine is used a number of times where dots are
% needed. Not in \gbc{shaded} (just below) but later in \gbc{shade} (an
% older command taking paths in graph coordinates), \gbc{polkadot} and
% some grid-making commands.
%    \begin{macrocode}
vardef setdot (expr apath, sc) =
  if cycle apath:  interior
  else:           picpath
  fi
%<MF>    (apath scaled emax (ceiling (sc), minpenwd))
%<MP>    (apath scaled emax (sc, minpenwd))
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{shadepic}
% We want to shade regions with a very regular pattern of black and white
% pixels for best appearance. Experiments show that symmetric dots
% (e.g., circles, squares) work better than non-symmetric (e.g.,
% rectangular). Circular dots are not significantly better than square at
% the size needed. I believe that the default result of \gbc{shade} looks
% reasonably good on my system. (That happens to produce two 3-pixel by
% 3-pixel square dots in a 8-pixel square on a 360dpi printer.) So we try
% to produce something similar. That is, the shading picture is 1.6bp
% (8 pixels at 360dpi) square.
%
% As a compromise (symmetric dots look better, but rectangular dots give
% more gray levels) we allow dots to be rectangles $k\times (k+1)$-pixels
% (assuming the aspect ratio is 1). This produces twice the number of
% gray levels. In my 360dpi example we get 15 gray levels. The two
% farthest apart (4 by 4 dots versus 3 by 4 dots) differ by 1/8 in
% fraction of area of coverage (which we equate to grayness).
%
% Why can't we have 64 grey levels in a $8\times 8$ square? Clearly we
% can in principle turn on any number of the 64 pixels. Unfortunately,
% spread out patterns (which look best) tend to consume memory, while
% clumpy patterns are hard to make good-looking. Compensating for aspect
% ratios unequal to 1 is also pretty hard to do automatically.
%
% The parameter \gbc{dims} needs to be a pair variable, and it will be
% assigned the actual dimensions of the picture returned. These routines
% are complicated by the fact that we may have an aspect ratio unequal to
% $1$. When \mfc{aspect_ratio = 1} the basic concept is simple: make an
% $n\times n$ square with two dots, each nearly $k \times k$ and nearly
% square, where $2k^2/n^2$ is the gray level needed.
%
% The calculations assume a gray level greater than $1/2$, so the final
% picture will be mostly white (for darker grays, we use the
% complementary gray level to construct the `reverse video', and then
% reverse back). Under this assumption, we concentrate all the black
% pixels into the lower left and upper right quadrant of the picture we
% are creating, so most of the calculation determines one of these
% quadrants. The scratch variables \gbc{_hp} and \gbc{_vp} give the number
% of horizontal and vertical pixels in the lower left quadrant,
% \gbc{_dotwd} and \gbc{_dotht} do the same for the actual dot. Then
% \gbc{_shp} is first set equal to one dot; then a copy of itself is added
% in the upper right quadrant. Finally, the suffix parameter \gbc{dims} is
% equated to the nominal width and height of the picture, and either
% \gbc{_shp} or its reverse is returned.
%    \begin{macrocode}
%<*MF>
numeric shadepicsize;  shadepicsize := 0.8bp;
vardef shadepic (suffix dims) (expr grparam) =
  pair dims;
  setnumeric (_frac) 2*emin (grparam, 1 - grparam);
  save _hp, _vp, _dotwd, _dotht;
  if aspect_ratio < 1:
    _vp := emax (2, hround (shadepicsize.o_));
    _hp := hround (_vp._o_);
    _dotwd := hround (_hp*sqrt _frac);
    _dotht := if _dotwd = 0:  0
              else:  hround (_hp*_vp*_frac/_dotwd)
              fi;
  else:
    _hp := emax (2, hround (shadepicsize));
    _vp := hround (_hp.o_);
    _dotht := hround (_vp*sqrt _frac);
    _dotwd := if _dotht = 0:  0
              else:  hround (_hp*_vp*_frac/_dotht)
              fi;
  fi
  dims := ( _hp, _vp._o_ );
  newpicture (_shp);
  addto _shp contour rect (origin, (_dotwd, _dotht));
  picdot (_shp, _shp, dims);
  dims := 2dims;  mono (_shp);
  if grparam >= .5:  _shp
  else:  (interior (rect (origin, dims))) picsub _shp
  fi
enddef;

%</MF>
%    \end{macrocode}
%
% \DescribeRoutine{shaded}
% This fills the interior of a contour (device coordinates) with copies of
% \gbc{shadepic}. The routine \gbc{fillwith} is defined later, but its
% name reflects its effect: a bounding rectangle (corners at \gbc{ll} and
% \gbc{ur}) is filled with copies of a picture (in this case, the result
% of \gbc{shadepic}), the picture having nominal dimensions \gbc{shdims}
% in this case.
%
% It may seem odd that black and white return the same thing. That is
% because white is handled in the calling routine by subtracting the
% black result.
%
% The \gbc{setbbox} command was defined earlier, in
% section~\ref{utilities}. The bounding rectangle it obtains is only
% approximate in \MF{}, but that is sufficient, since we only use it to
% produce things that are eventually clipped.
%
% We return \gbc{picpath} for non-cycles because I once thought to make
% \gbc{shaded} a replacement for \gbc{setdot} to get gray dots (in the
% \gbc{polkadot} routine). That turns out not to work, but this sort of
% thing is also done in most of the rendering commands that require a
% closed path.
%    \begin{macrocode}
vardef shaded (expr clr) expr c =
  if cycle c:
%<*MP>
    newpicture (v);
    addto v contour c _wc_ clr;
    v
%</MP>
%<*MF>
    if (clr <= black) or (clr >= white):
      interior c
    else:
      save shdims, shpic;
      picture shpic;  pair shdims;
      shpic := shadepic (shdims) (clr);
      setbbox (ll, ur) c;
      newpicture (vsh);
      fillwith (vsh) (shpic, shdims, ll, ur);
      clipto (vsh) c;  vsh
    fi
%</MF>
  else:  picpath c % should we? or just make it null?
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{fillwith}
% This is one of the ways we obtain something other than a solid fill. The
% routines \gbc{polkadot}, \gbc{tess} and (in \MF) \gbc{shade} and
% \gbc{shaded} all use it.
%
% It takes a picture expression \gbc{pic}, along with its dimensions (the
% pair \gbc{dims}) in device coordinates, plus the opposite corners,
% \gbc{ll} and \gbc{ur}, of a boundingbox rectangle, and draws that
% rectangle filled with copies of \gbc{pic}. Starting with \mfpic{}
% version 0.8, it adds to a predefined picture passed by name. Thus the
% calling routine must make sure that picture is initialized (it need
% not be \mfc{nullpicture}).
%
% One might do this with two nested loops, but it turns out to be much
% faster (surprisingly much!) to do two separate loops: the second one
% stacking copies of the row built by the first loop.
%
% We try to do any rounding that might have been forgotten. This code
% takes a mode's aspect ratio into account so that (most) calling routines
% don't have to. (That is, \gbc{dims} should be measured in horizontal
% pixels, while \gbc{fwdims} is in actual pixels. This could have been
% written in terms of \gbc{picdot}, which already handles aspect, but it
% has got to be more efficient to do the aspect ratio calculations once
% rather than every time through the loop.)
%    \begin{macrocode}
vardef fillwith (suffix v) (expr pic, dims, ll, ur) =
  newpicture (b);
%<*MF>
  save fwdims, _ll, _ur;  pair fwdims, _ll, _ur;
  fwdims := goodpair (dims);
  _ll   := floorpair  (ll.t_);
  _ur   := ur.t_;
  for s = xpart _ll step xpart fwdims until xpart _ur:
    addto b also pic shifted (s, 0);
  endfor
  for s = ypart _ll step ypart fwdims until ypart _ur:
    addto v also b shifted (0, s);
  endfor
  mono (v);
%</MF>
%<*MP>
  for s = xpart ll step xpart dims until xpart ur:
    addto b also pic shifted (s, 0);
  endfor
  for s = ypart ll step ypart dims until ypart ur:
    addto v also b shifted (0, s);
  endfor
%</MP>
enddef;

%    \end{macrocode}
%
% \subsection{Hatching}\label{basichatching}
%
% \DescribeRoutine{thatchf}
% This is the all-purpose macro called by the other macros that
% fill a region with hatching. It takes the name of a picture \gbc{v},
% a transform expression \gbc{CT}, a numeric expresion \gbc{sp} giving the
% space between hatch lines, and two pairs, \gbc{a} and \gbc{b},
% that represent the lower left and upper right limits of a rectangle.
% The expression \gbc{sp} must be nonzero. The calling macros should take
% care of that.
%
% It modifies the picture by adding to it the rectangle full of
% hatching lines spaced \gbc{sp} apart. The rectangle is initially upright
% and the lines horizontal, but they are drawn transformed by the
% transform \gbc{CT}. This is how diagonal hatching is accomplished: the
% transform is a rotation.
%
% We guard against \gbc{ypart a} being greater than \gbc{ypart b} or
% \gbc{sp} being negative: \gbc{_sp} is \gbc{sp} modified to have the same
% sign as \gbc{ypart (b - a)}. Thus, repeatedly adding it to \gbc{ypart a}
% gets one to \gbc{ypart b}. We make the starting value an integer
% multiple of \gbc{_sp} to make sure adjacent regions don't have jarringly
% misaligned hatch lines. (I guess that's the reason; this algorithm
% predates me.)
%    \begin{macrocode}
def thatchf (suffix v) (expr CT, sp, a, b) =
  begingroup
    setnumeric (_sp) signof (ypart b - ypart a) abs(sp);
    for _y = _sp*( ceiling ((ypart a)/_sp) ) step _sp until ypart b:
      shpath (v, hatchpen)
        ( ( (xpart a, _y)--(xpart b, _y) ) transformed CT );
    endfor
%<MF>    mono (v);
  endgroup
enddef;

%    \end{macrocode}
%
% \subsection{Tiles}\label{tiles}
%
% Tesselations are a type of fill in which a rectangular pattern is
% repeated throughout a region. The repeated rectangle is called a tile.
% We provide here an environment in which the drawing commands add to a
% picture variable other than \mfc{currentpicture}. We do this very simply
% by redefining \gbc{active_plane}, localizing the redefinition between
% \gbc{tile} and \gbc{endtile}
%
% \DescribeRoutine{tile}
% The macro \gbc{tile} accepts one suffix parameter, the name of the tile,
% followed by three numeric expressions and a boolean. \gbc{unit} should
% be a dimension in device units and is the unit of length for all high
% level drawing commands within the environment. \gbc{width} and
% \gbc{height} specify the size of the tile in multiples of \gbc{unit}, and
% \gbc{clipit} is a boolean that determines if the resulting picture is
% clipped to the rectangle these parameters determine. For example,\\
% \indent \gbc{tile (fred)(1in, 1, 2, true)} \\
% starts a tile named \gbc{fred} which will be 1 inch wide and 2 inches
% tall, and any marks that extend beyond this rectangle are clipped off.
% The tile is enclosed in a group to delimit these changes to the basic
% drawing parameters.
% \DescribeRoutine{endtile}
% The macro \gbc{endtile} merely implements the clipping and then closes
% the group.
%
% In \MF, the picture should be a whole number of pixels in size, so that
% the tiles fit perfectly together. The fact that shifts must be integer
% values is only mildly relevant, because the placement code does the
% rounding.
%
% For tesselation (filling with tiles) we need to know various properties
% of the tile, so a tile is a composite object consisting of a picture,
% \gbc{fred.pic} in our example (the actual tile) and a pair
% \gbc{fred.dims} of the dimensions (in device units). We used to
% save the \gbc{clipit} parameter in \gbc{atile.clipon}, but it was never
% used. We also used to have separate numerics \gbc{atile.wd} and
% \gbc{atile.ht} but they only got used together as a pair.
%    \begin{macrocode}
def tile (suffix atile) (expr unit, width, height, clipit) =
  picture atile.pic;  atile.pic := nullpicture;
  pair atile.dims;
%<MF>  atile.dims := round ((width, height)*unit);
%<MP>  atile.dims := (width, height)*unit;
  begingroup
%    \end{macrocode}
% We do a subset of what we do in \gbc{beginmfpic}, redefining
% \gbc{active_plane} so that all drawing commands that add to it will
% contribute to the tile, and adapting \gbc{ztr} to the tile dimensions.
% Re also redefine \gbc{xneg}, et al., for the benefit of \gbc{levelset}.
%    \begin{macrocode}
    save active_plane;
    def active_plane =  atile.pic  enddef;
    save ztr, vtr;  transform ztr, vtr;
    ztr := identity scaled unit;  vtr := ztr;
    save xneg, xpos, yneg, ypos;
    xneg := 0;  xpos := width;  yneg := 0;  ypos := height;
%    \end{macrocode}
% To implement \gbc{clipit}, we set the current clipping path array
% \gbc{ClipPath[\,]} to the boundary of the tile. Note that this turns off
% user-defined clipping paths, which are unlikely to be correct for the
% local tile coordinates.
%    \begin{macrocode}
    save ClipOn;  boolean ClipOn;
    if clipit:
      ClipOn := true;
      setarray (path) (ClipPath) (rect(origin, atile.dims));
    else:
      ClipOn := false;
    fi
enddef;
def endtile =
    DoClip (active_plane);
  endgroup
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{is_tile}
% To test whether \gbc{atile} is really a tile, just see if the needed
% components are defined and of the correct type.
%    \begin{macrocode}
vardef is_tile (suffix atile) =
  (known atile.pic ) and (picture atile.pic) and
  (known atile.dims) and (pair  atile.dims )
enddef;

%    \end{macrocode}
%
% \section{Bounding Boxes of Paths}\label{bboxes}
%
% To fill a region with other than a solid fill, we normally fill a
% rectangle with copies of a picture (or a path) and then clip to the
% boundary curve. In order not to place too many copies, we try to find a
% rectangle that is not too much larger than that region. For this we have
% the macro \gbc{getbbox} which takes two pair variable and a path
% expression, and sets the pairs to the lower left corner and upper right
% corner, respectively, of a rectangle enclosing the path. The bounding
% box macros are used on paths in device coordinates, but there is no
% intrinsic reason that has to be so: they will return the bounding box in
% whatever coordinates the supplied path is in.
%
% \DescribeRoutine{getbbox}
% One can get a rather loose bounding rectangle by using the fact that
% each segment of a path (from \mfc{point j of g} to \mfc{point j+1 of g})
% is contained in the convex set determined by all 4 control points for
% that segment. So we get a containing rectangle by getting the smallest
% and largest values of the $x$- and $y$-coordinates of all those points.
% We can get a considerably tighter fit if we cut each segment in half
% (or more) before doing that. A calling routine is expected to save and
% declare the suffixes \gbc{ll} and \gbc{ur}. Within \grafbase{} commands,
% \gbc{getbbox} is always called by \gbc{setbbox}, which does this.
%
% \DescribeRoutine{ctrlsbbox}
% There is a difference between ``\mfc{postcontrol 0 of (subpath (j,j+1/2)
% of p)}'' and ``\mfc{postcontrol j of p}''. To gain the tighter box we have
% to look at the former. \gbc{ctrlsbbox} just updates the previously found
% corners \gbc{ll} and \gbc{ur} of the bounding box based on the controls
% of the path segment \gbc{p}, and the calling routine \gbc{getbbox} passes
% it half a segment at a time. We don't examine the endpoints of
% the half-segment: one has already been examined by \gbc{getbbox} and
% the other (a subdivision point of an original segment) lies on the line
% segment connecting two control points, and so can't increase the bbox.
%
% We've given this potentially unlimited accuracy by allowing the number
% of subdivisions (\gbc{bbox_split}) to be arbitrary. We choose 2 for the
% default. The \gbc{setsplit} command (subsection~\ref{utilities}) ensures
% that \gbc{_s} is integral and positive, just in case \gbc{bbox_split}
% somehow isn't.
%
% This description applies only to \MF, because \MP{} has built-in
% facilities for determining the bounding box.
%
% \RoutineIndex{pnt}
% \RoutineIndex{pre}
% \RoutineIndex{post}
% I got tired of typing long expressions like ``\gbc{(precontrol length
% p of p)}'', and now use the following abbreviations.
%    \begin{macrocode}
vardef pnt@#  (expr p) =  point       @# of p  enddef;
vardef pre@#  (expr p) =  precontrol  @# of p  enddef;
vardef post@# (expr p) =  postcontrol @# of p  enddef;

def getbbox (suffix ll, ur) expr g =
%<MP>  ll := llcorner g;  ur := urcorner g;
%<*MF>
  setsplit (_s) bbox_split;
  ur := ll := pnt 0 (g);
  for _j = 1 upto length g:
    ll := pairmin (ll, pnt[_j] (g));  ur := pairmax (ur, pnt[_j] (g));
  endfor
  for _j = 1 upto _s*(length g):
    ctrlsbbox (subpath ((_j-1)/_s, _j/_s) of g) (ll, ur);
  endfor
%</MF>
  if showbbox:  noclip ( safedraw rect (ll, ur) );  fi
enddef;

%<*MF>
numeric bbox_split;  bbox_split := 2;
def ctrlsbbox (expr p) (suffix ll, ur) =
  ll := pairmin ( pairmin (ll, post0 (p)), pre 1 (p) );
  ur := pairmax ( pairmax (ur, post0 (p)), pre 1 (p) );
enddef;

%</MF>
%    \end{macrocode}
%
% We also have \gbc{tightbbox} and \gbc{tbbox} in \MF{} but these are no
% longer used so we'll omit them from \grafbase, but keep them in the
% documentation for now.
%
% \DescribeRoutine{tightbbox}
% Calculate tight bounding box points \gbc{ll} and \gbc{ur} for path
% \gbc{g}. The tight bounding box is accurate to the limits of the
% \mfc{solve} macro, which is the numeric \mfc{tolerance}, which we set to
% \mfc{.5} (accurate enough, assuming pixel units). This is only called by
% \gbc{tbbox}, which is never used.
%
% \DescribeRoutine{_xlimit}
% \gbc{xlimit(x)} returns a value of true if the path \gbc{g} doesn't
% cross the vertical line at \gbc{x}.
% \DescribeRoutine{_ylimit}
% \gbc{ylimit(y)} is the same for the horizontal line at \gbc{y}.
%    \begin{macrocode}
%<*unused>
def tightbbox (expr g) (suffix ll, ur) =
begingroup
  interim tolerance := .5;
  ll := ( (solve _xlimit (-infinity, xpart pnt 0 (g))),
          (solve _ylimit (-infinity, ypart pnt 0 (g))) );
  ur := ( (solve _xlimit ( infinity, xpart pnt 0 (g))),
          (solve _ylimit ( infinity, ypart pnt 0 (g))) );
endgroup
  if showbbox:
    noclip ( safedraw rect (ll, ur) );
  fi
enddef;

vardef _xlimit (expr x) =
  ((x, -infinity)--(x, infinity)) misses g
enddef;

vardef _ylimit (expr y) =
  ((-infinity, y)--(infinity, y)) misses g
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{tbbox}
% \gbc{tbbox} simply calls \gbc{tightbbox} on each of an array of paths
% and takes the maximum of all the upper right corners and the minimum of
% all the lower left. Same syntax as \gbc{tightbbox} except that, instead
% of a path parameter, \gbc{g} must be the name of an array of paths.
% This macro is never used elsewhere in \grafbase.
%    \begin{macrocode}
vardef tbbox (suffix g) (suffix ll, ur) =
  save _gll, _gur;  pair _gll, _gur;
  tightbbox (g1, ll, ur);
  for _idx = 2 upto g:
    tightbbox (g[_idx], _gll, _gur);
    ll := pairmin (ll, _gll);  ur := pairmax (ll, _gur);
  endfor
  if showbbox:  noclip ( safedraw rect (ll, ur) );  fi
enddef;

%</unused>
%    \end{macrocode}
%
% \section{Device Coordinate Rendering Commands}\label{basicrendering}
%
% We use the word `rendering' to refer to commands that accept a path
% expression as one parameter and use it to modify the \gbc{active_plane}.
% All the commands in this section expect paths, pairs and dimensions in
% device coordinates.
%
% \subsection{Drawing}\label{basicdrawing}
%
% \DescribeRoutine{safedraw}
% \gbc{safedraw} accepts a path expression, and adds the result to
% \gbc{active_plane}. It is the first drawing command to draw
% exclusively on \gbc{active_plane}. This is the first of many uses of
% \gbc{coloraddon}. In \MP{} it is basically the primitives  \mfc{addto}
% and \mfc{withcolor} applied to \gbc{active_plane}, but in \MF{} it adds
% when the color is less than 1 (gray or black), otherwise it subtracts
% (white).
%
% \RoutineIndex{colorsafedraw}
% The command \gbc{safedraw} merely calls \gbc{colorsafedraw}, which then
% calls \gbc{picpath}, which calls \gbc{shpath}. One reason for this
% roundabout sequence is to support older files (where \gbc{colorsafedraw}
% was not defined). Another is that color handling in \MF{} requires a
% picture with pixels of weight 1 or 0 only (\gbc{picpath}). Moreover,
% \gbc{shpath} guarantees that the mode's aspect ratio is respected.
%    \begin{macrocode}
def safedraw =  colorsafedraw (drawcolor)  enddef;
def colorsafedraw (expr clr) expr d =
  begingroup
    setpicture (v) picpath d;
    DoClip (v);  coloraddon (clr, v);
  endgroup
enddef;

%    \end{macrocode}
%
% \subsection{Filling}\label{basicfilling}
%
% \DescribeRoutine{NoCycle}
% This is a common warning for all those commands that require a cycle
% (closed path) but an open path is supplied. In addition to the warning
% in those commands, we also call \gbc{safedraw} for debugging purposes.
%
% \DescribeRoutine{safefill}
% \RoutineIndex{colorsafefill}
% The basic \gbc{safefill} simply calls the colored version with the
% default parameter \gbc{fillcolor}. \gbc{colorsafefill} takes a color as
% its first parameter and a path expression as second. These commands fill
% the path in the \gbc{active_plane}. In \MF, when the color is strictly
% between $0$ and 1, a gray fill is simulated with the \gbc{shaded} macro.
%
% To simulate the effect of painting over in gray, the \MF{} version
% clears the region before adding the shaded fill.
%
% \DescribeRoutine{safeunfill}
% \gbc{safeunfill} is just \gbc{safefill} with the color \mfc{background}.
% In \MF{}, when \gbc{background = white = 1}, this is detected by
% \gbc{coloraddon} which then subtracts the picture. We do this inside
% \gbc{noclip}, just because it seems a user would expect clipping only
% when things are \emph{added}. In \MP{} the white is indeed added, but
% conceptually, material is cleared away.
%    \begin{macrocode}
def NoCycle (expr s) expr p =
  GBwarn  s & " cannot be applied to an open path."
            & " The path will be drawn instead.";
  safedraw p;
enddef;

%<MF>vardef isgray (expr X) =  (X > black) and (X < white)  enddef;
%<MF>
def safefill =  colorsafefill (fillcolor)  enddef;
vardef colorsafefill (expr clr) expr c =
  if cycle c:
    setpicture (v) interior c;
    DoClip (v);
%<*MF>
    if isgray (clr):
      _subto (active_plane) (v);
      v := nullpicture;
      v := shaded (clr) c;
    fi
%</MF>
    coloraddon (clr, v);
  else:  NoCycle("fill") c;
  fi
enddef;
def safeunfill expr c =
  if cycle c:  noclip (colorsafefill (background) c);
  else:  NoCycle("unfill") c;
  fi
enddef;

%    \end{macrocode}
%
% \subsection{Clipping}\label{clipping}
%
% \DescribeRoutine{safeclip}
% This applies \gbc{clipto} to the active drawing plane. It follows the
% pattern started with \gbc{safefill} where commands that require a cycle
% will \gbc{safedraw} non-cyclic paths.
%    \begin{macrocode}
def safeclip expr c =
  if cycle c:  clipto (active_plane) c;
  else:  NoCycle("clip") c;
  fi
enddef;

%    \end{macrocode}
%
% \section{Graph Coordinate Rendering}\label{rendering}
%
% \DescribeRoutine{store}
% Now we come to the highest level rendering operations. These are the
% commands written to the output file by \mfpic. They accept a path in
% \emph{graph} coordinates, convert it to device coordinates, rendering
% the result, and return the original path. This way one can render a
% path and pass it on to the preceding command for further processing.
% This is how \mfpic{} implements multiple prefix macros. However, this
% cannot be kept up because \MF{} abhors an isolated expression. Therefore
% we provide a command that accepts a path and doesn't pass it on. In
% theory, it could do nothing, but in \mfpic{} we store the path in
% \gbc{curpath}, making every \mfpic{} figure a path assigment command
% and the rendering is `merely' a side-effect.
%
% \DescribeRoutine{stored}
% The macro \gbc{stored} performs \gbc{store}, but passes the same path as
% its return value. This is used by \mfpic{} to implement the \cs{store}
% command, allowing it to also be a prefix macro
%
% I don't know if \gbc{store} needs to employ \mfc{hide()}, but it seems
% not to hurt.
%    \begin{macrocode}
def store (suffix fs) expr f =
hide (
  if (not path f) and (not pair f):
    GBerrmsg ("Improper expression type.")
      "The second argument to `store' must be a path or pair.";
  fi
  if not path fs:  path fs;  fi
  fs := f
)
enddef;
vardef stored (suffix fs) expr f =  store (fs) f;  f  enddef;

%    \end{macrocode}
%
% \subsection{Drawing}\label{drawing}
%
% \DescribeRoutine{drawn}
% \RoutineIndex{colordrawn}
% The command \gbc{drawn} merely calls \gbc{colordrawn} with the default
% color \gbc{drawcolor}. Then \gbc{colordrawn} takes a color \gbc{clr}
% and a path expression \gbc{f} and returns the same path. In between,
% \gbc{zconv (f)} is subjected to \gbc{colorsafedraw}.
%    \begin{macrocode}
def drawn =  colordrawn (drawcolor)  enddef;
vardef colordrawn (expr clr) expr f =
  colorsafedraw (clr) (zconv (f));  f
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{colorwiggle}
% This is a multi-tasking command that can draw either zigzag or
% sinewave shapes depending on the boolean first parameter. For \mfc{true}
% we get smooth wiggles, for \mfc{false} we get jagged ones. In the
% smooth case, a tension parameter allows an adjustment to the smoothness.
% The command
% \DescribeRoutine{zigzag}
% \RoutineIndex{colorzigzag}
% \gbc{zigzag} calls it with the value \mfc{false} and an arbitrary
% value of the tension;
% \DescribeRoutine{sinewave}
% \RoutineIndex{colorsinewave}
% \gbc{sinewave} calls it with \mfc{true}, allowing it to pick up the
% tension parameter. All expect a quadruple of dimensions to follow
%
% The reason for using a loop (at the end) that draws the \gbc{sinewave}
% path in pieces, is that all the turning can quickly exceed \MF{}'s limit
% on the autorounding stack. I'd never heard of this stack until I ran
% this without a loop and received the ``capacity exceeded'' message. This
% turns out to be a problem mostly when the ratio of \gbc{len} to
% \gbc{wid} is too small and the `humps' of the sine are more like
% `bulbs'. However it is always a problem with \gbc{corkscrew} (below).
%    \begin{macrocode}
def zigzag =  colorzigzag (drawcolor)  enddef;
def colorzigzag (expr clr) =  colorwiggle (false, clr, 0)  enddef;
def sinewave =  colorsinewave (drawcolor)  enddef;
def colorsinewave =  colorwiggle (true)  enddef;

vardef colorwiggle (expr smth, clr, tens, blen, elen, len, wid) expr f =
  convertpath (g) f;
  setuplengtharray (cumlen, totlen, ct) g;
  save B;
  if cycle f:
    B := 0;
  else:
    B := abs(blen)/_rescale_factor;
    totlen := totlen - B - abs(elen)/_rescale_factor;
  fi
  setnumeric (n) 2*round (totlen/len*_rescale_factor);
  if n < 2:
    colorsafedraw (clr) g;
  else:
    save T, U, X, Y, Z, p;
    pair U, X, Y, Z;  path p;
    T := if cycle f:  0  else:  gettime (cumlen, ct) (B) fi;
    Z := pnt[T] (g);
    p :=if not cycle f:  (subpath (0,T) of g)
          if smth:  {curl 0} ..tension tens..  else:  --  fi
        fi
        for i = 1 upto n:
          hide(
            T := gettime (cumlen, ct) (B+(i/n)*totlen);
            X := Z;  Z := pnt[T] (g);
            Y := .5[X,Z];  U := sgn (Z-X);
          )
          (Y + (U zscaled (0, if even i:  -  fi  wid)))
            if smth:  {U}..tension tens..  else:  --  fi
        endfor
        if cycle f:  cycle
        else:  if smth:  {curl 0}  fi  (subpath (T, length g) of g)
        fi;
    newpicture (v);
%<MP>    shpath (v, drawpen) (p);
%<*MF>
    if smth:
      save n, k;
      n := length p;  k = n div 50;
      for i = 0 step 50 until 50*(k-1):
        shpath (v, drawpen) (subpath (i,i+50) of p);
      endfor
      shpath (v, drawpen) (subpath (50k,n) of p);
    else:  shpath (v, drawpen) (p);
    fi
%</MF>
    DoClip(v);  coloraddon (clr, v);
  fi
  f
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{corkscrew}
% \RoutineIndex{colorcorkscrew}
% The definition of \gbc{corkscrew} shares a lot of code with \gbc{zigzag}
% and \gbc{sinewave}, but the middle is considerably different, so it is
% not really possible to make a multipurpose command that can do all
% three.
%    \begin{macrocode}
def corkscrew =  colorcorkscrew (drawcolor)  enddef;
vardef colorcorkscrew (expr clr, tens, blen, elen, len, wid) expr f =
  convertpath (g) f;
  setuplengtharray (cumlen, totlen, ct) g;
  save B;
  if cycle f:
    B := 0;
  else:
    B := abs(blen)/_rescale_factor;
    totlen := totlen - B - abs(elen)/_rescale_factor;
  fi
  setnumeric (n) round (totlen/len*_rescale_factor);
  if n < 2:
    colorsafedraw (clr) g;
  else:
    save T, U, X, Y, Z, p;
    pair U, X, Y, Z;  path p;
    T := if cycle f:  0  else:  gettime (cumlen, ct) (B) fi;
    Z := pnt[T] (g);
    p :=if (not cycle f) and (B > 0):  (subpath (0,T) of g)--  fi
        for i = 1 upto n:
          hide(
            T := gettime (cumlen, ct) (B+(i/n)*totlen);
            X := Z;  Z := pnt[T] (g);
            Y := .5[X,Z];  U := sgn (Z-X);
          )
          (X + (U zscaled (0,-wid))){ U}..tension tens..
          (Y + (U zscaled (0, wid))){-U}..tension tens..
        endfor
        if cycle f:  cycle
        else:
          {U}(Z + (U zscaled (0,-wid)))
          if elen <> 0:  --(subpath(T, length g) of g)  fi
        fi;
    newpicture (v);
%<MP>    shpath (v, drawpen) (p);
%<*MF>
    save n, k;
    n := length p;  k = n div 50;
    for i = 0 step 50 until 50*(k-1):
      shpath (v, drawpen) (subpath (i,i+50) of p);
    endfor
    shpath (v, drawpen) (subpath (50k,n) of p);
%</MF>
    DoClip(v);  coloraddon (clr, v);
  fi
  f
enddef;

%    \end{macrocode}
%
% \subsection{Filling, unfilling and clipping}\label{filling}
%
% \DescribeRoutine{filled}
% \RoutineIndex{colorfilled}
% The command \gbc{filled} calls \gbc{colorfilled} with the default color
% \gbc{fillcolor}. Then \gbc{colorfilled} takes a color \gbc{clr} and a
% path expression \gbc{c}, returning the same path after subjecting
% \gbc{zconv (c)} to \gbc{colorsafefill}.
% \DescribeRoutine{unfilled}
% The macro \gbc{unfilled} returns the path after running \gbc{safeunfill}.
%
% \DescribeRoutine{Clip}
% Finally, \gbc{Clip} is similar, running \gbc{safeclip}. The name
% \gbc{clip} (lowercase) is taken: it is a \MP{} primitive.
%    \begin{macrocode}
def filled =  colorfilled (fillcolor)  enddef;
vardef colorfilled (expr clr) expr c =
  colorsafefill (clr) zconv (c);  c
enddef;
vardef unfilled expr c =  safeunfill zconv (c);  c  enddef;
vardef Clip expr c =  safeclip zconv (c);  c  enddef;

%    \end{macrocode}
%
% \subsection{Shading}\label{shading}
%
% \DescribeRoutine{shade}
% Shading is accomplished differently in \MP{} from \MF; however, many of
% the same parameters are used for compatibility (so that \MP{} can be run
% on a \file{.mf} created for \grafbase{} by \mfpic). In \MP, shading is
% just filling with some level of gray. In \MF, we place a pattern of
% small dots with the size and spacing adjustable. For compatibility,
% \MP{} accepts these size and spacing parameters, but simply uses them to
% calculate the darkness of gray.
%
% Ideally (i.e., for best appearance) one would shade with single pixels
% placed in a regular pattern. Unfortunately, this is the most memory
% intensive for \MF, which stores bitmaps by scanning each row of pixels,
% and records where changes from black to white occur. We do use simple
% dots, but make them quite a bit larger than one pixel. By default,
% \gbc{0.5bp} in diameter, spaced (in \mfpic) a default \gbc{1pt} between
% centers.
%
% The shape and size of the dots can be selected by defining
% \gbc{shadedotpath} and \gbc{shadewd}. A closed path representing the
% boundary of one dot of unit size, \gbc{shadedotpath} is initialized to a
% circle.
%
% The parameter \gbc{sp} is the distance between the centers of the dots in
% device coordinates, and \gbc{f} is the path to be filled in \emph{graph
% coordinates}.
%
% As usual, if the path is not closed, we draw the curve instead. If the
% spacing is too small relative to \gbc{shadewd}, we fill the curve.
% Otherwise the \gbc{fillwith} macro is used to fill with copies of
% a dot picture.
%    \begin{macrocode}
numeric shadewd;  shadewd := 0.5bp;
path shadedotpath;
shadedotpath := fullcircle;

vardef shade (expr sp) expr f =
  convertpath (g) f;
%    \end{macrocode}
% It seems clear that the gray level (\gbc{gr}) should depend
% quadratically on \gbc{shadewd/sp}. Also, there is a point where the
% result is essentially black and a fill would be more efficient.
% The value .88 is arrived at empirically and is a compromise so that
% \MF{} and \MP{} produce similar levels of gray on both printers
% available to me.
%    \begin{macrocode}
  setnumeric (gr) 1 - (.88*abs(shadewd)/sp)**2;
  if not cycle g:  NoCycle("shade") g;
  elseif gr <= 0:  safefill g;
  else:
%<*MF>
    setbbox (ll, ur) g;
    ll := floorpair (ll);
%    \end{macrocode}
%
% What we do is draw a row of dots and stack the rows to fill a rectangle.
% We call \gbc{fillwith} to draw these copies. We have to produce this
% on a picture separate from \gbc{active_plane} so we can apply
% \gbc{DoClip}. Adding one complex picture onto another means there is a
% time at which the added picture and the result are in memory at the same
% time. To save a little memory (at the cost of a little speed), we
% divide the picture into two, each with half the complexity. We add the
% one, then (I hope) recover the memory by equating it to \mfc{nullpicture}
% before adding the second.
%
% Shifts of pictures need to be by integer number of pixels, but this is
% ensured by \gbc{fillwith}, using \mfc{ceiling} to define \gbc{dv} is
% more to ensure it is not rounded down to 0.
%    \begin{macrocode}
    % setpair (dv) ceiling (sp/(sqrt 2))*(1,1);
    % test hex spacing:
    setpair (dv) ( ceiling(.5sp), ceiling(.5sp*sqrt 3) );
    setpicture (sh) setdot (shadedotpath, abs(shadewd));
    newpicture (v);
    fillwith (v) (sh, 2dv, ll, ur);
    newpicture (w);
    addto w also v shifted goodpair (dv);
    DoClip (v);  DoClip (w);
    clipto (v) (g);  clipto (w) (g);
    _orto (active_plane, v);
    v := nullpicture;
    _orto (active_plane, w);
%</MF>
%    \end{macrocode}
% In \MP{} we just fill with gray. The gray level having been calculated
% at the beginning.
%    \begin{macrocode}
%<*MP>
    colorsafefill (gr*white) g;
%</MP>
  fi
  f
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{polkadot}
% The macro \gbc{polkadot} is intended to fill a region with \emph{large}
% dots. The diameter, \gbc{polkadotwd}, is initialized to \mfc{5bp}. The
% code is similar to that of \gbc{shade}, but here we attempt a hexagonal
% array: each dot surrounded by 6 equally spaced dots. Because of their
% larger size and presumably larger spacing, we can be a little less
% efficient and so we aim for improved visual appearance. We do what we
% can to avoid unsightly slivers of partial dots, and only draw a dot if
% its center lies in the bounding box.
%
% We also permit the circles to overlap, and only replace the code with a
% fill if the dots overlap so much that no background can show (this assumes
% that \gbc{polkadotpath} is a circle).
%
% If the space \gbc{sp} and \gbc{polkadotwd} are too small, there will
% be a great many tiny dots. It is quite easy to overflow \MP{} capacity
% and tiny paths don't rasterize at all well in \PS. In \MF, we already
% have \gbc{shade} to place tiny dots. Therefore, we merely fill if
% \gbc{sp} is less that a certain minimum, even if that minimum is greater
% than \gbc{polkadotwd}.
%    \begin{macrocode}
polkadotwd := 5bp;
mindotspace := 1bp;
path polkadotpath;  polkadotpath := fullcircle;

vardef polkadot (expr sp) expr f =
  convertpath (g) f;
  if not cycle g:  NoCycle("polkadot") g;
  elseif sp <= emax (2*polkadotwd/3, mindotspace):
    safefill g;
  else:
    setbbox (ll, ur) g;
%    \end{macrocode}
% As with \gbc{shade}, we shift alternate rows by half the spacing between
% dot centers. The vertical shift is slightly larger (relatively speaking)
% and the horizontal smaller. We apply a further horizontal and vertical
% shift to avoid small pieces of dots. What it does is take only those dots
% whose centers lie in the bounding box, and center the whole array
% relative to that box.
%    \begin{macrocode}
    save dx, dy, dshift;  pair dshift;
    dx := sp/2;  dy := dx*sqrt 3;
    dshift := (xpart(ur - ll) mod dx, ypart (ur - ll) mod dy)/2;
%    \end{macrocode}
% Here, \gbc{p} is the center of the first dot in the lower left corner.
%    \begin{macrocode}
    save p, dims;  pair p, dims;
    p := ll + dshift;
    dims :=  2(dx, dy);
%    \end{macrocode}
% The extra \MF{} code is to clear what's under the dots in case they
% are gray dots. And then to `gray' the dots when \gbc{fillcolor} demands
% it.
%    \begin{macrocode}
    setpicture (thepolkadot) setdot (polkadotpath, polkadotwd);
    newpicture (v);
    fillwith (v) (thepolkadot, dims, p, ur);
    fillwith (v) (thepolkadot, dims, p + (dx, dy), ur);
    DoClip (v);  clipto (v) g;
%<*MF>
    if isgray (fillcolor):
      _subto (active_plane) (v);
      v := nullpicture;
      thepolkadot :=
        shaded (fillcolor) polkadotpath scaled ceiling (polkadotwd);
      fillwith (v) (thepolkadot, dims, p, ur);
      fillwith (v) (thepolkadot, dims, p + (dx, dy), ur);
      DoClip (v);  clipto (v) g;
    fi
%</MF>
    coloraddon (fillcolor, v);
  fi
  f
enddef;

%    \end{macrocode}
%
% \subsection{Hatching}\label{hatching}
%
% \DescribeRoutine{thatch}
% \RoutineIndex{colorthatch}
% This command hatches the interior of path \gbc{f} (graph coordinates)
% with lines at angle \gbc{theta}, spaced \gbc{sp} apart (device
% coordinates). As usual an unclosed path is simply drawn. The thickness
% of the lines is determined by \gbc{hatchwd}. If \gbc{sp} is not greater
% than \gbc{abs(hatchwd)}, we simply fill. This will ensure \gbc{thatchf}
% is called only for positive \gbc{sp}.
%
% We find the bounding box of the backward rotated path, so when that box
% is filled with lines and rotated, it will cover the path. After calling
% \gbc{thatchf} we add the picture, clipped to the path.
%    \begin{macrocode}
def thatch =  colorthatch (hatchcolor)  enddef;
vardef colorthatch (expr clr) (expr sp, theta) expr f =
  convertpath (g) f;
  if not cycle g:  NoCycle("hatch") g;
  elseif sp <= abs(hatchwd):  colorsafefill (clr) g;
  else:
    newpicture (v);
    setbbox (ll, ur) g rotated -theta;
    thatchf (v, identity rotated theta, sp, ll, ur);
    DoClip (v);  clipto (v) (g);
    coloraddon (clr, v);
  fi
  f
enddef;

%    \end{macrocode}
%
% We offer some special cases, calling \gbc{thatch} with different angles.
% These take only the spacing (in device coordinates) and a path
% expression (in graph coordinates) as parameters.\\
% \DescribeRoutine{hhatch}
% \gbc{hhatch} has angle 0 and so produces horizontal lines;\\
% \DescribeRoutine{vhatch}
% \gbc{vhatch} produces vertical lines;\\
% \DescribeRoutine{lhatch}
% \gbc{lhatch} produces lines tilted to the left (running from upper left
% to lower right);\\
% \DescribeRoutine{rhatch}
% \gbc{rhatch} produces lines running from lower left to upper right;
% and\\
% \DescribeRoutine{xhatch}
% \gbc{xhatch} produces cross-hatching, and essentially runs \gbc{lhatch}
% and \gbc{rhatch}.
%
% Color is a parameter only for \gbc{colorxhatch}. The reason for that
% is to make code written by \mfpic{} simpler. The \mfpic{} commands for
% the others actual write calls to \gbc{thatch} or \gbc{colorthatch}.
%
%    \begin{macrocode}
def hhatch (expr sp) =  thatch (sp, 0)   enddef;
def vhatch (expr sp) =  thatch (sp, 90)  enddef;
def lhatch (expr sp) =  thatch (sp, -45)  enddef;
def rhatch (expr sp) =  thatch (sp, 45)  enddef;

def xhatch =  colorxhatch (hatchcolor)  enddef;
def colorxhatch (expr clr, sp) =
  colorthatch (clr) (sp, 45) colorthatch (clr) (sp, -45)
enddef;

%    \end{macrocode}
%
% \subsection{Tesselations}\label{tess}
%
% \DescribeRoutine{tess}
% Tesselation of the interior of a closed path means filling with copies
% of a \emph{tile} (see subsection~\ref{tiles}). The path is in graph
% units, the tile is a suffix parameter and is the name of a previously
% defined tile. In fact, one can create the picture any way one likes (it
% doesn't have to be with the \gbc{tile} environment). Thus \gbc{tess
% (fred) f;} will work as long as \gbc{fred.pic} is a picture and
% \gbc{fred.dims} is a pair giving its dimensions.
%    \begin{macrocode}
vardef NoTile (suffix atile) expr g =
  GBwarn str atile & " is not a valid tile for tess()."
                   & " The path will be drawn instead.";
  safedraw g;
enddef;
vardef tess (suffix atile) expr c =
  convertpath (_g) c;
  if not cycle _g:  NoCycle("tess") _g;
  elseif not is_tile (atile):  NoTile (atile) _g;
  else:
    setbbox (_ll, _ur) _g;
    newpicture (_ts);
    fillwith (_ts) (atile.pic, atile.dims, _ll, _ur);
    DoClip (_ts);  clipto (_ts) _g;
    _orto (active_plane, _ts);
  fi
  c
enddef;

%    \end{macrocode}
%
% \subsection{Dots and dashes}\label{dashes}
%
% \MP{} already has commands for drawing a dashed or dotted curve,
% but \MF{} does not. Considerable effort went into making this possible
% (before \MP{} even existed). The code is now reasonably fast and the
% result is actually better quality than \MP{}'s native commands so we use
% the same code in both versions. It does, however, use pretty much
% memory in \MP{}.
%
% The \grafbase{} dashing code is designed to produce a whole number of
% dashes on any curve to which it is applied, and (usually) to begin and
% end with half a dash (so that when dashed curves abut, the result looks
% decent). \MP{}'s own facilities do neither of these. In addition, the
% dotting code is flexible enough that copies of any picture (not just a
% circular dot) can be used to trace a path.
%
% The general command is \gbc{gendashed}, which takes a suffix parameter
% (the name of a \emph{dashing pattern}, see below) and a path expression
% in graph coordinates.
%
% A dashing pattern \gbc{pat} consists of three arrays, \gbc{pat.start},
% which is used to draw the beginning of the path (half a dash in the
% default \gbc{dashed} command), \gbc{pat.finish}, which is used to draw
% the other end, and \gbc{pat.rep}, which is the repeating pattern for
% drawing the rest of the curve. Each of these is an \emph{array} of
% numerics. These should be lengths, in device units, and represent the
% lengths of dashes and spaces.
%
% We start with some variables and their defaults, some of which are no
% longer used. \gbc{segment_split} is used in the code for finding the
% approximate length of a curve. This is needed to make adjustments in the
% length of dashes and spaces so that a whole number of repeated patterns
% are used. \gbc{dashsize} and \gbc{dashgap} are no longer used.
% Originally they gave the lengths of default dashes and the spaces in
% between. \gbc{dash_start} and \gbc{dash_finish} are the fractions of a
% dash length that are used at the start and finish if the command
% \gbc{dashpat} is used to create the dashing pattern.
%
% And \gbc{_rescale_factor} is used to adjust numbers downward and avoid
% arithmetic overflow. For a 1200dpi \MF{} mode, a curve 4 inches long
% will be over \mfc{infinity} pixels in length, but only 40 deci-inches.
% Our default for this variable is just that: 1/10 inch.
%    \begin{macrocode}
if unknown segment_split:  segment_split := 8;  fi
if unknown dashsize:  dashsize  := 3bp;  fi
if unknown dashgap:  dashgap := dashsize + 2penwd;  fi
if unknown dash_finish:  dash_finish := .5;  fi
if unknown dash_start:  dash_start := .5;  fi
if unknown _rescale_factor:  _rescale_factor := 0.1in;  fi

%    \end{macrocode}
%
% \DescribeRoutine{gendashed}
% The main idea is to have a list of lengths represent the repeating
% pattern of dashes and dots. These lengths represent a dash length,
% followed by a gap length, etc., so there are an even number. To start
% dashing a path, we normally take a fraction (\gbc{dash_start}) of the
% first dash, then the rest of the pattern. We continue by repeating the
% pattern as many times as will fit, then we finish off with a fraction
% (\gbc{dash_finish}) of the first dash. A dash of length 0 is a dot. A
% gap of length 0 is OK, but useless unless it's between a dot and a dash,
% and you arrange for the dot's size to be different from \gbc{penwd}.
%
% We generalize this so that \gbc{pat.start} and \gbc{pat.finish} can be
% any patterns, not necessarily related to \gbc{pat.rep}. Also `dots' can be
% symbols like \gbc{Triangle}.
%
% When we tried to deal with arrays of dashing patterns, it became
% rather a pain to deal with three arrays of arrays. So now we allow the
% suffix \gbc{pat} to be a single array and call \gbc{mkdasharrays} to
% produce \gbc{pat.start} and \gbc{pat.finish}. It returns \gbc{true} if
% all three arrays are successfully produced.
%
% The variable \VariableIndex{last_dot_size} \gbc{last_dot_size} is
% intended to allow the clearing path of arrowhead commands to encompass
% a final dot larger than \gbc{penwd}.
%    \begin{macrocode}
numeric last_dot_size;  last_dot_size := 0;
vardef gendashed (suffix pat) expr f =
  convertpath (_g) f;
  save _dpat;
  if not mkdasharrays (pat) (_dpat):
    GBwarn "Dash pattern " & str pat
      & " undefined. Path will be drawn instead.";
    safedraw _g;
  elseif _dpat.rep < 2:
    safedraw _g;
  else:
%    \end{macrocode}
%
% After the following loop, \gbc{_dl.s} is the total length of the
% corresponding \gbc{pat.s} in multiples of \gbc{_rescale_factor}, and
% \gbc{_dpat.s[i]} has been converted to these units.
%    \begin{macrocode}
    save _dl;
    forsuffixes _s = start, rep, finish:
      _dl._s := 0;
      for i = 1 upto _dpat._s:
        _dpat._s[i] := _dpat._s[i]/_rescale_factor;
        _dl._s := _dl._s + _dpat._s[i];
      endfor
    endfor

    if _dl.rep = 0:
      GBwarn "Dash pattern " & str pat & " has length 0. "
        & "Path will be drawn instead.";
      safedraw _g;
    else:
%    \end{macrocode}
% Here \gbc{_g} is our path in device units, but \gbc{setuplengtharray}
% computes lengths in multiples of \gbc{_rescale_factor} to avoid having
% paths of length \gbc{infinity}.
%
% This is how we process a path mathematically: let $f(t)$, $0 \le t \le
% k$ be the formula for the path \gbc{f}, $k$ being the number of segments
% of \gbc{f},  we consider the polygon connecting the points $f(0), f(1/s),
% f(2/s),\ldots,f(k)$ (where $s$ is \gbc{segment_split}) and compute the
% length of \emph{that} path. Actually, we compute and save the cumulative
% lengths at each vertex of this polygon, since we use that later to
% determine `when' (i.e., at what values of $t$) to place a dot or draw
% a dash. The command \gbc{setuplengtharray} does this, storing the
% cumulative lengths in the array \gbc{_cumlen} and the total in
% \gbc{_totlen}. It also initializes \gbc{_ct} the index into that array.
%    \begin{macrocode}
      setuplengtharray (_cumlen, _totlen, _ct) _g;
%    \end{macrocode}
% Now we adjust the dashes so that a whole number of patterns make up
% the lengths of the approximating polygon. \gbc{scale_adjust} returns
% the scaling factor, equates \gbc{_n} to the total number of
% \gbc{pat.rep} to use. If the path length is too small compared to the
% length of the start and finish patterns, this is equated to $-1$ as a
% flag to draw the path instead. (recall \gbc{_dl.s} holds the length of
% part \gbc{s}).
%
% After this we rescale the dashes and spaces stored in \gbc{_dpat}, and
% the length of the patterns in \gbc{_dl}.
%    \begin{macrocode}
      save _n, _sf, _no_dots;
      boolean _no_dots;  _no_dots := true;
      _sf := scale_adjust (_n, _dl) (_totlen);
      if _n < 0:  safedraw _g;
      else:
        forsuffixes _s = start, rep, finish:
          for _i = 1 upto _dpat._s:
            if (_dpat._s[_i] = 0) and _no_dots:  _no_dots := false;
            else:  _dpat._s[_i] := _dpat._s[_i]*_sf;
            fi
          endfor
          _dl._s := _dl._s*_sf;
        endfor
%    \end{macrocode}
% The user has the capability to use something other than a small disk for
% a dot by defining \gbc{plot_pic} (and preferably also storing its
% diameter in \gbc{plot_pic.size}). The utility \gbc{makesymbol} is
% defined later. It examines \gbc{plot_pic} and makes a picture depending
% on what type of variable it is. The default \gbc{dotpath} is
% \mfc{fullcircle}, but user may also change that to get different dots.
% \gbc{makesymbol} scales by \gbc{penwd} \emph{only if the first
% parameter is a path}. This is how to increase the dot size (the code in
% \gbc{plot} uses this.)
%    \begin{macrocode}
        if _no_dots:
        else:
          if unknown plot_pic:
            save plot_pic;  path plot_pic;
            plot_pic := dotpath;
          fi;
          last_dot_size :=
            if known plot_pic.size:  plot_pic.size  else:  penwd fi;
          setpicture (dashingdot) makesymbol (plot_pic, last_dot_size);
        fi
%    \end{macrocode}
% The macro \gbc{dashit} draws the dashes, computing where they go and
% drawing the appropriate subpaths of \gbc{_g} or placing a dot at the
% appropriate point. \gbc{dashit} returns nothing and assumes all the
% information accumulated so far, so it can only be called by
% \gbc{gendashed}.
%
% \gbc{_t} and \gbc{_d} are temporary variables used by
% \gbc{dashit}, but we declare them here since we initialize them
% differently for each call. \gbc{_d0} and \gbc{_d1} hold the
% position along the curve of the ends of a dash in distance from the
% start; \gbc{_t0} and \gbc{_t1} are the same, but in terms of time.
% A macro \gbc{gettime} converts the first to the second. It uses the
% cumulative length array \gbc{_cumlen} for this, and maintains
% \gbc{_ct} as the current index into that array.
%    \begin{macrocode}
        save _t, _d, _v;
        picture _v;  _v := nullpicture;
        _d0 := 0;  _t0 := 0;
        dashit (_dpat.start) (_v);
%    \end{macrocode}
% The parameters to \gbc{dashit} are the name of the part of the dashing
% pattern that is being drawn, and a temporary picture variable. The
% latter holds the picture until \gbc{DoClip} can process it.
% The code of \gbc{dashit} leaves \gbc{_d0} pointing to the current
% position on the curve, but for safety and to reduce accumulated
% round-off error, we initialize it to what it should be before each call.
%
% The repeating pattern has the tendency to use lots of memory. Previously
% I added all the dashes to \gbc{_v} and then added it all at once to
% \gbc{active_plane}. The purpose was to be able to \gbc{DoClip} it once,
% and add it once with \gbc{coloraddon} to get it drawn in color under \MF.
% This was simplest, but a memory hog requiring $O(n)$ in memory, where
% $n$ is the number of repeated patterns. Then we tried clipping and adding
% within \gbc{dashit}. This was terribly slow, requiring $O(n)$ in time.
% Now we use a standard programming trick: accumulate $m < n$ repetitions
% before adding them, the memory should be $O(m)$ and the time $O(n/m)$.
% Making $m$ about $\sqrt n$ seems to work well.
%    \begin{macrocode}
        if _n > 0:
          save _m;  _m := ceiling sqrt(_n);
          for _j = 0 step _m until _n - 1:
            for _i = 0 upto _m - 1:
              exitif (_i + _j) > _n - 1;
              _d0 := _dl.start + (_j + _i)*_dl.rep;
              _t0 := gettime (_cumlen, _ct) (_d0);
              dashit (_dpat.rep) (_v);
            endfor
            DoClip (_v);
            coloraddon (drawcolor, _v);
            _v := nullpicture;
          endfor
        fi

        _d0 := _totlen - _dl.finish;
        _t0 := gettime (_cumlen, _ct) (_d0);
        dashit (_dpat.finish) (_v);
        DoClip (_v);
        coloraddon (drawcolor, _v);
      fi
    fi
  fi
  f
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{makelengtharray}
% This takes an array name and a path expression (which is assumed to be
% in device coordinates), computes the array of partial lengths (of the
% polygon approximation), and returns the total length. To avoid numeric
% overflow we rescale the lengths and so the array elements are in units
% of \gbc{_rescale_factor}. At one point we used to rescale the path,
% but that turned out to be unnecessary and made it harder to accomplish
% several of the things we now do with this. We also save a little
% memory by making the path a suffix parameter, which avoids the memory
% used for the `capsule' of an expression parameter. So far all uses apply
% it to a path variable and we have to remember to keep it that way.
%    \begin{macrocode}
vardef makelengtharray (suffix clen) suffix p =
  setsplit (_s) segment_split;
  numeric clen[];
  clen := _s * length p;  clen0 := 0;
  for _i = 1 upto clen:
    clen[_i] := clen[_i-1] + abs (pnt[_i/_s] (p) - pnt[(_i-1)/_s] (p)) /
        _rescale_factor;
  endfor
  clen[clen]
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{scale_adjust}
% Here \gbc{n} is a suffix defined by the calling routine,
% \gbc{pl.\{start\|rep\|finish\}} are the lengths of corresponding parts
% of a dashing pattern, \gbc{lngth} is the length of some path (determined
% by the calling routine). It determines how many times \gbc{pl.rep} goes
% into \gbc{lngth - pl.start - pl.finish}. If this is negative it remains
% negative, otherwise it is rounded.  \gbc{scale_adjust} then determines
% and returns the scaling factor \gbc{sf} required to make
% \gbc{sf*(pl.start + n*pl.rep + pl.finish)} equal to \gbc{lngth}.
%    \begin{macrocode}
vardef scale_adjust (suffix n, pl) (expr lngth) =
  n := (lngth - pl.start - pl.finish)/pl.rep;
  n := if n < 0:  -1  else:  round(n) fi;
  lngth/(pl.start + emax (n, 0)*pl.rep + pl.finish)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{gettime}
% \gbc{arr} is an increasing array of lengths, defined by the calling
% routine. \gbc{ct} is current index into that array; it will vary with
% subsequent calls. Calling routine initializes it before the first call,
% \gbc{gettime} updates it. \gbc{lngth} is a length interpreted as the
% length along the path associated to the array.
%
% Since this array is generated by splitting the segments of the path at
% times \gbc{i/segment_split} we first determine in which of these splits
% the given distance is (i.e., find \gbc{ct} so that \gbc{lngth} lies
% between \gbc{arr[ct-1]} and \gbc{arr[ct]}). To avoid problems with
% round-off error, bad length parameter, etc., we force \gbc{lngth} to
% satisfy this for some index between the current value of \gbc{ct} and
% \gbc{arr} inclusive.
%
% Once we know what segment we are in, we determine the time by linear
% interpolation between the times corresponding to \gbc{ct} and
% \gbc{ct+1}. Note: in the \mfc{forever} loop, the exit must come before
% the increment. The function \gbc{inrange} is defined in
% section~\ref{axes}. It checks if the third argument is between the
% first two, or equal to one of them.
%    \begin{macrocode}
vardef gettime (suffix arr, ct) (expr lngth) =
  setnumeric (_gtl) emax (arr[ct], emin (arr[arr], lngth));
  setsplit (_s) segment_split;
  forever:  exitif inrange (arr[ct], arr[ct+1]) (_gtl);
    next ct;
  endfor
  if arr[ct] = arr[ct+1]:  ct
  else:  ( ct + (_gtl - arr[ct]) / (arr[ct+1] - arr[ct]) )
  fi  /_s
enddef;

def next suffix X =  X :=  X + 1;  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{dashit}
% No variables are saved or initialized; \gbc{gendashed} defines array
% \gbc{_cumlen}, path \gbc{_g}, and initializes \gbc{_d0},
% \gbc{_t0} and \gbc{_ct}.
%
% \gbc{pos} is one of the dashpattern arrays, so it consists of numerics
% interpreted as lengths of dashes (odd index) and spaces (even index). In
% the first case \gbc{_d0} and \gbc{_t0} will already be pointing to
% the beginning of the dash and we get to the end of the dash by adding
% the length of a dash (\gbc{pos[_j]}) to \gbc{_d0} (getting
% \gbc{_d1}) and calling \gbc{gettime} (getting \gbc{_t1}). We draw
% the subpath between those points. Unless \gbc{pos[_j] = 0}, in which case
% a dot is placed.
%
% For even \gbc{j} (a space) we are at \gbc{_d1} and \gbc{_t1} and
% we increment them to get \gbc{_d0} and \gbc{_t0} for the next
% iteration.
%    \begin{macrocode}
def dashit (suffix pos) (suffix pic) =
  for _k = 1 upto pos:
    if odd _k:
      if pos[_k] = 0:
        _d1 := _d0;  _t1 := _t0;
        picdot (pic, dashingdot, pnt [_t0] (_g));
      else:
        _d1 := _d0 + pos[_k];
        _t1 := gettime (_cumlen, _ct) (_d1);
        shpath (pic, drawpen) (subpath (_t0, _t1) of _g);
      fi
    else:
      _d0 := _d1 + pos[_k];
      _t0 := gettime (_cumlen, _ct) (_d0);
    fi
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{dashpat}
% This is a utility to convert a list of lengths to an array. It is almost
% the same as \gbc{list}, but it does make sure the array functions as
% a dashing pattern. It needs to consist of either $1$ item (the signal to
% draw a solid line) or an even number of items. So we add a zero length
% term if the size is $0$ or odd and bigger than $1$.
%
% \DescribeRoutine{mkdasharrays}
% The dashing code in \gbc{gendashed} is written so the the beginning and
% ending patterns can be different from the repeating patterns. If so,
% they must be named \gbc{pat.start}, \gbc{pat.rep} and \gbc{pat.finish}.
% If one of these three is not a known array but \gbc{pat} is, this macro
% creates the array in a generic way. When created this way, \gbc{pat.rep}
% is a copy of \gbc{pat}. So is \gbc{pat.start}, except only a fraction of
% the first dash is used, while \gbc{pat.finish} is just the first dash of
% \gbc{pat} reduced by the factor \gbc{dash_finish}.
%    \begin{macrocode}
def dashpat (suffix pat) (text t) =
  list (pat) (t);
  if (pat = 0) or (odd (pat) and (pat > 1)):
    pat[incr pat] := 0;
  fi
enddef;

vardef mkdasharrays (suffix src, dest) =
  save _bad;  boolean _bad;  _bad := false;
  forsuffixes _s = start, rep, finish:
    numeric dest._s, dest._s[];
    boolean _bad._s;
    if knownnumericarray src._s:
      copyarray (src._s) (dest._s);
      _bad._s := false;
    else:
      _bad := _bad._s := true;
    fi
  endfor
  % _bad = one of the three arrays not copied.
  if _bad:
    if knownnumericarray src:  _bad := false;
      if _bad.rep:      % make dest.rep = src
        copyarray (src) (dest.rep);
      fi
      if _bad.start:    % shrink first dash to get dest.start
        copyarray (src) (dest.start);
        dest.start1 := dash_start*src1;
      fi
      if _bad.finish:   % use partial first dash for dest.finish
        dest.finish := 1;
        dest.finish1 := dash_finish*src1;
      fi
    fi
  fi
  not _bad
enddef;

%    \end{macrocode}
%
% The \mfpic{} command \cs{dashed} is now implemented by making a
% dashpattern from the two arguments and calling gendashed. That is the
% definition of \gbc{DASHED}.
%
% \gbc{dashed} takes parameters which are the length and the space (device
% coordinates) and a path (graph coordinates). It returns the path.
%    \begin{macrocode}
vardef Dashed (expr dlen, dgap) expr f =
  save dashes;  dashpat (dashes) (dlen, dgap);
  gendashed (dashes) f
enddef;

def DASHED =  Dashed  enddef;
%<MF> def dashed =  Dashed  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{doplot}
% \gbc{doplot} places symbols at positions along a path determined by
% \gbc{dgap} (space between symbols), they are scaled by \gbc{sc} and the
% actual symbol is \gbc{spath}. Currently this may be one of three things:
% \begin{enumerate}
%  \item A path, giving the shape of the dot, which should be defined in
%     units so that the desired size is obtained under scaling by
%     \gbc{sc}. Normally this means one unit across.
%  \item A picture. This is used unscaled, it being presumed that it has
%     been prepared by a user to the correct size.
%  \item (\MP{} only) a string.
% \end{enumerate}
% All these are converted to a picture by the \gbc{makesymbol} command and
% it is assigned to \gbc{plot_pic}, which \gbc{gendashed} has been
% trained to use when dots are needed.
%
% After this \gbc{gendashed} is called with a pattern where the dashes are
% 0 length, the signal that dots are to be used.
% \DescribeRoutine{dotted}
% The macro \gbc{dotted} is implemented by calling \gbc{doplot} with
% \gbc{dotpath} as the the symbol.
%    \begin{macrocode}
vardef doplot (expr spath, sc, dgap) expr f =
  save dots;  dashpat (dots) (0, dgap);
  setpicture (plot_pic) makesymbol (spath, sc);
  plot_pic.size := sc;
  gendashed (dots) f
enddef;

path dotpath;  dotpath := fullcircle;
def dotted =  doplot (dotpath)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{plotnodes}
% These are useful little utilities to draw the points on top of the
% curve through them. \gbc{plotnodes} differs from \gbc{plotsymbol}
% (defined later) in that it takes a path parameter (rather than a list of
% points) and returns that path (so it works with \mfpic{} as a prefix
% macro). It also uses \gbc{drawcolor}. Otherwise it calls the same code.
%
% \DescribeRoutine{showcontrols}
% \gbc{showcontrols} was mainly for debugging; it draws a line segment
% connecting the control points of each node. It optionally draws a symbol
% at each control point. We use \mfc{0} for a `symbol' to indicate we
% don't wish to draw a symbol there. The default color for both the
% symbols and the segment is \gbc{pointcolor}.
%    \begin{macrocode}
vardef plotnodes (expr symbol, size) expr f =
  if size > 0:
    save pln;  pair pln[];
    pln := 0;
    for _a = 0 upto (length f) if cycle f:  - 1 fi:
      pln[incr pln] := pnt[_a] (f);
    endfor
    dosymbols (drawcolor, symbol, size) (pln);
  fi
  f
enddef;

def showcontrols =  colorshowcontrols (pointcolor)  enddef;
vardef colorshowcontrols (expr clr, syma, symb, size) expr f =
  save shpre, shpost;
  pair shpre[], shpost[];
  shpre := 0;  shpost := 0;
  for a = 0 upto (length f) if cycle f:  - 1 fi:
    shpre [incr shpre]  := pre [a] (f);
    shpost[incr shpost] := post[a] (f);
    colorsafedraw (clr)
      (zconv (shpre[shpre]--pnt[a](f)--shpost[shpost]));
  endfor
  if size > 0:
    if not numeric syma:  dosymbols (clr, syma, size) (shpre) ;  fi
    if not numeric symb:  dosymbols (clr, symb, size) (shpost);  fi
  fi
  f
enddef;

%    \end{macrocode}
%
% \subsection{Double-line drawing}\label{doubleline}
%
% \DescribeRoutine{doubledraw}
% This invokes \gbc{parapath}, which has issues with smooth but wiggly
% paths. An easier approach would be to draw a wide line and erase a
% narrow one in the middle. However, that would not be transparent in
% \MP{}.
%    \begin{macrocode}
def doubledraw =  colordoubledraw (drawcolor)  enddef;
vardef colordoubledraw (expr clr, sep) expr f =
  convertpath (g) f;
  colorsafedraw (clr) (parapath ( sep/2) g);
  colorsafedraw (clr) (parapath (-sep/2) g);
  f
enddef;

%    \end{macrocode}
%
% \section{Points Symbols and Other Pictures}\label{symbols}
%
% \DescribeRoutine{centerit}
% This accepts a picture and returns the same picture centered. This
% is close to impossible in \MF, so we only do it in \MP. Actually, we
% no longer use it, because in the one case where we did
% (\gbc{makesymbol}), it seemed to restrict the user's choices too much.
%    \begin{macrocode}
%<*MP>
vardef centerit (expr pic) =
   pic shifted -(0.5[urcorner pic, llcorner pic])
enddef;

%</MP>
%    \end{macrocode}
%
% \DescribeRoutine{makesymbol}
% This utility takes \emph{any} expression and scale and returns a picture.
% If the expression \gbc{spath} is a cycle it returns the interior, for
% other paths, a drawing of the path, in either case scaled by \gbc{sc}.
% If already a picture, it returns it. In \MP, if it is a string, it
% returns a picture containing that string drawn in the \mfc{defaultfont}.
% In any other case, the default dot is returned.
%    \begin{macrocode}
vardef makesymbol (expr spath, sc) =
  if picture spath :
%<MF>    setpicture (v) spath;  mono (v);  v
%<MP>    spath
  elseif path spath:  setdot (spath, sc)
%<MP>  elseif string spath:
%<MP>    spath infont defaultfont scaled defaultscale
  else:
    GBwarn "Undefined symbol for plotting, "
         & "dotpath will be used instead.";
    setdot (dotpath, sc)
  fi
enddef;

%    \end{macrocode}
%
% Points are filled or unfilled circles. They are implemented with
% \gbc{plotsymbol}, but the code differs in that filled or unfilled
% circles are determined by a parameter rather than the type of curve.
% In addition, for unfilled circles, it clears the pixels inside the circle.
%
% \DescribeRoutine{bpoint}
% \gbc{bpoint} is basicly a shorthand for a scaled circle shifted to a
% point. The scale and the point are in device coordinates. We don't use
% it anywhere in \grafbase{} anymore.
%    \begin{macrocode}
vardef bpoint (expr ptwd, b) =
  fullcircle scaled ptwd shifted b
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{pointd}
% This draws disks with diameter \gbc{ptwd}, filled or unfilled based on
% the boolean \gbc{filled}, at the graph coordinate coordinates in the
% list \gbc{t}. In case \gbc{filled} is true, \gbc{pointd} calls
% \gbc{plotsymbol (SolidCircle)} otherwise we make \gbc{clearsymbols} true
% (so that the area where each point is drawn will be cleared before
% drawing it) and call \gbc{plotsymbol (Circle)}.
%    \begin{macrocode}
def pointd (expr ptwd, filled) (text t) =
  if filled:
    plotsymbol (SolidCircle, ptwd) (t);
  else:
    begingroup;
      setboolean (clearsymbols) true;
      plotsymbol (Circle, ptwd) (t);
    endgroup
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{plotsymbol}
% \RoutineIndex{colorplotsymbol}
% The \gbc{plotsymbol} command places a symbol centered at each of the
% graph coordinate points in the list. The symbol placed is the first
% parameter, which would normally be a path, but can be a picture or, in
% \MP, a string. Like the \gbc{doplot} command, it calls \gbc{makesymbol}.
% If \gbc{spath} is of type path, and is cyclic, it is drawn filled. This
% is because we call \gbc{makesymbol} on it, and that subjects it to
% \gbc{setdot}, which has that behavior. For other types of symbols, we
% simply convert them to pictures with \gbc{makesymbol} and then place
% them. Unlike \gbc{pointd} above, the interior of the path is not erased
% by default. However, in the special case where the symbol is an open
% path, if its first point is equal to its last point, and
% \gbc{clearsymbols} is true, then the interior of the path obtained by
% \gbc{\& cycle} is cleared before the path itself is drawn. We copy the
% text list to an array and call \gbc{dosymbols} so that \gbc{plotnodes},
% \gbc{plotsymbol} and \gbc{showcontrols} can share the code.
%
% \DescribeRoutine{dosymbols}
% Since \gbc{dosymbols} uses identical code twice (once to clear, once to
% draw), we put that code in \gbc{addsymbols}.
% \DescribeRoutine{addsymbols}
% And finally, \gbc{addsymbols} draws copies of a symbol at a given array
% of points with a given color.
%    \begin{macrocode}
boolean clearsymbols;  clearsymbols := false;
vardef clearable (expr pth) =
  if path pth:
    ( pnt0 (pth) = pnt[length pth] (pth) ) and (not cycle pth)
      and (length pth > 0)
  else:  false
  fi
enddef;

def clearopenpath expr f =
  if clearable (f):  safeunfill f & cycle;  fi
enddef;

def plotsymbol =  colorplotsymbol (pointcolor)  enddef;
def colorplotsymbol (expr clr, spath, sc) (text t) =
  if sc > 0:
  begingroup
    setpairs (_cpls) (t);
    if _cpls > 0:  dosymbols (clr, spath, sc) (_cpls);  fi
  endgroup
  fi
enddef;

def dosymbols (expr clr, spath, sc) (suffix arr) =
  if clearsymbols and  clearable (spath):
    addsymbols (background, makesymbol (spath&cycle, sc)) (arr);
  fi
  addsymbols (clr, makesymbol (spath, sc)) (arr);
enddef;

def addsymbols (expr clr, symb) (suffix arr) =
  newpicture (_pls);
  for _idx = 1 upto arr:
    picdot (_pls, symb, zconv (arr[_idx]));
  endfor
  DoClip (_pls);  coloraddon (clr, _pls);
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{putimage}
% This is designed to allow \mfpic{} users to make a picture (created with
% \cs{mfpimage} or \cs{tile} perhaps), and put a copy at several
% locations. This allows more complex things than \cs{plotsymbol} and
% more flexibility than \cs{tess}. As the picture is should be completely
% prepared in advance, there is no color or size parameter. Moreover,
% \gbc{pic} might be quite complex, so we don't use \gbc{picdot} which
% takes the added picture as an expression, but rather repeat its code.
% The indirection of adding to \gbc{_pti} is normal when we want to
% respect clipping without clipping what is already drawn. Saving memory
% at some sacrifice of speed, we clip and reset with each addition. (It is
% uncertain whether adding multiple clipped pictures includes the clipping
% path for each addition. I'll have to test the memory use of this code
% versus adding all the copies to \gbc{_pti} and clipping once.)
%    \begin{macrocode}
def putimage (suffix pic) (text t) =
  newpicture (_pti);
  for _itm = t:
    addto _pti also
%<MF>      (pic shifted goodpair (zconv (_itm)));
%<MP>      (pic shifted zconv (_itm));
    DoClip (_pti);  addto active_plane also _pti;
    _pti := nullpicture;
  endfor
%<MF> mono active_plane
enddef;

%    \end{macrocode}
%
% \section{Axes, Tic Marks, and Grids}\label{axes}
%
% \DescribeRoutine{arrowdraw}
% This is used in \gbc{vectorfield} and to draw axes. It returns nothing.
% This doesn't follow the usual pattern of drawing a path and returning
% it. This approach makes the old \cs{axes}, \cs{xaxis} and \cs{yaxis}
% commands in \mfpic{} impossible to dash or dot. The newer axis drawing
% commands permit this and so use other code.
%
% We simply call \gbc{headpath} with default values, but add \gbc{drawn}
% to make sure the path is drawn, and precede it with \gbc{store} so
% \MF{} won't complain of an isolated expression. The new \mfpic{}
% method of drawing an axis is to apply \cs{arrow} to the path
% \gbc{axisline}. Thus it can also be \gbc{dashed}, \gbc{dotted},
% etc.
%
% The order is significant if axis and head are different colors. This
% order puts the head on top of the shaft.
%    \begin{macrocode}
def arrowdraw (expr hlen) (expr f) =
  store (curpath) headpath (hlen, 0, 0) drawn f;
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xaxis}
% The macro \gbc{xaxis} draws the $x$-axis through the point $(0,0)$ in
% graph coordinates. The only parameter is the length of the arrowhead in
% device coordinates.
% \DescribeRoutine{yaxis}
% The Macro \gbc{yaxis} draws the $y$-axis.
%
% \DescribeRoutine{axes}
% \gbc{axes} draws both axes with the same length of head.
%    \begin{macrocode}
def xaxis (expr hlen) =  arrowdraw (hlen) ((xneg, 0)--(xpos, 0));
enddef;
def yaxis (expr hlen) =  arrowdraw (hlen) ((0, yneg)--(0, ypos));
enddef;
def axes (expr hlen) =  xaxis (hlen);  yaxis (hlen);  enddef;

%    \end{macrocode}
%
% For axes at the borders of the graph coordinates, we allow for them to
% be shifted inwards. The amount of the shift is given by \gbc{laxis} for
% the left side axis, \gbc{baxis} for the bottom axis, etc. They are in
% graph coordinates.
%
% \DescribeRoutine{axisline}
% The commands \gbc{axisline.x}, etc., return the appropriate straight
% line at  the appropriate location. These are vardefs rather than
% variables so they can be affected by changing shift values.
%
% \DescribeRoutine{axis}
% Finally, the commands \gbc{axis.x}, etc. examine their suffix and
% apply \gbc{headpath} to the corresponding axis line. With a recent
% change in \mfpic{} code, it is no longer used there. Instead, code
% is written that allows the head to be drawn after the line is.
% Note it is a vardef, and so returns the line as a path..
%    \begin{macrocode}
laxis := baxis := raxis := taxis := 0;

vardef xlow  =  xneg + laxis  enddef;
vardef xhigh =  xpos - raxis  enddef;
vardef ylow  =  yneg + baxis  enddef;
vardef yhigh =  ypos - taxis  enddef;

vardef axisline.x =  (xlow, 0)--(xhigh, 0)  enddef;
vardef axisline.y =  (0, ylow)--(0, yhigh)  enddef;
vardef axisline.l =  axisline.y shifted (xlow, 0)  enddef;
vardef axisline.b =  axisline.x shifted (0, ylow)  enddef;
vardef axisline.r =  axisline.y shifted (xhigh, 0)  enddef;
vardef axisline.t =  axisline.x shifted (0, yhigh)  enddef;

vardef axis@# (expr len) =  headpath (len, 0, 0) axisline@#  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{borderrect}
% These are mostly for the simplification of \mfpic{} and readability of
% code. The command \gbc{borderrect} produces the border of the picture
% in graph coordinates, taking into account the four margins.
%
% \DescribeRoutine{between}
% The boolean \gbc{between} checks if the last argument is strictly
% between the first two (which must be in order).
%
% \DescribeRoutine{inrange}
% The boolean \gbc{inrange} checks if the last argument is in the closed
% interval determined by the first two (which must be in order).
%
% \DescribeRoutine{inbounds}
% The boolean \gbc{inbounds} checks if the argument (a pair) is in
% the closed border rectangle (\gbc{borderrect}). It is not yet used in
% \grafbase{}, though it would seem it ought to be useful.
%    \begin{macrocode}
vardef borderrect =
  rect((xlow,ylow),(xhigh,yhigh))
enddef;

vardef between (expr A, B, X) = (A <  X) and (X <  B)  enddef;
vardef inrange (expr A, B, X) = (A <= X) and (X <= B)  enddef;

vardef inbounds (expr Z) =
  inrange (xlow, xhigh) (xpart Z) and inrange (ylow, yhigh) (ypart Z)
enddef;

%    \end{macrocode}
%
% Possible binary relation versions. The last is just a reversal of the
% order of the first. These are not yet used in \grafbase{}.
%    \begin{macrocode}
tertiarydef X isbetween P =  between (xpart P, ypart P, X)  enddef;
tertiarydef X isinrange P =  inrange (xpart P, ypart P, X)  enddef;
tertiarydef P contains X  =  between (xpart P, ypart P, X)  enddef;

%    \end{macrocode}
%
% Tick marks can be on the inside or outside of a border axis,
% above or below any horizontal axes, left or right of any vertical axis
% or centered on any axis. The following numerics are merely used to
% convert the names to numeric code that the drawing routine will examine.
%
% However, it is no accident that \gbc{onbottom = onright} and that
% \gbc{centered} is halfway between \gbc{onright} and \gbc{onleft}. The
% code uses the numeric values to compute a shift, and one can supply an
% expression like \gbc{.33ontop+.67onbottom]} and then 1/3 of each mark
% will be above (and 2/3 will be below) the axis.
%
% The negative value of \gbc{inside} and \gbc{outside} is a flag that they
% are to be treated differently. The others have the property that the
% direction is the direction of the axis rotated a certain way (e.g.,
% $90$ degrees from \mfc{up} points \mfc{left}, $-90$ points \mfc{right}).
% But \gbc{inside} is right of the left axis and left of the right axis.
%    \begin{macrocode}
numeric inside, outside, centered, onleft, onright, ontop, onbottom;
inside   := -2;
outside  := -1;
onright  :=  1;
onleft   :=  2;
centered := .5[onright, onleft];
onbottom := onright;
ontop    := onleft;
%    \end{macrocode}
%
% We interact with \mfpic{} by allowing the user to change the value of
% \gbc{ltick}, for example, with a command like
% \cs{setaxismarks l}\marg{outside}. Here we set the defaults.
%    \begin{macrocode}
ltick := rtick := ttick := btick := inside;
xtick := ytick := centered;

%    \end{macrocode}
%
% \DescribeRoutine{axismarks}
% This utility macro draws the tick marks on an arbitrary axis. The
% different commands \gbc{xmarks}, etc., call this command with particular
% values of these parameters.
% \begin{itemize}
% \item \gbc{inang} is the direction one must rotate the axis to point
%     inside. This is always $\pm90$ degrees. The $x$-axis  and $y$-axis
%     are treated just like bottom and left axis in this respect.
% \item \gbc{tp}    is the tick position (e.g., \gbc{inside} or
%     \gbc{ontop}).
% \item \gbc{loc}   is the location of the 0-point of the axis (graph
%     coordinates).
% \item \gbc{pdir}  is \mfc{right} or \mfc{up}, indicating the positive
%     direction on the axis.
% \item \gbc{len}   is the length of a tick mark, supplied as an argument
%     to the individual axis mark commands.
% \item \gbc{t}     is the list of positions, also supplied.
% \end{itemize}
%    \begin{macrocode}
vardef axismarks (expr inang, tp, loc, pdir) (expr len) (text t) =
 save _tp, _U, _P, _tic, _ticang;
 pair _U, _P;  path _tic;
%    \end{macrocode}
% For \gbc{onleft}, \gbc{onright}, \gbc{ontop} or \gbc{onbottom}, which
% are positive, don't examine \gbc{inang} but for \gbc{inside/outside}
% use it to determine what inside means. \gbc{_ticang} will be the angle
% to rotate \gbc{pdir} to set the direction of the tic mark.
%
% Then we shift the numeric value of \gbc{tp} by one, so \gbc{centered}
% corresponds to $.5$ and the rest to either $0$ or $1$.
%    \begin{macrocode}
 _ticang := if tp < 0:  inang  else:  90 fi;
 _tp := abs(tp) - 1;
%    \end{macrocode}
% Except, we go through the following shenanigans so that the marks are
% always perpendicular to the axis, even if a coordinate transform will
% slant the axis. After this \gbc{_U} should point in direction of inside,
% onleft or ontop.
%    \begin{macrocode}
 _U := unitvector (vconv (pdir)) rotated _ticang;
%    \end{macrocode}
% Next, we use \gbc{_tp} to calculate the mark. For example, if
% \gbc{tp = inside}, then \gbc{_tp = 1}. Since \gbc{_U} points toward
% inside, \gbc{_tic} will go from \mfc{(0,0)} to a point a distance
% \gbc{len} in the direction of \gbc{_U}.
%    \begin{macrocode}
 _tic := (-_U--(0,0)) shifted (_tp*_U) scaled len;
%    \end{macrocode}
% Finally, for each numeric value in the list \gbc{t}, draw the tic
% shifted to the corresponding point on the axis.
%    \begin{macrocode}
 for _a = t:
   safedraw (_tic shifted zconv (loc + _a*pdir));
 endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xmarks}
% And now the specialized command for each axis. Inside and outside
% really make no sense for the $x$- and
% \DescribeRoutine{ymarks}
% \RoutineIndex{lmarks}
% \RoutineIndex{bmarks}
% \RoutineIndex{rmarks}
% \RoutineIndex{tmarks}
% $y$-axis, but since a bottom axis is usually used for $x$ and a left
% axis for $y$, we give \gbc{xmarks} the same first parameter as
% \gbc{bmarks} and \gbc{ymarks} the same as \gbc{lmarks}.
%    \begin{macrocode}
def xmarks =  axismarks ( 90, xtick, origin, right)  enddef;
def ymarks =  axismarks (-90, ytick, origin, up)    enddef;
def lmarks =  axismarks (-90, ltick, (xlow, 0), up)    enddef;
def bmarks =  axismarks ( 90, btick, (0, ylow), right)  enddef;
def rmarks =  axismarks ( 90, rtick, (xhigh, 0), up)    enddef;
def tmarks =  axismarks (-90, ttick, (0, yhigh), right)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{vargrid}
% \RoutineIndex{vgrid}
% Mainly for the purpose of visualising coordinates, \gbc{vargrid}
% draws a dot of size \gbc{dsize} at every point whose coordinates
% are are \gbc{(n*xsp, m*ysp)}, \gbc{n} and \gbc{m} being integers.
% \gbc{dsize} is in device coordinates, the spacings are in graph
% coordinates.
% \DescribeRoutine{grid}
% The macro \gbc{grid} is for backward compatibility, calling
% \gbc{vargrid} with a default \gbc{dsize} of \mfc{.5bp}. The old name
% \gbc{vgrid} incorrectly suggests a connection to \gbc{vgridlines}.
%    \begin{macrocode}
path griddotpath;  griddotpath := fullcircle;
def grid =  vargrid (0.5bp)  enddef;
vardef vargrid (expr dsize, xsp, ysp) =
  save gdot, gridpic;  picture gdot, gridpic;
  gdot := setdot (griddotpath, dsize);
  gridpic := nullpicture;
  for n = ceiling ((xlow)/xsp) upto floor ((xhigh)/xsp):
  for m = ceiling ((ylow)/ysp) upto floor ((yhigh)/ysp):
    picdot (gridpic, gdot, zconv ((n*xsp, m*ysp)));
  endfor
  endfor
  coloraddon (pointcolor, gridpic);
enddef;
def vgrid =  vargrid  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{gridlines}
% This is more what I think of when I hear `grid', but the name was
% already taken. The macro \gbc{gridlines} draws horizontal and vertical
% lines through all the points that \gbc{grid} would draw.
% \DescribeRoutine{hgridlines}
% The macro \gbc{hgridlines} draws only the horizontal lines through the
% same points, while
% \DescribeRoutine{vgridlines}
% \gbc{vgridlines} draws only vertical lines.
%    \begin{macrocode}
def hgridlines (expr ysp) =
  for n = ceiling ((ylow)/ysp) upto floor ((yhigh)/ysp):
    safedraw zconv ((xlow, n*ysp)--(xhigh, n*ysp));
  endfor
enddef;
def vgridlines (expr xsp) =
  for n = ceiling ((xlow)/xsp) upto floor ((xhigh)/xsp):
    safedraw zconv ((n*xsp, ylow)--(n*xsp, yhigh));
  endfor
enddef;
def gridlines (expr xsp, ysp) =
  vgridlines (xsp);  hgridlines (ysp);
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{vectorfield}
% This command produces a field of arrows from a pair-valued formula (text
% parameter \gbc{fcn}) in a region described by a boolean-valued
% expression (text parameter \gbc{cond}). This routine simply makes
% functions (\mfc{vardef}\,s) out of the expressions and calls
% \DescribeRoutine{mkvectorfield}
% \gbc{mkvectorfield}, which steps through the points described by
% \gbc{xsp} and \gbc{ysp} and places an arrow (actually, any path) at
% each. The arrow path is given by the function \gbc{vf}. The arrow is
% placed at the point only if the function \gbc{isOK} returns true. It
% also omits points that lie in the axis margins.
%
% \DescribeRoutine{plrvectorfield}
% The polar version differs only in the distribution of the arrows. They
% are placed at regular intervals of $r$ an $\theta$. The text parameters
% should be expressions in \gbc{r} and \gbc{t}, but are otherwise the
% same. In particular, \gbc{fcn} should return \MF{} pairs, not polar
% coordinate pairs. The function \gbc{polar} can be used to convert if
% necessary. Its code is very similar, except for the boolean code needed
% to keep the vectors within the bounds of the graph.
% \DescribeRoutine{mkplrvectorfield}
% It calls \gbc{mkplrvectorfield}, which is a lot like the non-polar
% version, except it first calculates the extremes of the polar variables
% with \gbc{getpolarbounds} and relies on the boolean to keep it out of
% the axis margins.
%    \begin{macrocode}
def vectorfield (expr len, xsp, ysp) (text fcn) (text cond) =
  save _vf, _is_OK;
  vardef _vf (expr x,y) =  ((0,0)--(fcn)) shifted (x,y)  enddef;
  vardef _is_OK (expr x,y) =  cond  enddef;
  mkvectorfield (len, xsp, ysp) (_vf, _is_OK);
enddef;

vardef mkvectorfield (expr len, xsp, ysp) (suffix vf, isOK) =
  for n = ceiling ((xlow)/xsp) upto floor ((xhigh)/xsp):
  for m = ceiling ((ylow)/ysp) upto floor ((yhigh)/ysp):
    if isOK (n*xsp,m*ysp):  arrowdraw (len) (vf(n*xsp,m*ysp));  fi
  endfor
  endfor
enddef;

def plrvectorfield (expr len, rsp, tsp) (text fcn) (text cond) =
  save _vf, _is_OK, _A, _B, _C, _D;
  _A := xlow;  _B := xhigh;
  _C := ylow;  _D := yhigh;
  vardef _vf (expr r,t) =  ((0,0)--(fcn)) shifted (r*dir t)  enddef;
  vardef _is_OK (expr r,t) =
    save _X, _Y;  _X := r*cosd t;  _Y := r*sind t;
    (cond) and between (_A, _B) (_X) and between (_C, _D) (_Y)
  enddef;
  mkplrvectorfield (len, rsp, tsp) (_vf, _is_OK);
enddef;

vardef mkplrvectorfield (expr len, rsp, tsp) (suffix vf, isOK) =
  save rmin, rmax, tmin, tmax;
  getpolarbounds;
  if rmin = 0:
    if isOK (0,tmin):  arrowdraw (len) (vf (0,tmin));  fi
    rmin := rsp;
  fi
  for n = ceiling (rmin/rsp) upto floor (rmax/rsp):
  for m = ceiling (tmin/tsp) upto floor (tmax/tsp):
    if isOK (n*rsp,m*tsp):  arrowdraw (len) (vf (n*rsp,m*tsp));  fi
  endfor
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{patcharcs}
% The macro \gbc{patcharcs} draws on a picture \gbc{X} the arcs
% \gbc{tstart}${}\le \theta \le{}$\gbc{tstop} with radii starting at
% \gbc{rstart}, stepping by \gbc{rstep} until \gbc{rstop}.
% \DescribeRoutine{patchrays}
% The macro \gbc{patchrays} draws the radial lines with $r$ coordinate
% varying between \gbc{rstart} and \gbc{rstop} at angles from \gbc{tstart}
% to \gbc{tstop} stepping by \gbc{tstep}.
% \DescribeRoutine{plrpatch}
% And \gbc{plrpatch} simply calls them both, and adds the resulting
% pictures to \gbc{active_plane}.
%    \begin{macrocode}
def patcharcs (suffix X) (expr rstart, rstop, rstep, tstart, tstop) =
  for rad = (if rstart = 0:  rstep  else:  rstart fi)
      step rstep until rstop:
    orto (X, picpath zconv (arcplr (origin, tstart, tstop, rad)) );
  endfor
enddef;
def patchrays (suffix X) (expr tstart, tstop, tstep, rstart, rstop) =
  for _ang = tstart step tstep until tstop:
    orto (X) (picpath zconv ((rstart*dir _ang)--(rstop*dir _ang)));
  endfor
enddef;

def plrpatch (expr rstart, rstop, rstep, tstart, tstop, tstep) =
begingroup
  newpicture (v);
  patcharcs (v) (rstart, rstop, rstep, tstart, tstop);
  coloraddon (drawcolor, v);
  v := nullpicture;
  patchrays (v) (tstart, tstop, tstep, rstart, rstop);
  coloraddon (drawcolor, v);
endgroup
enddef;

%    \end{macrocode}
%
% Polar coordinate grids are analogous to \gbc{gridlines} and \gbc{grid}.
% They first draw a grid large enough to cover the whole graph, then clip
% it to the graph boundaries. Since three of the four require
% calculating the dimensions of a polar coordinate patch that completely
% covers the graph rectangle, we isolate that code in
% \gbc{beginpolargrid}, defined later.
%
%
% \DescribeRoutine{gridarcs}
% \gbc{gridarcs} creates arcs having radii that are integer multiples of
% \gbc{rstep} and
% \DescribeRoutine{gridrays}
% \gbc{gridrays} draws radial lines at angles that are multiples of
% \gbc{tstep}.
% \DescribeRoutine{polargrid}
% The command \gbc{polargrid}simply calls the first two.
%
% \DescribeRoutine{polargridpoints}
% On the other hand, \gbc{polargridpoints} draws dots at the points where
% the lines and arcs in \gbc{poloargrid} would intersect. The `step'
% parameters are in graph coordinates. \gbc{beginpolargrid} also
% declares the picture variable \gbc{gridpic}, while \gbc{endpolargrid}
% clips the resulting picture and adds it to \gbc{active_plane}.
%
% The \gbc{rmin}, etc., returned by \gbc{beginpolargrid} are modified to
% fit the grid established by the step sizes. A ray could happen to be one
% of the graph's sides, so we use \mfc{ceiling} and \mfc{floor} which
% doesn't change integer values. However, the arc with radius \gbc{rmin}
% or \gbc{rmax} could touch the graph rectangle in at most 4 points, so we
% use \mfc{floor (x + 1)} and \mfc{ceiling (x - 1)} to start and stop
% before the edge of the graph.
%    \begin{macrocode}
def gridarcs (expr rstep) =
  beginpolargrid;
    if rmin = 0:
      picdot (gridpic, setdot (griddotpath, penwd), zconv (origin));
    fi
    rmin := rstep * floor (rmin/rstep + 1);
    rmax := rstep * ceiling (rmax/rstep - 1);
    patcharcs (gridpic) (rmin, rmax, rstep, tmin, tmax);
  endpolargrid (drawcolor, .5penwd);
enddef;

def gridrays (expr tstep) =
  beginpolargrid;
    tmin := tstep * ceiling (tmin/tstep);
    tmax := tstep * floor (tmax/tstep);
    patchrays (gridpic) (tmin, tmax, tstep, rmin, rmax);
  endpolargrid (drawcolor, .5penwd);
enddef;

def polargrid (expr rstep, tstep) =
  gridarcs (rstep);  gridrays (tstep);
enddef;

def polargridpoints (expr dsize, rstep, tstep) =
  beginpolargrid;
    setpicture (gdot) setdot (griddotpath, dsize);
    if rmin = 0:
      picdot (gridpic, gdot, zconv (origin));
      rmin := rstep;
    fi
    for n = ceiling (rmin/rstep) upto floor (rmax/rstep):
    for m = ceiling (tmin/tstep) upto floor (tmax/tstep):
      picdot ( gridpic, gdot, zconv ( polar ((n*rstep, m*tstep)) ) );
    endfor
    endfor
  endpolargrid (pointcolor, .5dsize);
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{beginpolargrid}
% The macro \gbc{beginpolargrid} calls \gbc{getpolarbounds} to compute the
% bounds (on $r$ and $\theta$) of the smallest polar coordinate patch that
% covers the graph rectangle.
% \DescribeRoutine{getpolarbounds}
% That command leaves the values in \gbc{rmin}, \gbc{rmax}, \gbc{tmin} and
% \gbc{tmax}. Then \gbc{beginpolargrid} initializes \gbc{gridpic} whereon
% the grids are drawn.
%    \begin{macrocode}
def beginpolargrid =
begingroup;
  save rmax, rmin, tmax, tmin;
  getpolarbounds;
  newpicture (gridpic);
enddef;

def getpolarbounds =
    save p, r, t;
    pair p[];
    p0 := (xneg, yneg);  p1 := (xneg, ypos);
    p2 := (xpos, ypos);  p3 := (xpos, yneg);
%    \end{macrocode}
% This loop finds the radial coordinate of each corner of the graph and
% finds the maximum while doing so.
%    \begin{macrocode}
    r0 := abs(p0);  rmax := r0;
    for j = 1 upto 3:
      r[j] := abs(p[j]);
      if rmax < r[j]:  rmax := r[j];  fi
    endfor
%    \end{macrocode}
% When the origin is inside the graph rectangle we need the full range
% of $r$ and $\theta$. When the origin is one of the corners, the angles
% can just be read off. Otherwise, to find the range of $\theta$ we
% essentially rotate one corner to have angle zero, get the angles to all
% corners and rotate back. This guarantees that the wedge with
% \gbc{tmin}${} < \theta < {}$\gbc{tmax} includes the graph.
%    \begin{macrocode}
    rmin := 0;
    if between (xneg, xpos) (0) and between (yneg, ypos) (0):
      tmin := 0;  tmax := 360;
    elseif (p0 = origin):  tmin :=    0;  tmax :=  90;
    elseif (p1 = origin):  tmin :=  -90;  tmax :=   0;
    elseif (p2 = origin):  tmin := -180;  tmax := -90;
    elseif (p3 = origin):  tmin :=   90;  tmax := 180;
    else:
      tmax := tmin := t0 := angle p0;
      for j = 1 upto 3:
        t := t0 + anglefromto (p0, p[j]);
        if tmax < t:  tmax := t;  fi
        if tmin > t:  tmin := t;  fi
      endfor
%    \end{macrocode}
% The minimum value of $r$ can be one of 9 possibilities: if the four
% sides of the graph are extended infinitely far in both directions, the
% origin can be in any one of the 9 regions formed. We've already disposed
% of the inside of the graph. This code considers the remaining regions in
% the following order: (1)~above or below, (2)~left or right, and (3)~one
% of the four corner regions.
%    \begin{macrocode}
      if between (xneg, xpos) (0):
        rmin := emin (abs(yneg), abs(ypos));
      elseif between (yneg, ypos) (0):
        rmin := emin (abs(xneg), abs(xpos));
      else:
        rmin := min (r0, r1, r2, r3);
      fi
    fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{endpolargrid}
% The \gbc{clr} is \gbc{drawcolor} for line grids, \gbc{pointcolor} for
% dot grids. The \gbc{size} is half the width of the grid's lines or half
% the width of the grid's dots. The purpose is to make sure dots and lines
% on the graph's edge aren't cut off. For dots I should probably put this
% decision in the code that draws them on \gbc{gridpic}.
%    \begin{macrocode}

def endpolargrid (expr clr, size)=
    clipto (gridpic) rect ( zconv ((xneg, yneg)) - size*(1,1),
                            zconv ((xpos, ypos)) + size*(1,1) );
    coloraddon (clr, gridpic);
  endgroup
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{polarpatch}
% Finally, this just does \gbc{plrpatch}, but also draws the ending
% boundaries, in case they are not an integer number of steps from the
% start.
%    \begin{macrocode}
vardef polarpatch (expr rstart, rstop, rstep, tstart, tstop, tstep) =
  plrpatch (rstart, rstop, rstep, tstart, tstop, tstep);
  safedraw zconv ( arcplr (origin, tstart, tstop, rstop) );
  safedraw zconv ( ((rstart, 0)--(rstop, 0)) rotated tstop );
enddef;

%    \end{macrocode}
%
% \section{Path Construction}\label{pathconstruction}
%
% This section is devoted to commands that accept a list or array of
% points and produce a path, usually (but not necessarily) through those
% points. In addition there are a few commands that find some of the key
% points, lines and circles associated with a triangle. No \mfpic{}
% interface is yet available for the triangle commands.
%
% \subsection{Piecewise linear paths}\label{linear}
%
% \DescribeRoutine{rect}
% Most of the macros that only define paths are coordinate independent.
% The simplest is \gbc{rect}. It accepts two pair expressions and produces
% the upright rectangle with those points at opposite corners. It might be
% noted that if the corners really are lower left and upper right, then
% the path is anticlockwise, If they are on the other diagonal, the
% path is clockwise. The path is a cycle (closed). The starting/ending
% point (needed for arrows and the like) is the first point of the two.
%
% \DescribeRoutine{triangle}
% Produces a closed path joining three points with straight lines; first
% named point \gbc{A} is \mfc{point 0 of triangle (A, B, C)}, etc.
%    \begin{macrocode}
vardef rect (expr ll, ur) =
  ll--(xpart ur, ypart ll)--ur--(xpart ll, ypart ur)--cycle
enddef;
vardef triangle (expr A, B, C) =  A--B--C--cycle  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{regularpolygon}
% The first argument is the number of sides, the second is an array name
% to hold the list of vertices. The third argument contains two
% equations separated by a semicolon, preferably the location of two of
% the vertices, or the location of the center and one vertex. That plus
% the equations in the \mfc{for}-loop give \gbc{n+1} equations to
% determine the \gbc{n} vertices and the center. Note that the vertices
% are numbered anticlockwise.
%
% The equations must \emph{not} be equations that are satisfied by all
% $n$-gons regardless of size and position. For example:\\
% \indent \gbc{regulapolygon(4)(Ted)(Ted0 := (0,0);Ted1 + Ted3 = (0,0))}\\
% The second of these says the middle lies halfway between the extremes,
% and is already a consequence of the code. It goes without saying that
% equations that cannot be satisfied by any regular polygon are also out.
%    \begin{macrocode}
vardef regularpolygon (expr n) (suffix Bob) (text eqns) =
  pair Bob[];  Bob := emax (round (abs (n)), 2);
  eqns;
  for _uncle = 1 upto Bob - 1:
    (Bob1 - Bob0) rotated (360/Bob*_uncle) = Bob[_uncle+1] - Bob0;
  endfor
  mkpoly (true) (Bob)
enddef;

%    \end{macrocode}
%
% The following set of commands take a path as argument, but it is
% intended that it be a triangle. Even then, they work correctly only if
% it is a cycle.
%
% These produce the perpendicular from \,\gbc{point n of t}\, to the
% (extension of) the opposite side (i.e., the altitude).
% \DescribeRoutine{altitudept}
% The first one determines where the altitude meets the opposite side, and
% the
% \DescribeRoutine{altitude}
% second just connects the two points. Since \gbc{altitudept} is always
% \gbc{point 1 of altitude}, it is actually redundant. However, the
% command \gbc{medianpt} (defined below) is used outside of the
% construction of \gbc{median}, so it seemed possible the \gbc{altitudept}
% might be useful also.
%
% We need a cycle so that points $n+1$ and $n+2$ will wrap around to the
% start of the path when necessary.
%    \begin{macrocode}
vardef altitudept expr n of t =
  save A, B, C, zz;  pair A, B, C, zz;
  B := pnt[n + 1] (t);
  C := pnt[n + 2] (t);
  zz = whatever[B,C];
  zz = pnt[n](t) + whatever*((C-B) rotated 90);
  zz
enddef;

vardef altitude expr n of t =
  (pnt[n](t))--(altitudept n of t)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{medianpt}
% These next two produce the midpoint of the side opposite
% \,\gbc{point n of t}\, and the
% \DescribeRoutine{median}
% line connecting those two points.
%    \begin{macrocode}
vardef medianpt expr n of t =
  0.5[pnt[n + 1] (t), pnt[n + 2] (t)]
enddef;

vardef median expr n of t =
  (pnt[n](t))--(medianpt n of t)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{anglebisectorpt}
% The first produces the point on the side opposite \,\gbc{point n of t}\,
% where the angle bisector at that corner crosses it and the second
% produces
% \DescribeRoutine{anglebisector}
% the line that bisects that angle.
%    \begin{macrocode}
vardef anglebisectorpt expr n of t =
  save A, B, C;  pair A, B, C;
  A := pnt[n    ] (t);
  B := pnt[n + 1] (t);
  C := pnt[n + 2] (t);
  save zz;  pair zz;
  zz = whatever[B,C];
  zz = A + whatever*((B-A) rotated (.5*cornerangle (A,B,C)));
  zz
enddef;

vardef anglebisector expr n of t =
  (pnt[n](t))--(anglebisectorpt n of t)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{cornerangle}
% This calculates the angle at the corner of a triangle. Specifically,
% the angle (between $-180$ and $180$) required to rotate the vector
% \gbc{B-A} into \gbc{C-A}. For degenerate triangles the seemingly
% arbitrary values 60 and 90 are designed to match the assumptions used
% in the arc commands. But also to guarantee that the three
% \gbc{cornerangle}\,s add up to $\pm180$. \gbc{cornerangle (A,B,C)} gives
% the angle at \gbc{A}, positive if \gbc{A--B--C--cycle} is
% anticlockwise.
%    \begin{macrocode}
vardef anglefromto (expr u, v) =
  if (u = origin) or (v = origin):  0
  else:  angle (v rotated (-angle u))
  fi
enddef;

vardef cornerangle (expr A, B, C) =
  if (A = B) or (A = C) :
    if (B = C) :  60
    else:         90
    fi
  else:  anglefromto (B - A, C - A)
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mkpath}
% This accepts the name of an array of pairs and produces a path
% that connects them. The first and third parameters are booleans. If
% \gbc{smooth} is \mfc{true} a smooth path is produced, otherwise a
% polyline. If \gbc{cyclic} is \mfc{true} the path is closed. The work is
% actually done by \gbc{mksmooth} or \gbc{mkpoly}.
%    \begin{macrocode}
vardef mkpath (expr smooth, tens, cyclic) (suffix pts) =
  if smooth:  mksmooth (tens)
  else:      mkpoly
  fi  (cyclic, pts)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mkpoly}
% This produces the path of line segments connecting \gbc{pts1},
% \gbc{pts2}, etc., closing it up if the boolean \gbc{cyclic} is true.
% It is also used with an array of paths instead of points.
%    \begin{macrocode}
vardef mkpoly (expr cyclic) (suffix pts) =
  for _i = 1 upto pts-1:  pts[_i]-- endfor
  pts[pts] if cyclic:  -- cycle  else:  {0,0}  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{polyline}
% This is the \mfpic{} interface. Instead of an array name, it accepts a
% list of pair expressions, forms an array from them and calls
% \gbc{mkpoly}.
% \DescribeRoutine{NoPoints}
% \mfc{NoPoints} prints a warning and sets the array to a single point,
% the origin.
%    \begin{macrocode}
vardef polyline (expr cyclic) (text t) =
  setpairs (_pl) (t);
  if _pl=0:  NoPoints ("polyline", _pl);  fi
  mkpoly (cyclic, _pl)
enddef;

def NoPoints (expr s) (suffix pts) =
  GBwarn  s & " attempted with empty list.";  pts[incr pts] := origin;
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{turtle}
% \emph{Turtle graphics} was a teaching tool to get youngsters used to the
% concept of programming while also teaching geometry. The students fed an
% Apple II computer a sequence of angles and distances, and a small
% triangle on the screen (the `turtle') would turn the indicated angle
% and travel the indicated distance, tracing a polyline on the screen.
%
% The argument of \gbc{turtle} is a list of pairs. The first is the
% starting point, the rest are vector displacements (moves). The
% distance and incremental angles of the original turtle graphics would
% require keeping track of the current angle and using the \gbc{polar}
% command.
%    \begin{macrocode}
vardef turtle (text t) =
  setnumeric (_tu) 0;
  setpair (_tmp) origin;
  pair _tu[];
  for _a = t:
    _tmp := _tmp + _a;
    _tu[incr _tu] := _tmp;
  endfor
  if _tu = 0:  NoPoints("turtle", _tu);  fi
  mkpoly (false, _tu)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{brownianpath}
% I needed the following to illustrate Brownian motion. It takes a given
% starting point, a given number of steps and a scaling factor. It
% generates a sequence of random points, each one being chosen randomly
% using a Gaussian distribution centered at the previous point. Strictly
% speaking this is a Gaussian random walk, not Brownian motion. A true
% Brownian motion would be a limit of these, with \gbc{num} tending to
% $\infty$ and \gbc{sc} tending to 0.
%    \begin{macrocode}
vardef brownianpath (expr start, num, sc) =
  setnumeric (_brp) 1;
  setpair (_tmp) start;
  pair _brp[];  _brp1 := _tmp;
  for _idx := 1 upto num:
    _tmp := _tmp + sc/(sqrt 2)*(normaldeviate,normaldeviate);
    _brp[incr _brp] := _tmp;
  endfor
  mkpoly (false, _brp)
enddef;

%    \end{macrocode}
%
% \subsection{Smooth paths}\label{smooth}
%
% We added an optional parameter for the tension of smooth curves to
% \mfpic. It used to be implemented this way: functions that implement a
% tension parameter set \gbc{cur_tension} and called \gbc{mksmooth}, which
% used that tension in its formation of a path. Since \gbc{mksmooth} was
% only ever used in this way, I decided to change its syntax to include a
% tension parameter. Only the functions \gbc{tcurve} and \gbc{mkpath}
% actually call \gbc{mksmooth} directly, most other path building commands
% with tension parameters call \gbc{mkpath} or \gbc{mkfcn} (which calls
% \gbc{mkpath}).
%
% \DescribeRoutine{mksmooth}
% This takes a tension value, a boolean, and the name of an array of
% points, draws the curve connecting them and closes it up if the boolean
% is true. It draws the curve forcing it to have the same direction at a
% point as the line segment connecting the preceding and following points.
% This is normally best if the curve direction changes relatively modestly
% from point to point. For example, if the polyline would be convex, then
% this smooth version would be pretty close to being convex. If the convex
% polygon has several consecutive sides that are in the same direction,
% all but the first and last of these segments in the smooth version would
% be straight.
%    \begin{macrocode}
vardef mksmooth (expr tens, cyclic) (suffix pts) =
  if pts = 1:  onepointpath (cyclic, pts1)
  else:
    settension (_tn) tens;  fixtension (_tn);
    pts1  if cyclic:  {pts[2]-pts[pts]}  fi
    for _i = 2 upto pts-1:
      ..tension _tn..pts[_i]{pts[_i+1]-pts[_i-1]}
    endfor
    ..tension _tn..pts[pts]
    if cyclic:  {pts[1]-pts[pts-1]}..tension _tn..cycle  fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mktenser}
% This is just like \gbc{mksmooth}, except the tension value is preceded
% by \mfc{atleast}. At this writing only \gbc{mkconvex} uses it (as a
% fallback when there are three or fewer points to connect).
%    \begin{macrocode}
vardef mktenser (expr tens, cyclic) (suffix pts) =
  if pts = 1:  onepointpath (cyclic, pts1)
  else:
    settension (_tn) tens;  fixtension (_tn);
    pts1  if cyclic:  {pts[2]-pts[pts]}  fi
    for _i = 2 upto pts-1:
      ..tension atleast _tn..pts[_i]{pts[_i+1]-pts[_i-1]}
    endfor
    ..tension atleast _tn..pts[pts]
    if cyclic:  {pts[1]-pts[pts-1]}..tension atleast _tn..cycle  fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mkconvex}
% This could have been very much like \gbc{mksmooth}, using
% \mfc{tension atleast} instead of \mfc{tension} (i.e., exactly
% \gbc{mktenser} above). Unfortunately This destroys smoothness at the
% beginning and end of any sequence of three or more points that lie on a
% straight line. Some geometric situations absolutely prevent smoothness,
% but this certainly isn't one of them. Two consecutive points
% identical isn't either, but it does. We let the user or calling command
% arrange for it not to happen (for example, using \gbc{setuniquepairs}
% instead of \gbc{setpairs}.
%
% What we do is weight the direction to be used at each point by how flat
% the polygon is on the two sides of each point, the flatter side
% getting the most weight. If a point is collinear with the next two,
% the flatness is infinite and the curve is forced in that direction. We
% measure the flatness using the square root of the area of the triangle
% made by the given point and the the next two points. Those three points
% lie on a line just when the area is $0$. Using this measure of flatness
% can be disputed (two triangles can be equally flat in terms of angles
% but different in terms of areas), but it has the advantage that if the
% points are subjected to an affine transformation, the weighting is
% unchanged and the constructed direction vectors transform the same way.
% Of course, this is also true of other measures. I've tested only
% the square root of area and it works well for simple cases.
%
% The only way to ensure that the whole path transforms the same as the
% points is to explicitly calculate the controls (in a manner that
% transforms the same way as the points). I will do this if I can figure
% out how it \emph{should} be done. It appears to be a convex programing
% problem.
%
% The end segments of a noncycle are problematic. One could argue that
% this command doesn't apply to noncycles and just truncate the cyclic
% path. I decided to treat them specially and hope the result is useful.
%    \begin{macrocode}
vardef mkconvex (expr tens, cyclic) (suffix pts) =
  save _B, _d, _tmp;  pair _d[];
  settension (_tn) tens;  fixtension (_tn);
  if pts < 4:  mktenser (_tn, cyclic) (pts)
  else:
    for _j = 2 upto pts - 1:
      _B[_j] := sqrt(abs((pts[_j]-pts[_j-1])xprod(pts[_j+1]-pts[_j])));
    endfor
    if cyclic:
      _B1     := sqrt(abs((pts1    -  pts[pts])xprod(pts2 - pts1)));
      _B[pts] := sqrt(abs((pts[pts]-pts[pts-1])xprod(pts1 - pts[pts])));
    else:
      _B1     := _B2;
      _B[pts] := _B[pts-1];
    fi
    for _j = 2 upto pts - 1:
      _tmp := _B[_j-1] + _B[_j+1];
      _d[_j] :=
        if _tmp = 0:  origin % signal to use curl1
        else:
          ( _B[_j+1]*(pts[_j] - pts[_j-1]) +
            _B[_j-1]*(pts[_j+1] - pts[_j]) )/_tmp
        fi;
    endfor
    if cyclic:
      _tmp := _B[pts] + _B2;
      _d1 :=
        if _tmp = 0:  origin
        else:
          (_B2*(pts1 - pts[pts]) + _B[pts]*(pts2 - pts1))/_tmp
        fi;
      _tmp := _B[pts-1] + _B1;
      _d[pts] :=
        if _tmp = 0:  origin
        else:
          ( _B1*(pts[pts] - pts[pts-1]) +
            _B[pts-1]*(pts1 - pts[pts]) )/_tmp
        fi;
    else:
      _d1 := origin;  _d[pts] := origin;
    fi
    pts1
    for _j = 1 upto pts-1:
      {if _d[_j] = origin:  curl1  else:  _d[_j] fi}
      ..tension atleast _tn..pts[_j+1]
    endfor
      {if _d[pts] = origin:  curl1  else:  _d[pts] fi}
    if cyclic:  ..tension atleast _tn..cycle  fi
  fi
enddef;

%    \end{macrocode}
%
% The old \cs{curve} command in \mfpic{} permitted no tension parameter
% and wrote a \grafbase{} \gbc{curve} command.
% \DescribeRoutine{curve}
% For backward compatibility we keep that name, but simply call the
% \gbc{tcurve} command with the default value for tension.
% \DescribeRoutine{tcurve}
% \gbc{tcurve} converts a list of pairs to an array, then calls
% \gbc{mksmooth} on the array.
%
% The next pair call \gbc{mkconvex}, which tries to produce a convex curve
% when the points form a convex polygon.
% \DescribeRoutine{ccurve}
% The first, \gbc{ccurve}, merely calls the second with a default texnsion,
% while
% \DescribeRoutine{tccurve}
% creates an array from the list of pairs and calls \gbc{mkconvex} on it.
%    \begin{macrocode}
numeric default_tension;  default_tension := 1;
def curve =  tcurve (default_tension)  enddef;
vardef tcurve  (expr tens, cyclic) (text t) =
  setpairs (_tc) (t);  mksmooth (tens, cyclic, _tc)
enddef;

def ccurve =  tccurve (default_tension)  enddef;
vardef tccurve (expr tens, cyclic) (text t) =
  setuniquepairs (_tcc) (t);  mkconvex (tens, cyclic, _tcc)
enddef;

%    \end{macrocode}
%
% It seemed odd that we had no way for an \MF-savvy user to easily get
% the standard \mfc{p..q..r} kind of path. For such a simple one,
% \cs{mfobj} with the explicit path expression would work, but when one
% has to add a tension to it, it is nice to have an abbreviation. That's
% what these are for.
%
% \DescribeRoutine{mkbezier}
% The command \gbc{mkbezier} takes an array argument and produces either
% an open or cyclic path with a given tension.
% \DescribeRoutine{bezier}
% The macro \gbc{bezier} does nothing more than call \gbc{tbezier} with
% the default tension,
% \DescribeRoutine{tbezier}
% which takes a list of points and creates an array for \gbc{mkbezier} to
% act on.
%    \begin{macrocode}
vardef mkbezier (expr tens, cyclic) (suffix pts) =
  settension (_tn) tens;  fixtension (_tn);
  pts1
    for _i = 2 upto pts:  ..tension _tn..pts[_i] endfor
    if cyclic:  ..tension _tn..cycle  else:  {0,0}  fi
enddef;

def bezier =  tbezier (default_tension)  enddef;
vardef tbezier (expr tens, cyclic) (text t) =
  setpairs (_tbs) (t);
  if _tbs=0:  NoPoints ("bezier", _tbs);  fi
  mkbezier (tens, cyclic) (_tbs)
enddef;

%    \end{macrocode}
%
% It also seemed we ought to allow \mfpic{} users to easily reproduce the
% effect of a sequence of \LaTeX's  \cs{qbezier} commands. That's what
% these next are for.
%
% These commands and the various splines below don't use tension as they
% have their control points explicitly given, not computed from the
% tension value by \MF. The \gbc{qbezier} command does not produce a
% smooth path unless the controls are explicitely chosen for that. The
% spline commands will almost always produce a smooth path.
%
% \DescribeRoutine{mkqbezier}
% \gbc{mkqbezier} requires an even number of points for a cyclic path,
% an odd number for a noncyclic path. It does not check for this, but
% the calling macro \gbc{qbezier} does. If the parity is incorrect, it
% repeats the last point in the list. This has the effect of making
% the last link a straight line.
%
% \DescribeRoutine{qbezier}
% The \gbc{qbezier} command takes a list of points and creates an array
% from then before calling \gbc{mkqbezier}.
%
% \DescribeRoutine{mkcbezier}
% This is like \gbc{mkqbezier}, but needs a multiple of 3 for a closed
% cubic bezier, one more (the endpoint) for an open cubic bezier.
%
% \DescribeRoutine{cbezier}
% Like \gbc{qbezier}, but calls \gbc{mkcbezier}.
%    \begin{macrocode}
vardef mkqbezier (expr cyclic) (suffix pts) =
  pts1
  if pts=1: {0,0}
  else:
    for _i = 2 step 2 until pts - 1:
      ..controls 1/3[pts[_i], pts[_i-1] ] and 1/3[pts[_i], pts[_i+1] ]..
      pts[_i+1]
    endfor
    if cyclic:
      ..controls 1/3[ pts[pts], pts[pts - 1] ]
        and 1/3[ pts[pts], pts1 ]..cycle
    fi
  fi
enddef;

vardef qbezier (expr cyclic) (text t) =
  setpairs (_qbz) (t);
  if _qbz=0:  NoPoints ("qbezier", _qbz);
  else:
    if (cyclic and odd _qbz) or (not cyclic and even _qbz):
      _qbz[incr _qbz] := _qbz[_qbz-1];
    fi
    mkqbezier (cyclic) (_qbz)
  fi
enddef;

vardef mkcbezier (expr cyclic) (suffix pts) =
  pts1
  if pts=1:  {0,0}
  else:
    for _i = 1 step 3 until pts - 3:
      ..controls pts[_i+1] and pts[_i+2] .. pts[_i+3]
    endfor
    if cyclic:
      ..controls  pts[pts - 1] and pts[pts]..cycle
    fi
  fi
enddef;

vardef cbezier (expr cyclic) (text t) =
  setpairs (_cbz) (t);
  if _cbz=0:  NoPoints ("qbezier", _cbz);
  else:
    % Need 0 mod 3 for cyclic, otherwise 1 mod 3
    setnumeric (_mdt) _cbz mod 3;
    if cyclic:
      if _mdt <> 0:  _cbz[incr _cbz] := _cbz[_cbz-1];  fi
      if _mdt = 1 :  _cbz[incr _cbz] := _cbz1;  fi
    else: % need 1 more, duplicate next to last
      if _mdt = 0:
        _cbz := _cbz + 1;
        _cbz[_cbz]   := _cbz[_cbz-1];
        _cbz[_cbz-1] := _cbz[_cbz-2];
      fi
      if _mdt = 2: % need 2 more, duplicate last 2.
        _cbz := _cbz + 2; % add 2 slots
        _cbz[_cbz]   := _cbz[_cbz-2]; % fill them
        _cbz[_cbz-1] := _cbz[_cbz-2]; % with last node
        _cbz[_cbz-2] := _cbz[_cbz-3]; % orig last slot = orig previous.
      fi
    fi
    mkcbezier (cyclic) (_cbz)
  fi
enddef;

%    \end{macrocode}
%
% When calling \gbc{curve} or \gbc{tcurve} there can be a problem
% with the resulting path: even with high tension one is not guaranteed
% that a sequence of points with increasing $x$-coordinate will produce a
% path with increasing $x$-coordinate. The \gbc{fcnspline} command will do
% what we want, but we have no control over the path, apart from the
% equations at the ends.
%
% The requirement to guarantee that a path have increasing $x$-coordinates
% is that the control points of the segment connecting
% $(x\sb{j}, y\sb{j})$ to the next $(x\sb{j+1}, y\sb{j+1})$ have their
% $x$-part in the interval $x\sb{j} < x < x\sb{j+1}$.
%
% Therefore, if we wish to plot a curve connecting points with increasing
% $x$-coordinates and believe that the resulting path should be the graph
% of a function, we pretty much have to select the control points
% ourselves. By default we choose the two controls so the \mfc{xpart}s
% divide the $x$-interval into three equal parts. This makes the B\'ezier
% $f(t)$ linear in the $x$-part and so has the added `advantage' that in
% each segment, $y$ is a cubic function of $x$. It is not a spline, as
% the computation of the controls uses only the two nearest points, plus
% we allow them to be modified by an additional parameter.
%
% Another concern is what direction to place the controls. In
% \gbc{mksmooth} we ask the direction at a given point to be the average
% of the straight line directions to adjacent points. We do the same
% here, though it is not clear if this is best.
%
% Finally, we permit a tension of sorts by dividing the distance to the
% controls by a parameter normally equal to \gbc{default_tension}.
%
% \DescribeRoutine{fcncontrol}
% This computes the control point for the points on the path, following
% the above description. If by chance some $x$ interval is zero, we
% make the controls equal to the nodes, which gives a straight vertical
% (the $y$ values differ because we use \gbc{setuniquepairs}. This also
% abandons smoothness there.
%
% The method selecting the controls is new with \mfpic{} version 0.8.
% Following discussions with Stephan Hennig in \texttt{comp.text.tex} I
% came to the conclusion that the method used ought to satisfy the
% following: if the data are xscaled or yscaled, the control vectors ought
% to scale the same way. The current version does that, the previous one
% did not.
%
% \DescribeRoutine{mkfcnpath}
% This produces the path, calling \gbc{fcncontrol} to produce the controls.
%
% \DescribeRoutine{fcncurve}
% This is the \mfpic{} interface; \gbc{fcncurve} calls \gbc{functioncurve}
% with the default tension, and
% \DescribeRoutine{functioncurve}
% then takes a list of points, converts it to an array, and calls
% \gbc{mkfcnpath} to build the path.
%    \begin{macrocode}
vardef fcncontrol (expr ftens, X, Y, Z) =
  Y if (xpart(Z-Y) <> 0) and (xpart(Y-X) <> 0):
      + xpart(Z-Y)/3/xpart(Z-X)*(Z - X)/ftens  fi
enddef;

vardef mkfcnpath (expr ftens) (suffix q) =
  settension (_tn) ftens;
  if _tn <= 0:  _tn := 1;  fi
  for _i = 1 upto q - 1:
    q[_i]..controls fcncontrol (_tn) (q[_i-1], q[_i],  q[_i+1])
           and fcncontrol (_tn) (q[_i+2], q[_i+1], q[_i])..
  endfor
  q[q]{0,0}
enddef;

def fcncurve =  functioncurve (default_tension)  enddef;
def tfcncurve =  functioncurve  enddef;
vardef functioncurve (expr ftens) (text t) =
  settension (_ftens) ftens;  if _ftens < 1/3:  _ftens := 1/3;  fi
  setuniquepairs (_fc) (t);
  if _fc > 1:  _fc0 :=  _fc1;  _fc[_fc+1] := _fc[_fc];  fi
  mkfcnpath (_ftens) (_fc)
enddef;

%    \end{macrocode}
%
% \subsection{Splines with explicit controls}\label{splines}
%
% For these quadratic B-splines, a list of pairs representing the control
% points must be given. The nodes of the path  and the cubic Bezi\'er
% controls required to produce a quadratic B-spline are computed. The
% nodes are just half way between the one control point and the next.
%
% \DescribeRoutine{openqbs}
% For simplicity, the list is converted to an array \gbc{_oq} first. In
% the closed version
% \DescribeRoutine{closedqbs}
% additional array elements are created at the end, repeating two of the
% beginning elements. Finally,
% \DescribeRoutine{mkqbs}
% \gbc{mkqbs} is called. This draws an open spline based on the points in
% an array \gbc{b}. The additional array elements defined by
% \gbc{closedqbs} cause the resulting path to end where it began and a
% simple \mfc{\&cycle} closes it.
%
% \DescribeRoutine{qspline}
% The \mfpic{} commands \cs{qspline} and \cs{closedqspline} now call
% \gbc{qspline} with appropriate boolean, for consistency with other
% commands that have the same argument structure. The commands
% \gbc{openqbs}, and \gbc{closedqbs} are no longer needed, but are kept
% for backward compatability. The most efficient setup would be to give
% \gbc{mkqbs} a boolean argument, but that could break old files.
%    \begin{macrocode}
def openqbs  =  qspline (false)  enddef;
def closedqbs =  qspline (true)  enddef;

vardef mkqbs (suffix b) =
  0.5[ b1, b2]
  if b<3:  {0,0}
  else:
    for _i = 2 upto b-1:
      ..controls 1/6[ b[_i], b[_i-1] ] and 1/6[ b[_i], b[_i+1] ]..
      0.5[ b[_i], b[_i+1] ]
    endfor
  fi
enddef;

vardef qspline (expr cyclic) (text t) =
  setpairs (_qs) (t);
  if _qs=0:  NoPoints ("qspline", _qs);  fi
  if _qs=1:  _qs[incr _qs] := _qs1;  fi
  if cyclic:
    _qs[incr _qs] := _qs1;  _qs[incr _qs] := _qs2;
  fi
  mkqbs (_qs) if cyclic:  & cycle  fi
enddef;

%    \end{macrocode}
%
% These cubic B-splines also require a list of `control' points. Each of
% the points $Q$, combined with the next one $Q'$, determine two more
% points that divide the segment from $Q$ to $Q'$ into thirds. These new
% points become the two control points of a \MF{} B\'ezier segment. The
% nodes of these segments are half way between the second control of one
% segment and the first control of the next.
%
% \DescribeRoutine{mkcbs}
% The main code is in \gbc{mkcbs}, which results in an open curve. For
% backward compatibility, the alias
% \DescribeRoutine{mkopencbs}
% \gbc{mkopencbs} is supplied.
%
% \DescribeRoutine{mkclosedcbs}
% Earlier versions of \gbc{mkclosedcbs} would partly redefine its suffix
% parameter (for example, using \gbc{b[incr b]:=b1}. I decided this
% shouldn't change the array variable (imagine using two such functions on
% the same array). Now the code has been rearranged so the cubic case is
% handled just like the quadratic. We retain \gbc{mkclosedcbs} only for
% backward compatibility.
%
% \DescribeRoutine{opencbs}
% These are the versions taking a list of points instead of an array name.
% They create a temporary array and call \gbc{mkcbs}, with
% \DescribeRoutine{closedcbs}
% \gbc{closedcbs} extending the array, just like the quadratic versions.
%
% \DescribeRoutine{cspline}
% The \mfpic{} commands \cs{cspline} and \cs{closedcspline} now call
% \gbc{cspline} with appropriate boolean, for consistency with other
% commands that have the same argument structure. The commands
% \gbc{mkopencbs}, \gbc{opencbs}, and \gbc{closedcbs} are no longer
% needed, but are kept for backward compatability. The most efficient
% setup would be to give \gbc{mkcbs} a boolean argument, but that could
% break old files.
%    \begin{macrocode}
vardef mkcbs (suffix b) =
  (b[1]+4b[2]+b[3])/6
  if b < 4:  {0,0}
  else:
    for _i = 3 upto b-1:
      ..controls 1/3[ b[_i-1], b[_i] ] and 1/3[ b[_i], b[_i-1] ]
      .. (b[_i-1] + 4b[_i] + b[_i+1])/6
    endfor
  fi
enddef;

def mkopencbs =  mkcbs  enddef;
vardef mkclosedcbs (suffix b) =
  mkcbs (b) & opencbs (b[b-2],b[b-1],b[b], b1, b2, b3) & cycle
enddef;

def opencbs  =  cspline (false)  enddef;
def closedcbs =  cspline (true)  enddef;

vardef cspline (expr cyclic) (text t) =
  setpairs (_cs) (t);
  if _cs=0:  NoPoints ("cspline", _cs);  fi
  for _idx = _cs upto 2:  _cs[incr _cs] := _cs[_idx];  endfor
  if cyclic:
    for _idx = 1 upto 3:  _cs[incr _cs] := _cs[_idx];  endfor
  fi
  mkcbs (_cs) if cyclic:  & cycle  fi
enddef;

%    \end{macrocode}
%
% \subsection{Splines with computed controls}\label{computedsplines}
%
% A cubic spline through a set of points is a curve obtained by joining
% each point to the next with a cubic parametrized curve, where adjoining
% cubics must have matching first and second derivative at their common
% point. In the previous section's \gbc{mkcbs}, the control points must be
% supplied, it being up to the user to arrange (if necessary) that the
% spline produced passes through given points. If, instead, these points
% are given, it is possible to compute the necessary controls.
% Unfortunately, the controls are not uniquely determined unless the curve
% is required to be closed. For open curves, there is need for two
% additional conditions at the end points. A `relaxed spline' is produced
% if we require that the second derivative is $0$ at those points.
%
% For a closed curve, the equality of the first and second derivatives at
% the common beginning/ending point gives the needed additional equations.
%
% Note that this equates \emph{time} derivatives, so this works best when
% points are relatively evenly spaced and so the speed is relatively
% uniform. If points are differently spaced then the relatively slower
% speed between closely spaced points allows sharper turns without large
% second derivatives. Curves produced tend to have a more natural look,
% and relaxed splines are most suitable for smoothing data that is
% obtained by taking observations at evenly space times. Still, the
% technique is somewhat unstable when points are closely spaced, for
% example when a small change in the position of one point can produce a
% large change in its direction when viewed from another point.
%
% \DescribeRoutine{init_spline_eqns}
% In this command we generate the equations common to all cubic
% splines: the equality of derivatives at all interior points.
% This command accepts a suffix \gbc{pts}, which is the array of points
% to be connected. It initializes the variables \gbc{_spl_pre[\,]} and
% \gbc{_spl_post[\,]} to unknown arrays of pairs. These will hold the
% control points.
%
% \DescribeRoutine{closed_spline_eqns}
% The next two macros contain the additional equations: for a closed
% spline these are the same as the interior equation, but at the first and
% last point in the array.
% \DescribeRoutine{relaxed_spline_eqns}
% For relaxed splines they force the second derivative to be 0 at the
% first and last point.
%
% The macro \gbc{mksplinepath} simply assembles the previously computed
% points and controls into a path.
%
% \DescribeRoutine{mkspline}
% The macro \gbc{mkspline} issues the common equations and then either the
% closed equations (\gbc{closed = true}) or the relaxed equations
% (\gbc{closed = false}), before calling \gbc{mksplinepath}.
%
% The knowledgeable user can call \gbc{init_spline_eqns}, append any
% choice of equations for the end segments, and then call
% \gbc{mksplinepath~(false)} to produce any sort of open spline.
%
% \DescribeRoutine{dospline}
% This version accepts a list of pairs and produces a spline through
% them. It simply stores the list in an array and calls the appropriate
% version that operates on an array. This is the command passed by \mfpic{}.
%    \begin{macrocode}
def init_spline_eqns (suffix pts) =
  save _spl_pre, _spl_post;
  pair _spl_pre[], _spl_post[];
  for j= 2 upto pts - 1:
   _spl_post[j] + _spl_pre[j] = 2pts[j];
   _spl_pre[j+1]+2_spl_pre[j] = 2_spl_post[j]+_spl_post[j-1];
  endfor
enddef;

def closed_spline_eqns (suffix pts) =
  _spl_post1 + _spl_pre1 = 2pts1;
  _spl_post[pts] + _spl_pre[pts] = 2pts[pts];
  _spl_pre2 + 2_spl_pre1 = 2_spl_post1 + _spl_post[pts];
  _spl_pre1+2_spl_pre[pts] = 2_spl_post[pts]+_spl_post[pts-1];
enddef;

def relaxed_spline_eqns (suffix pts) =
  _spl_pre2 + pts1 = 2_spl_post1;
  pts[pts] + _spl_post[pts-1] = 2_spl_pre[pts];
enddef;

vardef mksplinepath (expr closed) (suffix pts) =
  pts1..controls _spl_post1 and
  for j = 2 upto pts if not closed:  -1 fi:
    _spl_pre[j]..pts[j]..controls _spl_post[j] and
  endfor
    if closed:  _spl_pre1..cycle  else:  _spl_pre[pts]..pts[pts]  fi
enddef;

def mkspline (expr closed) (suffix pts) =
  init_spline_eqns (pts);
  if closed:  closed_spline_eqns (pts);
  else:     relaxed_spline_eqns (pts);
  fi
  mksplinepath (closed) (pts)
enddef;

vardef dospline (expr closed) (text the_list) =
  setpairs (_sp) (the_list);
  if _sp=0:  NoPoints ("dospline", _sp);  fi
  if _sp=1:  _sp[incr _sp] := _sp1;  fi
  mkspline (closed) (_sp)
enddef;

%    \end{macrocode}
%
% The above computations produce a $2$-dimensional spline. A $1$-dimensional
% cubic spline would be a function $f(t)$ with numeric values rather
% than pair values. Such are often used to interpolate functions. That is,
% given pairs $(x\sb j,y\sb{j})$, and assuming they lie on the graph of
% some function (generally unknown), fill in the graph with $y = f(x)$
% where $f$ is a cubic function of $x$ in each interval $x\sb j < x < x\sb
% {j+1}$, making sure that the resulting graph is as smooth as possible at
% the points $x\sb j$.
%
% The requirements on our $2$-dimensional path are the following:
% \begin{enumerate}
%   \item The $j$th link should connect $(x\sb{j},y\sb{j})$ to $(x\sb{j+1},
%       y\sb{j+1})$.
%   \item The $x$-part of that link should increase linearly from $x\sb{j}$ to
%     $x\sb{j+1}$ as $t$ goes from $0$ to $1$.
%   \item The $y$-part should be a cubic $y = f(x)$.
%   \item The $x$-derivatives $df/dx$ and $d^2f/dx^2$ should match at the
% connecting points.
% \end{enumerate}
%
% Two necessary equations for converting between $x$ and $t$ coordinates
% are:
% \begin{equation}\label{first}
%         x = x\sb{j} + t \Delta x\sb{j}
% \end{equation}
% (where $\Delta x\sb{j} = x\sb{j+1} - x\sb{j}$) and
% \begin{equation}\label{second}
%       \frac{df}{dt} = \frac{dx}{dt}\frac{df}{dx} =
%           \Delta x\sb{j} \frac{df}{dx}.
% \end{equation}
% Thus we want to choose controls so that (\ref{first}) is maintained and
% so that $x$-derivatives match. It turns out that this requires controls
% at
% \begin{equation}
% \begin{array}{c}
%   (x\sb{j}, y\sb{j}) - (\Delta x\sb{j-1}, s\sb{j} \Delta x\sb{j-1})/3\\
%   (x\sb{j}, y\sb{j}) + (\Delta x\sb{j}  , s\sb{j} \Delta x\sb{j}  )/3
% \end{array}
% \end{equation}
% where $s\sb{j}$ is the slope (derivative) at $x\sb{j}$. This provides
% matching first derivatives automatically (equation (\ref{second})) and
% also (\ref{first}). To get matching second derivatives we need the same
% conditions as in parametric splines. We use these equations simplified to
% the form:
% \begin{displaymath}
%  s\sb{j+1} \Delta x\sb{j} - 2s\sb{j} (\Delta x\sb{j} +
%     \Delta x\sb{j-1}) + s\sb{j-1}\Delta x\sb{j-1}
%       =  3y\sb{j+1} - 3y\sb{j-1}.
% \end{displaymath}
% There can be almost any equations at the end points. For a relaxed
% spline we equate the second derivatives to 0. To get a periodic
% function, we equate the slope and second derivative at beginning to
% those at the end. This makes it possible to put a shifted copy of the
% graph with starting point at the end of the original and have the same
% smoothness at that connection as at the other points.
%
% \DescribeRoutine{init_fcnspl_eqns}
% This declares the temporary arrays \gbc{_dx[\,]} (the set of $dx\sb j$)
% and \gbc{_sl[\,]} (the desired slopes) and issues the common equations.
% The parameter \gbc{pts} is the array of $(x,y)$ values.
%
% \DescribeRoutine{periodic_fcnspl_eqns}
% For the periodic case we use \gbc{periodic_fcnspl_eqns} to generate the
% additional equations and for the
% \DescribeRoutine{relaxed_fcnspl_eqns}
% relaxed case we use \gbc{relaxed_fcnspl_eqns}. As before, one can
% produce custom splines by issuing the common equations and then ones own
% equations.
%
% \DescribeRoutine{mkfcnsplpath}
% Then we assemble the path from the computed information by calling the
% command \gbc{mkfcnsplpath}.
%
% \DescribeRoutine{mkfcnspline}
% These commands emits the appropriate equations then assemble the path.
% The if the first parameter is true it uses the periodic equations,
% otherwise the relaxed equations.
%
% \DescribeRoutine{fcnspline}
% Finally, this command is the one written by \mfpic{}. It copies a list
% of pairs into an array and calls the appropriate command to process
% them.
%    \begin{macrocode}
def init_fcnspl_eqns (suffix pts) =
  save _dx, _sl;  numeric _dx[], _sl[];
  _dx1 := xpart (pts2 - pts1);
  for j = 2 upto pts - 1:
    _dx[j] := xpart (pts[j+1] - pts[j]);
    _sl[j + 1]*_dx[j] + _sl[j-1]*_dx[j-1] + 2_sl[j]*(_dx[j] + _dx[j-1])
            = 3*ypart(pts[j+1] - pts[j-1]);
  endfor
enddef;

def periodic_fcnspl_eqns (suffix pts) =
  _sl1 = _sl[pts];
  _sl2*_dx1 + 2_sl1*_dx1 + 2_sl[pts]*_dx[pts-1] + _sl[pts-1]*_dx[pts-1]
          = 3 * ypart (pts[2] - pts[pts-1]);
enddef;

def relaxed_fcnspl_eqns (suffix pts) =
  _sl2*_dx1 + 2_sl1*_dx1 = 3 * ypart(pts2 - pts1);
  _sl[pts-1]*_dx[pts-1] + 2_sl[pts]*_dx[pts-1]
        = 3 * ypart(pts[pts] - pts[pts-1]);
enddef;

vardef mkfcnsplpath (suffix pts) =
  pts1..controls (pts1 + (1, _sl1)/3*_dx1) and
    for j = 2 upto pts - 1:
      (pts[j] - (1, _sl[j])/3*_dx[j-1]) ..pts[j]..
        controls (pts[j] + (1,_sl[j])/3*_dx[j]) and
    endfor
    (pts[pts] - (1,_sl[pts])*_dx[pts-1]/3)..pts[pts]
enddef;

vardef mkfcnspline (expr periodic) (suffix pts) =
  init_fcnspl_eqns (pts);
  if periodic:  periodic_fcnspl_eqns (pts);
  else:         relaxed_fcnspl_eqns (pts);
  fi
  mkfcnsplpath (pts)
enddef;

vardef fcnspline (expr periodic) (text the_list) =
  setpairs (_fs) (the_list);
  if _fs<2:
    if _fs=0:  NoPoints ("fcnspline", _fs);  fi
    onepointpath (false, _fs1)
  else:
    mkfcnspline (periodic) (_fs)
  fi
enddef;

%    \end{macrocode}
%
% \subsection{Arcs, circles and ellipses}\label{arcs}
%
% We have multiple commands that generate circular arcs, differing in
% how the arc is specified. All are (in part) based on the following
% \gbc{mkarc}. However, perfectly reasonable arcs can have centers so far
% away that requiring the center among the parameters can cause numeric
% overflow.
%
% I'd like to use some scheme that avoids this. It is possible, given
% three reasonably spaced points on an arc with angle less than 90
% degrees between each, to draw the arc without finding the center.
% However, I am not sure how to reduce any given format to this
% information
%
% Another problem is that of accuracy. If the angle is small, accuracy is
% not usually a problem, but if an angle is close to 360, and the
% endpoints are known, then finding the center (or finding other points on
% the arc without knowing the center) is unstable.
%
% There is really no problem with \gbc{mkarc} itself: if you can express
% both \gbc{center} and \gbc{begpt} in \MF, then the other values on the
% arc should be no problem.
%
% Care has been taken that changing the sign of various parameters
% produces reasonable results. And there should be no more problem for
% arcs with sweep larger than 360 degrees than with less.
%
% \DescribeRoutine{mkarc}
% This takes the center, starting and ending point (pair expressions) and
% the angle, and returns the arc defined pretty much the way \file{plain.mf}
% defines \mfc{quartercircle}.
%
% It would be easier to do something like we frequently do with
% \mfc{fullcircle}: make an arc of unit radius, and then rotate, scale
% and shift it into place. However, I would like to accomplish at least
% the following: if an endpoint of the arc is among the parameters, or is
% straightforwardly implied by them, then the corresponding endpoint of
% the path created should test equal to that point. Shifting works OK, but
% scaling and rotating cause roundoff differences.
%
% Note that \gbc{mkarc} has parameters that may over-determine the arc.
% It is only called by arc-making commands that have calculated these
% parameters and, I hope, ensured they are compatible. \gbc{mkarc}'s job
% is mainly to ensure that the arc begins at \gbc{begpt} and ends at
% \gbc{endpt} (exactly). A \gbc{sweep} of $0$ is actually incompatible
% with any case where \gbc{begpt<>endpt} unless \gbc{center} is
% literally at infinity, but we allow it even though I am pretty sure
% the other arc commands all filter out that case.
%    \begin{macrocode}
vardef mkarc (expr center, begpt, endpt, sweep) =
  if (sweep = 0):  begpt--endpt
  else:
    setnumeric (n) ceiling (abs(sweep)/45);
    setpair (d) (begpt - center) rotated (signof (sweep) 90);
    begpt{d}
    for j = 1 upto n-1:
      ..(begpt rotatedabout (center, j/n*sweep)){d rotated (j/n*sweep)}
    endfor ..endpt{d rotated sweep}
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arc}
% The most basic: center of circle, starting point of arc, and angle
% subtended. Another name for \gbc{arc} is \gbc{arccps}, (\gbc{cps} is
% for ``center, point, sweep'').
%    \begin{macrocode}
vardef arc (expr center, begpt, sweep) =
  if (center = begpt) or (sweep = 0):  begpt--begpt
  else:
    mkarc (center, begpt, begpt rotatedabout (center, sweep), sweep)
  fi
enddef;
def arccps =  arc  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arcpps}
% In this form we are given two points and the angle of the arc between
% them. If the points are equal or the sweep makes the arc undefined, we
% return a line segment. If the sweep is less than 90 degrees we use the
% idea from the code of \mfc{quartercircle}, except, when the sweep is
% greater than 45 degrees we let \MF{} find the midpoint \gbc{m} of the
% arc. Otherwise, we get the center \gbc{c} of the circle and call
% \gbc{mkarc}.
%
% The code for finding \gbc{c} and \gbc{m} used to be separate commands,
% \gbc{arccenter} and \gbc{midarc}. However, this is the only place we
% used them and the several cases that they had to consider are reduced
% because the \mfc{if} in this command takes care of some of them.
%
% The code for finding \gbc{m} uses the fact that chord and the line
% from one of its endpoints to the midpoint subtend a circular arc of
% \gbc{sweep/2} and so the angle between them is half that, \gbc{sweep/4}.
% The code gets the intersection between the line in that direction and
% the perpendicular bisector of the chord.
%
% We find the center by intersecting two lines. One is the radius from
% one end of the chord. Then we branch on two cases: if the chord is close
% to a diameter, use its perpendicular bisector as the other line,
% otherwise use the radius from the opposite end of the chord. Here
% \gbc{cd} is a vector in the direction of the chord from \gbc{begpt} to
% \gbc{endpt}. The angle \gbc{ang} is the amount we have to rotate
% \gbc{cd} about \gbc{begpt} to make it point toward the center of the
% circle. This gives the radius mentioned above.
%    \begin{macrocode}
vardef arcpps (expr begpt, endpt, sweep) =
  if begpt = endpt:  begpt--endpt
  else:
    setpair (cd) unitvector (endpt-begpt);
    if abs(sweep) <= 45:
      begpt{cd rotated (-sweep/2)}..endpt{cd rotated (sweep/2)}
    elseif abs(sweep) <= 90:
      save m;  pair m;
      m = begpt + whatever*( cd rotated (-sweep/4));
      m = 0.5[begpt, endpt] + whatever*(cd rotated 90);
      begpt{cd rotated (-sweep/2)}..m{cd}..endpt{cd rotated (sweep/2)}
    else:
      setnumeric (ang) 90 - ((sweep/2) mod 180);
      if abs(ang) = 90:
        GBwarn "undefined arc. A line segment will be used instead.";
        begpt--endpt
      else:
        save c;  pair c;
        c = begpt + whatever*(cd rotated ang);
        c = if abs(ang) < 30:
              (0.5)[begpt, endpt] + whatever*(cd rotated 90)
            else:
              endpt + whatever*(-cd rotated -ang)
            fi;
        mkarc (c, begpt, endpt, sweep)
      fi
    fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arcpp}
% In the macro \gbc{arcpp}, two points and the radius of the circle are
% given. Alone, this would determine two circles and therefore 4 arcs. We
% reduce the possibilities to two by assuming the arc is anticlockwise
% from the first point to the second if \gbc{rad} is positive, clockwise
% if negative. Then \gbc{arcpp} produces the one that has absolute value
% no more than 180 degrees if \gbc{small} is true, otherwise the other
% one.
% \DescribeRoutine{arcppr}
% The macro \gbc{arcppr} is just \gbc{arcpp} with the boolean argument
% \gbc{small} last (for compatibility with previous \mfpic{} versions).
%
% The code computes the angle of the arc and calls \gbc{arcpps}. If the
% radius is not larger than half the distance between the points, we make
% the angle $\pm 180$, which produces a half circle.
%    \begin{macrocode}
vardef arcpp (expr small, begpt, endpt, rad) =
  save full, diam, chord, ang;
  full := signof (rad) 360;
  diam  := 2rad;
  chord := abs(endpt-begpt);
  if chord < abs(diam):
    ang := if not small:  full -  fi  2*asin (chord/diam);
  else:  ang := signof (rad) 180;
  fi
  arcpps (begpt, endpt, ang)
enddef;
def arcppr (expr begpt, endpt, rad, small) =
  arcpp (small, begpt, endpt, rad)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arcplr}
% This one takes the center and polar coordinates of the ends relative to
% the center. We just call \gbc{mkarc} with the obviously computed
% endpoints and sweep.
%    \begin{macrocode}
vardef arcplr (expr center, frtheta, totheta, rad) =
  if rad = 0:  center--center
  else:
    mkarc (center, center + rad*dir frtheta,
                   center + rad*dir totheta, totheta - frtheta)
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arcalt}
% This one is the same as above, but with the same argument order as
% \gbc{sector}.
%    \begin{macrocode}
vardef arcalt (expr center, radius, frtheta, totheta) =
  arcplr (center, frtheta, totheta, radius)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arcppp}
% This last one finds the arc connecting three points in the order given.
% It works by calling \gbc{arcpps} twice, using first the sweep from
% \gbc{first} to \gbc{second}, and then the sweep from \gbc{second} to
% \gbc{third}. Each of these is twice the opposite angle of the triangle
% formed from these points, and calculated by \gbc{cornerangle}.
%    \begin{macrocode}
vardef arcppp (expr first, second, third) =
  arcpps (first, second, 2*cornerangle (third, first, second)) &
  arcpps (second, third, 2*cornerangle (first, second, third))
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{ellipse}
% We get an ellipse by xscaling and yscaling a unit circle, rotating it
% and then shifting it into position. All parameters are coordinate
% independent expressions, with obvious meaning (\gbc{center} is a pair, the
% rest numeric). \gbc{circle} is similar, but we only scale and shift.
%
% If either radius is negative, the sense of the ellipse is reversed and
% the starting point changes. If both are negative, only the starting
% point changes.
% \DescribeRoutine{circle}
% \gbc{circle} acts like \gbc{ellipse} with both radii the same.
%    \begin{macrocode}
vardef ellipse (expr center, radx, rady, angle) =
  fullcircle xscaled (2*radx) yscaled (2*rady) rotated angle
    shifted center
enddef;

vardef circle (expr center, rad) =
  fullcircle scaled (2*rad) shifted center
enddef;

%    \end{macrocode}
%
% The next four implement different ways of specifying a circle.
% \DescribeRoutine{circlecp}
% The first, \gbc{circlecp}, produces the circle with a given center
% passing through a given point.
% \DescribeRoutine{circleppp}
% The second, \gbc{circleppp}, produces the circle passing through three
% given points.
% \DescribeRoutine{circlepps}
% The third, \gbc{circlepps}, produces the circle passing through two
% given points in such a way that the arc from the first to the second has
% a given angle.
% \DescribeRoutine{circleppr}
% The fourth, \gbc{circleppr}, produces the circle with the given radius
% passing through the two points in such a way that the angle from the
% first point to the second is between $0$ and $180$ degrees if the switch
% \gbc{small} is true. If \gbc{small} is false, then the clockwise arc
% from first to second is between $180$ and $360$. If \gbc{rad} is
% negative, the circles switch and their orientation is reversed.
% \DescribeRoutine{circlepp}
% The last, \gbc{circlepp}, is just \gbc{circleppr} with a different order
% of arguments (for previous \mfpic{} versions).
%
% These could be implemented by finding the center and radius and calling
% \gbc{circle}. However, we call the arc commands so that those points
% specified in the parameters that lie on the circle will be nodes of the
% path produced, in the given order.
%    \begin{macrocode}
vardef circlecp (expr center, point) =
  mkarc (center, point, point, 360) & cycle
enddef;

vardef circleppp (expr one, two, three) =
  arcpps (one, two, 2*cornerangle (three, one, two))
    & arcpps (two, three, 2*cornerangle (one, two, three))
    & arcpps (three, one, 2*cornerangle (two, three, one))
    & cycle
enddef;

vardef circlepps (expr one, two, sweep) =
  save ang, full;
  full := signof (sweep) 360;
  ang := sweep mod full;
  arcpps (one, two, ang) & arcpps (two, one, full - ang) & cycle
enddef;

vardef circlepp (expr small, one, two, rad) =
 arcpp (small, one, two, rad) & arcpp (not small, two, one, rad) & cycle
enddef;

def circleppr (expr one, two, rad, small) =
  circleppr (one, two, rad, small)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{pathcenter}
% This finds the center of a circle. For other paths, the point found
% may be meaningless (but it will also obtain the center of an arc or a
% rectangle). It takes three supposedly distinct points on the path and
% finds the intersection of the perpendicular bisectors of two chords.
%    \begin{macrocode}
vardef pathcenter expr p =
  save a, cntr, n;  pair cntr, a[];
  n := length p;
  a1 = pnt   0   (p);
  a3 = pnt [n/2] (p);
  if cycle p:
    a2 = pnt [n/4] (p);
    a4 = pnt [3n/4] (p);
  else:
    a2 := a3;
    a4 := pnt[n] (p);
  fi
  cntr = .5[a1, a3] + whatever*((a3 - a1) rotated 90);
  cntr = .5[a2, a4] + whatever*((a4 - a2) rotated 90);
  cntr
enddef;

%    \end{macrocode}
%
% The next four commands create certain circles associated to
% triangles. The triangle is specified as a path expression, so they
% produce results for any path, but make sense only for a cyclic
% triangular path.
%
% \DescribeRoutine{circumcircle}
% This is just the circle through the three corners.
%
% \DescribeRoutine{incircle}
% The command \gbc{incircle} produces the circle inside the triangle that
% is tangent to all three sides. It makes use of the fact that the two
% tangent points on the sides adjacent to corner \gbc{A} (for example) are
% equidistant from \gbc{A}. The three equations then express the fact that
% the sum of the two distances from the tangent point to the corners on
% the same side add up to the length of the side.
%
% \DescribeRoutine{excircle}
% In \gbc{excircle}, a corner is given (by number from $0$ to $2$) and the
% circle is produced that is \emph{outside} the triangle and is tangent to
% the side opposite the point and tangent to the extensions of the other
% two sides.
%
% \DescribeRoutine{ninepointcircle}
% The ``nine-point circle'' passes through the following nine points: the
% midpoint of each side, the point on each side (extended, if necessary)
% where the altitude from the opposite corner meets it, and the midpoint
% of the segments connecting each corner to the intersection of the
% altitudes.
%    \begin{macrocode}
vardef circumcircle expr t =
  circleppp (pnt0 (t), pnt1 (t), pnt2 (t))
enddef;

vardef incircle expr t =
  save A, B, C;  pair A, B, C;
  A := pnt0 (t);
  B := pnt1 (t);
  C := pnt2 (t);
  save a, b, c, D, E, F;
  D := abs (B-A) = a + b;
  E := abs (C-B) = b + c;
  F := abs (A-C) = a + c;
  circleppp ((a/D)[A,B], (b/E)[B,C], (c/F)[C,A])
enddef;

vardef excircle expr n of t =
  save A, B, C;  pair A, B, C;
  A := pnt[n] (t);
  B := pnt[n + 1] (t);
  C := pnt[n + 2] (t);
  save a, b, c, D, E, F;
  D := abs (B-A) = a - b;
  E := abs (C-B) = b + c;
  F := abs (C-A) = a - c;
  circleppp ((a/D)[A,B], (b/E)[B,C], (c/F)[A,C])
enddef;

vardef ninepointcircle expr t =
  circleppp (medianpt 0 of t, medianpt 1 of t, medianpt 2 of t)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{pshcircle}
% Here is a couple of circles maybe only I need. They are the
% pseudohyperbolic circles in the unit disk and upper half-plane.
% One supplies a point that must be inside the unit circle or above
% the $x$-axis, and a radius that must be less than $1$. Some degenerate
% cases will not generate an error. We code this with a boolean that
% determines whether the disk or the half-plane is to be assumed.
%    \begin{macrocode}
vardef pshcircle (expr disk, ctr, rad) =
  if disk:
    if rad >= 1 :
      if rad > 1:
        GBerrmsg ("Impossible radius of pseudohyperbolic circle.")
          "The radius of a pseudohyperbolic circle can be at most 1.";
      fi
      circle ((0,0),1)
    elseif abs(ctr) >= 1 :
      if abs(ctr) > 1:
        GBerrmsg ("Impossible center of pseudohyperbolic circle.")
          "The center of a pseudohyperbolic circle must be in "
          & "the unit disk.";
      fi
      onepointpath (true,ctr)
    else:
      % compute Euclidean center and radius (and a denominator used twice
      % in calculations).
      save _r, _dnm;
      _r := abs(ctr);
      _dnm := 1 - _r*_r*rad*rad;
      circle ( (1 - rad*rad)/_dnm*ctr, rad*(1 - _r*_r)/_dnm)
    fi
  else:
    if rad >= 1 :
      GBerrmsg ("Impossible pseudohyperbolic circle.")
        "The radius of a pseudohyperbolic circle must be less than 1.";
      onepointpath (true,ctr)
    elseif ypart ctr <= 0:
      if ypart ctr < 0:
        GBerrmsg ("Impossible pseudohyperbolic circle.")
          "The center of a pseudohyperbolic circle must be in "
          & "the upper half-plane.";
      fi
      onepointpath (true,ctr)
    else:
      % compute Euclidean center and radius (and a denominator used twice
      % in calculations).
      % Euclidean center at  xpart ctr + (1 + R^2)/(1 - R^2)*ypart ctr
      % Euclidean radius  4R/(1 - R^2)*ypart ctr
      save _y, _dnm;
      _y := ypart ctr;
      _dnm := 1 - rad*rad;
      circle ( (xpart ctr, (1 + rad*rad)/_dnm * _y), 2rad/_dnm*_y)
    fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{barycenter}
% This is the average of the three corners of the triangle, or of any
% path. If \gbc{t} is an open path with length $n$ and the nodes are
% $x\sb0$ through $x\sb n$, the barycenter is
% $$ \frac{1}{n+1}\sum\sb{j=0}\sp{n} x\sb j. $$
% If \gbc{t} is a cycle with $x\sb n = x\sb0$, then it is
% $$ \frac{1}{n}\sum\sb{j=0}\sp{n-1} x\sb j. $$
%
% For a triangle the barycenter is the intersection of the medians. I
% don't recall if this is the center of any important circle.
%
% The centers of the various circles associated with triangles can be
% found with \gbc{pathcenter}. Or by intersecting various lines: the
% \emph{incenter} (center of the inscribed circle)is the intersection of
% the angle bisectors; the \emph{circumcenter} is the intersection of the
% prependicular bisectors.
%    \begin{macrocode}
vardef barycenter expr t =
  save m;  m := length t if not cycle t:  + 1 fi;
  pnt0(t)/m  for k = 1 upto m - 1:  + pnt[k](t)/m  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{sector}
% \gbc{sector} produces the closed path consisting of a straight line
% of length \gbc{rad} from \gbc{center} in the direction \gbc{frtheta},
% thence along an arc of the circle centered at \gbc{center} to angle
% \gbc{totheta}, and then along the straight line back to \gbc{center}.
%    \begin{macrocode}
vardef sector (expr center, rad, frtheta, totheta) =
  center -- arcalt (center, rad, frtheta, totheta) -- cycle
enddef;

%    \end{macrocode}
%
% \subsection{Plotting of functions}\label{functionplots}
%
% In these macros, if the boolean argument \gbc{sm} is true then the
% path returned will be a B\'ezier, otherwise it will be a polyline. If a
% \gbc{tens} parameter exists, then the smooth version will have that
% value of tension, otherwise the value of \gbc{default_tension} is used.
% These two parameters are simply passed to \gbc{mkpath} by \gbc{mkfcn},
% and all these macros call \gbc{mkfcn}.
%
% \DescribeRoutine{mkfcn}
% In this command the text parameter \gbc{pf} should be the name of a
% function of some sort that can take a numeric value in parentheses and
% return a pair expression. The parameters \gbc{bmin}, \gbc{bmax} and
% \gbc{bst} determine a sequence of numeric values starting at \gbc{bmin},
% stepping by \gbc{bst} and ending with \gbc{bmax}. These are fed to
% \gbc{pf} and the resulting pairs stored in an array. Then either
% \gbc{mkpoly} or \gbc{mksmooth} is called with the tension \gbc{tens} and
% the name of the array.
%
% For stability, we don't actually step by \gbc{bst}, but round
% \gbc{(bmax-bmin)/bst} and step that many equal steps. We first adjust
% the step size upward so the number of steps doesn't exceed
% \gbc{infinity}. The path is forced to begin at \gbc{pf(bmin)} and
% end at \gbc{pf(bmax)} even if that is not an integer multiple of
% \gbc{bst}.
%
% \DescribeRoutine{tfcn}
% The macro \gbc{tfcn} is included for backward compatibility.
%    \begin{macrocode}
vardef mkfcn (expr sm, tens) (expr bmin, bmax, bst) (text pf) =
  save _p;  pair _p[];  _p := 0;
  save _dx, _n, _r;  numeric _dx, _n, _r;
  if bmax = bmin:  _n := 1;
  else:
    _r := bmax - bmin;
    _dx := max (abs(bst), nottoosmall*abs(_r), epsilon);
    _n := emax (round(abs(_r)/_dx), 1);
  fi
  for _i = 0 upto _n:  _p[incr _p] := pf(bmin + _i/_n*_r);  endfor
  mkpath (sm, tens, false, _p)
enddef;

def tfcn (expr sm) =  mkfcn (sm, default_tension)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{parafcn}
% This is like \gbc{mkfcn}, but the text argument is not a pair
% valued function, but rather a text parameter containing code that, when
% copied literally into a vardef, defines a function in which \gbc{t} is
% the argument, and which returns a pair.
%
% Older files are supported with a definition of \gbc{parafcn} that calls
% \gbc{tparafcn} with \gbc{default_tension}. I should have made this easier
% by reversing the smoothness and tension arguments, but for backward
% compatibility I have to leave it thus. Other commands implement \mfpic's
% tension options: \gbc{function} and \gbc{plrfcn}. They also have forms
% that accept a tension argument (\gbc{tfunction} and \gbc{tplrfcn}) and
% call them with the default tension.
%    \begin{macrocode}
def parafcn (expr sm) =  tparafcn (sm, default_tension)  enddef;
vardef tparafcn (expr sm, tn) (expr bmin, bmax, bst) (text pf) =
  save _fp;  vardef _fp (expr t) =  pf  enddef;
  mkfcn (sm, tn) (bmin, bmax, bst) (_fp)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{xfcn}
% This first converts its final argument, which should be a numeric
% valued function \gbc{f}, to a pair valued function \gbc{(x, f(x))}, then
% calls \gbc{mkfcn} to return the path that should be the graph of $f(x)$.
%    \begin{macrocode}
vardef xfcn (expr sm) (expr xmin, xmax, st) (text _fx) =
  save _fp;  vardef _fp (expr _x) =  (_x, _fx(_x))  enddef;
  mkfcn (sm, default_tension) (xmin, xmax, st) (_fp)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{function}
% This is like \gbc{xfcn} but its last argument, instead of a function,
% is a text argument that can be copied literally it into a vardef so as
% to define a pair valued function with a literal \gbc{x} as the argument.
%
% \DescribeRoutine{btwnfcn}
% This is mainly for the sake of simpler \mfpic{} output, implementing
% the \cs{btwnfcn} macro. One could code it in \mfpic{} macros with two
% calls to \gbc{function}.
%
% \DescribeRoutine{belowfcn}
% This is essentially \gbc{btwnfcn} with the first function identically
% 0, but it more efficiently graphs $0$ with one straight line
% rather than several end-to-end.
%    \begin{macrocode}
def function (expr sm) =  tfunction (sm, default_tension)  enddef;
vardef tfunction (expr sm, tens, xmin, xmax, st) (text _fx) =
  save _fp;  vardef _fp (expr x) =  (x, _fx)  enddef;
  mkfcn (sm, tens) (xmin, xmax, st) (_fp)
enddef;

def btwnfcn (expr sm) =  tbtwnfcn (sm, default_tension)  enddef;
vardef tbtwnfcn (expr sm, tn, xlo, xhi, st)(text _fx)(text _gx) =
  tfunction (sm, tn) (xlo, xhi, st) (_fx) --
  ( reverse tfunction (sm, tn) (xlo, xhi, st) (_gx) ) -- cycle
enddef;

def belowfcn (expr sm) =  tbelowfcn (sm, default_tension)  enddef;
vardef tbelowfcn (expr sm, tn, xlo, xhi, st)(text _fx) =
  (xlo,0)--(xhi,0)--
    (reverse tfunction (sm, tn, xlo, xhi, st)(_fx))--cycle
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{rfcn}
% This takes the name of a function \gbc{f} which is a numeric
% valued function of a numeric parameter. It interprets it as a polar
% curve $(\theta, f(\theta))$, converts that to a curve in rectangular
% coordinates and calls \gbc{mkfcn} on it.
%    \begin{macrocode}
vardef rfcn (expr sm, tmin, tmax, st) (text ft) =
  save _fq;  vardef _fq (expr t) =  (ft(t)) * (dir t)  enddef;
  mkfcn (sm, default_tension) (tmin, tmax, st) (_fq)
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{plrfcn}
% This is like \gbc{rfcn}, but with a text argument containing code that
% can be copied literally into a \mfc{vardef} creating a numeric function
% with a literal \gbc{t} as the parameter (representing $\theta$).
%
% \DescribeRoutine{btwnplrfcn}
% The macro \gbc{btwnplrfcn} is the polar version of \gbc{btwnfcn}.
%    \begin{macrocode}
def plrfcn (expr sm) =  tplrfcn (sm, default_tension)  enddef;
vardef tplrfcn (expr sm, tens, tmin, tmax, st) (text ft) =
  save _fq;  vardef _fq (expr t) =  (ft) * (dir t)  enddef;
  mkfcn (sm, tens) (tmin, tmax, st) (_fq)
enddef;

def btwnplrfcn (expr sm) =  tbtwnplrfcn (sm, default_tension)  enddef;
vardef tbtwnplrfcn (expr sm, tn, tlo, thi, st)(text _ft)(text _gt)=
  tplrfcn (sm, tn, tlo, thi, st) (_ft) --
  ( reverse tplrfcn (sm, tn, tlo, thi, st) (_gt) ) -- cycle
enddef;

def plrregion (expr sm) =  tplrregion (sm, default_tension)  enddef;
vardef tplrregion (expr sm, tn, tlo, thi, st) (text _ft) =
  (0,0)--tplrfcn (sm, tn, tlo, thi, st ) (_ft)--cycle
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mklevelset}
% This command assumes \gbc{inside_levelset} has been defined, which
% should be a boolean-valued function of two variables. It tries to create
% a path such that the expression is true inside the path and false
% outside it. The intended application is to obtain a path surrounding a
% region like $\lbrace (x,y) \mid F(x,y) > 0 \rbrace$.
%
% The parameters are
% \begin{itemize}
%   \item \gbc{sm}: Boolean, smooth path (true) or polyline.
%   \item \gbc{tens}: Numeric, the tension (if \gbc{sm} is true).
%   \item \gbc{X} and \gbc{Y}: A starting point where \gbc{_inside_}
%       should return true.
%   \item \gbc{t}: Numeric, a step size.
%   \item \gbc{a}, \gbc{b}, \gbc{c}, and \gbc{d}: Numeric, the limits
%       beyond which the search routine will not go. This is needed to
%       get a starting range for the binary chop method of \mfc{solve}.
%       In practice, the extent of the \mfpic{} figure will be used.
% \end{itemize}
%
% We use \mfc{solve} to find the first point $z\sb1$ to the right of the
% given point that is on the edge of the region. Then we find the first
% point of intersection between the circle at $z\sb1$ with radius \gbc{t}
% and the edge of the region. Continue from this new point to the next
% until (one hopes) we are within a distance \gbc{t} of the first point.
% The radius \gbc{t} should be in coordinates appropriate for the use:
% graph coordinates when used in \mfpic{} figures.
%
% The tolerance used in the first \mfc{solve} is \gbc{t/50}. In later uses
% it is an angle parameter and is set to a number of degrees sufficient to
% give a distance tolerance at least that.
%
% If the starting point \gbc{(X,Y)} does not actually satisfy the
% condition, a one point path is returned.
%    \begin{macrocode}
numeric tolerancefactor;
tolerancefactor := .02;

vardef mklevelset (expr sm, tens, X, Y, t, a, b, c, d) =
  save _inside_;
  vardef _inside_ (expr U, V) =
    inside_levelset (U, V) and between (a, b) (U) and between (c, d) (V)
  enddef;
  if not _inside_ (X, Y):
    GBwarn "Invalid seed point for levelset.";
    pairmax((a,c), pairmin((X,Y), (b,d)))&cycle
  else:
    save ls, W, A, B, prev, curr, seed;
    pair ls[], prev, curr, seed;
    seed := (X,Y);
    ls := 0;  W := 0;

    save _first_, _next_, get_next;
    vardef _first_ (expr U) =  _inside_ (U, Y)  enddef;
    vardef _next_ (expr ang) =
      _inside_ (X_curr + t * cosd ang, Y_curr + t * sind ang)
    enddef;
    def get_next (expr angA, angB) =
      X_curr := xpart curr;  Y_curr := ypart curr;
      ls[incr ls] := curr + t * dir (solve _next_ (angA, angB));
      prev := curr;  curr := ls[ls];
      W := W + anglefromto (prev - seed, curr - seed);
    enddef;

    interim tolerance := t*tolerancefactor;
    ls[incr ls] := (solve _first_ (X, b), Y);
    curr := ls[ls];
    interim tolerance := radian*tolerancefactor;
    get_next (180, 0);
    for n = 3 upto max_points:
      A := angle (curr - prev);
      get_next (A + 120, A - 120);
      exitif ((abs(W) > 180) or (ls > 10))
        and (abs(ls[ls] - ls1) < 1.2t);
    endfor
    mkpath (sm, tens, true) (ls)
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{levelset}
% This is the \mfpic{} interface. It checks the \gbc{t} parameter before
% passing it to \gbc{mklevelset}, making sure it is not zero, it passes
% appropriate limits, and defines boolean function \gbc{mklevelset}
% expects with literal \mfc{x} and \mfc{y} as parameters, using the text
% parameter \gbc{cond}.
%    \begin{macrocode}
numeric max_points;
max_points := 2000;

def levelset (expr s) =  tlevelset (s, default_tension)  enddef;
vardef tlevelset (expr smth, tens, seed, seg) (text cond) =
  save inside_levelset, _t;
  vardef inside_levelset (expr x, y) =  cond  enddef;
  _t := if seg <= 0:  emax (xpos-xneg, ypos-yneg)/max_points * 20
        else:  seg fi;
  mklevelset (smth, tens, xpart seed, ypart seed, _t)
    (xneg, xpos, yneg, ypos)
enddef;

%    \end{macrocode}
%
% Our next set of macros produce approximations to the solutions of
% differential equations. While we could have several different macros
% each using a different method (Euler, two-step Runge-Kutta, four-step
% Runge-Kutta, etc.), our point of view is that we just want to draw a
% reasonably accurate solution, so we only utilize one method: four-step
% Runge-Kutta. The variations we allow are the following:
% \begin{enumerate}
% \item Drawing the graph of a one-dimensional differential
%       equation,
%       \[ \frac{dy}{dx} = g(x,y)\,.\]
%
% \item Drawing the trajectory of a two-dimensional differential
%       equation,
%       \[ \left( \frac{dx}{dt},\frac{dy}{dt} \right) =
%       (f(x,y,t), g(x,y,t))\,.\]
% \end{enumerate}
% The first of these is implemented using the second with $f(x,y,t) \equiv
% 1$ and $g(x,y,t)$ not depending on $t$. The parameters passed include
% the starting point, the step size, the number of steps and an expression
% representing the right side of the equation.
%
% We do not use exactly the traditional Runge-Kutta method: we use the
% Runge-Kutta algorithm, but with a variable step size. The time step
% $\Delta t$ is chosen so that $|\mathbf{F}(x,y,t)|\Delta t$ equals the
% given step size parameter, and thus the parameter passed is actually a
% distance step. This makes drawing more stable, especially if the DE is
% one that produces an infinite path in finite time.
%
% This modification is itself unstable if $|\mathbf{F}|$ is very
% small (and impossible if it is zero), so we never use a $\Delta t$
% larger than the given step size parameter $\Delta s$. That is, we
% actually use $\Delta t = \Delta s/\max(1,|\mathbf{F}|)$.
%
% As with our other function-like paths, we offer two variants. The basic
% version has a final text parameter which is the name of a pair-valued
% function of a numeric (representing $t$) and a pair variable
% (representing $x$ and $y$). The other version takes a text
% parameter, which must be a pair-valued expression in \mfc{x}, \mfc{y}
% and \mfc{t}. This parameter is copied into the definition text of a
% function and then the first form is called with that function's name.
%
% Also like other function-like paths, we offer polygonal or smooth
% versions controlled by a boolean argument, and the smooth versions make
% use of a tension parameter.
%    \begin{macrocode}
def RKIV (expr sm) =  tRKIV (sm, default_tension)
enddef;
vardef tRKIV (expr sm, tens, zstart, ds, N) (text _RHS_) =
  save _trj, _ztr, _dz, _ztmp, _ctm;
  pair _trj[],          % The trajectory
       _ztr,            % current point
       _dz[],           % array[4] of displacements
       _ztmp;           % current point for calculating velocity
%
  _trj := N+1;          % ultimate size of _trj array
  _trj1 := _ztr := zstart;
  save _tt,             % current time
       _dt,             % current time step
       _th;             % current time plus half a step
  _tt := 0;
  for _idx := 2 upto _trj:
    _dt   := ds/emax(1,abs(_RHS_(_tt,_ztr)));
    _th   := _tt + .5_dt;
    _dz1  := _dt*_RHS_(_tt, _ztr);  % displacement based on current point
    _ztmp := _ztr + .5_dz1;         % 1st midpoint
  % use _th instead of twice calculating (_tt + .5_dt)
    _dz2  := _dt*_RHS_(_th, _ztmp); % displacement based on 1st midpoint
    _ztmp := _ztr + .5_dz2;         % 2nd midpoint
    _dz3  := _dt*_RHS_(_th, _ztmp); % displacement based on 2nd midpoint
    _ztmp := _ztr + _dz3;           % temporary end point
  % get time for next loop now since we need it right away in next line:
    _tt   := _tt + _dt;
    _dz4  := _dt*_RHS_(_tt, _ztmp); % displacement based on end point
  % get next point
    _ztr  := _ztr + (_dz1 + 2_dz2 + 2_dz3 + _dz4)/6;
    _trj[_idx] := _ztr;
  endfor
  mkpath (sm, tens, false, _trj)
enddef;

def xyRKIV (expr sm) = txyRKIV (sm, default_tension)
enddef;
vardef txyRKIV (expr sm, tens, zstart, ds, N) (text _RHS_) =
  save _fgxy, __fgxy;
  vardef __fgxy (expr t, x, y) =  _RHS_  enddef;
  vardef _fgxy (expr t, Z) = __fgxy(t, xpart Z, ypart Z) enddef;
  tRKIV (sm, tens, zstart, ds, N) (_fgxy)
enddef;

def odeRKIV (expr sm) = todeRKIV (sm, default_tension)
enddef;
vardef todeRKIV (expr sm, tens, xstart, ystart, ds, N)
(text _fxy) =
  txyRKIV (sm, tens, (xstart, ystart), ds, N) ((1, _fxy))
enddef;

%    \end{macrocode}
%
% \section{Modification of Paths}\label{modification}
%
% \subsection{Closing a path}\label{closing}
%
% In \MF{} one closes a path with any legal path connection between the
% last point and the keyword \mfc{cycle}. Connecting the last point to the
% first point is not enough. \Grafbase{} commands provide a few different
% ways. All the commads take an undelimited path expression as the last
% parameter and return a cycle (closed path). If the path is already
% closed, it is returned unchanged.
%
% All the closure commands have a version with a tension parameter when
% that makes sense. Those versions create the join with the supplied
% tension. The ones where it doesn't make sense are \gbc{lclosed},
% \gbc{cbclosed} and \gbc{qbclosed}. The first always uses a straight line
% and the other two require explicit controls.
%
% \DescribeRoutine{lclosed}
% This closes with a line segment. If the first and last point are
% already equal, we just use the \mfc{\&} path join.
%    \begin{macrocode}
vardef lclosed expr f =
  f
  if not cycle f:
    if pnt0(f) = pnt[infinity](f):  &  else:  --  fi  cycle
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{sclosed}
% This closes the path in the manner that \gbc{mksmooth} creates a path.
% This will change the first and last segment of the original path. In
% particular, if there are fewer than three segments, the whole path can
% be different.
% \DescribeRoutine{sclosedt}
% It has a variant \gbc{sclosedt} that takes a tension argument.
%    \begin{macrocode}
def sclosed =  sclosedt (default_tension)  enddef;
vardef sclosedt (expr t) expr f =
  if cycle f:  f
  else:  save n;  n := length f;
    if n = 0:  f&cycle
    elseif n = 1:  pnt0(f)..tension t..pnt1(f)..tension t..cycle
    else:
      (pnt0 (f)) { (pnt1(f)) - (pnt[n] (f)) }..tension t
         ..(subpath (1, n-1) of f)..tension t
         ..(pnt[n](f)) { pnt0(f) - pnt[n-1](f) }
         ..tension t..cycle
    fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{bclosed}
% This closes with the basic default \MF{} Bezi\'er. It is a smooth
% closure, but it does not have the same direction at the endpoints
% that \gbc{mksmooth (true)} would have produced. It has a tense variant
% \DescribeRoutine{bclosedt}
% \gbc{bclosedt}
%    \begin{macrocode}
def bclosed =  bclosedt (default_tension)  enddef;
vardef bclosedt (expr t) expr f =
  f
  if not cycle f:
    if pnt0(f) = pnt[infinity](f):  &  else:  ..tension t..  fi  cycle
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{uclosed}
% Same as \gbc{bclosed}. Retained for backward compatibility. There is
% a tense variant only for
% \DescribeRoutine{uclosedt}
% consistency.
%    \begin{macrocode}
def uclosed =  bclosed  enddef;
def uclosedt =  bclosedt  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{cbcontrols}
% This utility is for use in \gbc{cbclosed}. It converts Bezier segment
% key points of a path \gbc{f}, to cubic B-spline control points stored
% in an array \gbc{b}. The data needed are the first point and first two
% control points of a path. It is used twice in \gbc{cbclosed} on a path
% and on its reverse. The appropriate three points are passed in the array
% \gbc{t}.
% The B-spline points needed are \gbc{b1} and \gbc{b4}. The extra two
% points \gbc{b2} and \gbc{b3} divide the line from \gbc{b1} to \gbc{b4}
% into thirds and will be turned into Bezier control points of a new path
% segment.
%    \begin{macrocode}
def cbcontrols (suffix b, t) =
  b1 := 2[t3, t2];
  b2 := 2[t2, t1];
  b3 := 2[b1, b2];
  b4 := 2[b2, b3];
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{cbclosed}
% This closes a path with a cubic B-spline. If the path \gbc{f} had been
% produced by \gbc{opencbs}, then \gbc{q1} and \gbc{q4} would have been the
% last two points in the argument list, and \gbc{p4} and \gbc{p1} would
% have been the first two. We just use them and mimic the effect of
% \gbc{closedcbs}.
%    \begin{macrocode}
vardef cbclosed expr f =
  save n;  n := length f;
  if cycle f:  f
  elseif n = 0:  f&cycle
  else:
    save p, q, t;  pair p[], q[], t[];
    t1 := pnt0(f);  t2 := post0(f);  t3 := pre1(f);
    cbcontrols (p, t);       % defines p1 to p4
    t1 := pnt[n](f);  t2 := pre[n](f);  t3 := post[n-1](f);
    cbcontrols (q, t);       % defines q1 to q4
    f..controls q2 and q3..opencbs (q1,q4,p4,p1)
     ..controls p3 and p2..cycle
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{qbclosed}
% It seemed wrong to be able to close with a cubic B-spline but not a
% quadratic B-spline; therefore I have add such a possibility. We
% calculate B-spline controls \gbc{p[n]} that will agree with those of
% \gbc{f}, if \gbc{f} had been created as a quadratic B-spline. Note
% that \gbc{cbclosed} required three \MF{} links to close the curve;
% \gbc{qbclosed} only requires two (\gbc{mkqbs} on an array of $n$
% points makes $n-2$ links).
%    \begin{macrocode}
vardef qbclosed expr f =
  if cycle f:  f
  else:  save n;  n := length f;
    if n = 0:  f&cycle
    else:
      save p;  pair p[];  p := 4;
      p1 := (3/2)[pnt[n](f), pre[n](f)];
      p2 := 2[p1, pnt[n](f)];
      p4 := (3/2)[pnt 0 (f), post0 (f)];
      p3 := 2[p4, pnt 0 (f)];
      f & mkqbs (p) & cycle
    fi
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{makesector}
% This makes sense only if the path being modified is an arc. It closes
% the arc by connecting its ends to the center of the circle, as
% computed by \gbc{pathcenter}.
%    \begin{macrocode}
vardef makesector expr p =  (pathcenter p)--p--cycle  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{arccomplement}
% Getting the complement of an arc is easy if the arc is specified by
% three points. So we just select three points on the arc and do that.
% The \gbc{setpairs} statement makes \gbc{pp1}, \gbc{pp2} and \gbc{pp3}
% three points on the arc \gbc{p} in order. The arc we want goes from
% \gbc{pp3} to \gbc{pp1} with angle twice that of the corner angle at
% \gbc{pp2}.
%    \begin{macrocode}
vardef arccomplement expr p =
  if cycle p:  onepointpath (false, pnt0(p))
  else:
    setnumeric (nn) length p;
    setpairs (pp) (pnt0(p), pnt[.5nn](p), pnt[nn](p));
    arcpps (pp3,pp1,2*cornerangle(pp2,pp3,pp1))
  fi
enddef;

%    \end{macrocode}
%
% \subsection{Trimming a path}\label{trimming}
%
% \DescribeRoutine{cutoffbefore}
% This is a useful utility operation present in \file{plain.mp} (as
% \mfc{cutbefore}) but missing from \file{plain.mf}. We write a different
% version for our purposes: it has the syntax of most of our path
% modification commands, plus the first loop tries to avoid a bug (or
% perhaps inaccuracy) in \mfc{intersectiontimes} which can return an
% intersection time in a later segment of \gbc{f} than the first
% intersection point.
%
% \DescribeRoutine{cutoffafter}
% What it and \gbc{cutoffafter} do is return the second path with the
% part before\slash after the first path removed.
%    \begin{macrocode}
%<MF>path cuttings;
vardef cutoffbefore (expr b) expr f =
  save t, n;  n := length f;
  if n > 0:
    for k = 1 upto n:
      exitif (subpath (0,k) of f) intersects b;
    endfor
    if _Xtime < 0:
      cuttings := pnt0 (f){0,0};
      f
    else:
      cuttings := subpath (0,_Xtime) of f;
      subpath (_Xtime, n) of f
    fi
  else:  f
  fi
enddef;

vardef cutoffafter (expr b) expr f =
  setpath (g) cutoffbefore (b) reverse f;
  cuttings := reverse cuttings;
  reverse g
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{trimmedpath}
% This takes two lengths and a path and trims off the ends of the path
% that lie within the given lengths of the endpoints. The lengths are in
% device coordinates, the path in graph coordinates.
%    \begin{macrocode}
vardef trimmedpath (expr btrim, etrim) expr f =
  save g, h;  path g, h;
  g := invvconv (fullcircle scaled 2btrim) shifted pnt0(f);
  h := invvconv (fullcircle scaled 2etrim) shifted pnt[length f] (f);
  cutoffafter (h) cutoffbefore (g) f
enddef;

%    \end{macrocode}
%
% \subsection{Creating arrows}\label{arrows}
%
% First, some better \mfc{direction} commands. They makes use of the fact
% (easily proved) that a cubic B\'ezier
% \[
%   z\sb0(1 - t)^3 + 3z\sb1(1-t)^2t + 3z\sb2(1-t)t^2 + z\sb3t^3
% \]
% has a tangent at $z\sb0$ equal to the first one of $z\sb{j} -z\sb0$
% that is nonzero.
%
% \DescribeRoutine{__dir}
% \gbc{__dir} gets the direction at point 0 for an arbitrary path.
% \gbc{postdirection}
% \DescribeRoutine{postdirection}
% reduces to this case using \mfc{subpath}. If the postdirection is
% $(0,0)$, that means the path is trivial from that point to the end so we
% are effectively at an endpoint (noncyclic path) and we use the incoming
% direction. If that is $(0,0)$, the path is trivial.
% \DescribeRoutine{predirection}
% \gbc{predirection} just runs \gbc{postdirection} on the reversed path.
%
% \DescribeRoutine{trivial}
% This returns \mfc{true} if a path has direction vector $(0,0)$ (as
% determined by \gbc{__dir}).
%    \begin{macrocode}
vardef predirection@# (expr p) =
  - postdirection[length p - @#] (reverse p)
enddef;

vardef postdirection@# (expr p) =
  save _n;  _n := length (p);
  setpair (v) __dir (subpath (@#, @# + _n) of p);
  if v = origin:
    v := - __dir (subpath (@#, @# - _n) of p);
  fi
  v
enddef;

vardef __dir (expr p) =
  save v, w;  pair v, w;  w := pnt0 (p);
  v := origin;
  for n = 1 upto length (p):
    v := post[n-1] (p) - w;  exitif v <> origin;
    v := pre [ n ] (p) - w;  exitif v <> origin;
    v := pnt [ n ] (p) - w;  exitif v <> origin;
  endfor
  sgn v
enddef;

vardef trivial expr p =  (__dir (p) = origin)  enddef;

%    \end{macrocode}
%
% Arrowheads can be just two straight lines at an angle placed on the end
% of a curve, or it can be a filled triangle. \grafbase{} permits both,
% but it also allows the two lines (or the corresponding sides of the
% triangle) to be gracefully concave and tangent to the path at the
% endpoint of the path. The parameters controlling the shape of the arrowhead
% are the two numerics \gbc{hdwdr}, the ratio of the length to width of the
% arrowhead, and \gbc{hdten}, the tension in the two angled curves. By
% default, one side of an arrowhead is just the \MF{} path
% \mfc{a..b\marg{\meta{tangent}}}, where \mfc{a} is the base of the
% arrowhead (calculated from \gbc{hdwdr}) and \gbc{b} is the end of the
% path and \meta{tangent} is the direction of the path at that
% point. The curve can be straightened by increasing \gbc{hdten}, the
% head widened by increasing \gbc{hdwdr}
%
% The arrowhead is drawn by drawing two of the curves described above. If
% \gbc{hfilled} is \mfc{true}, the two base points (\gbc{a} above) are
% connected and the three sided region filled.
%    \begin{macrocode}
newinternal hdwdr, hdten;
boolean hfilled;

%    \end{macrocode}
%
% \DescribeRoutine{headshape}
% The following utility not only adjusts the above parameters, but
% creates the arrowhead paths as it does so. Call it with two pure numbers
% \gbc{wr} and \gbc{tens} for the \gbc{hdwdr} and \gbc{hdten}, and a
% boolean \gbc{fil} for \gbc{hfilled}. The paths include two harpoon
% tips, an arrowhead, and the paths that give regions that will be erased
% when requested.
%    \begin{macrocode}
def headshape (expr wr, tens, fil) =
  interim hdwdr := wr;
  interim hdten := if tens>0:  tens  else:  default_tension fi;
  if hdten < .75:  hdten := .75;  fi
  setboolean (hfilled) fil;
  mkheadpaths;
enddef;
def mkheadpaths =
  save Arrowhead, Leftharpoon, Rightharpoon;
  path Arrowhead, Leftharpoon, Rightharpoon,
       Arrowhead.clear, Leftharpoon.clear, Rightharpoon.clear;
  Rightharpoon := (0,0){down}..tension hdten..(.5hdwdr,-1);
  Rightharpoon.clear := Rightharpoon--(.5hdwdr,0)--cycle;
  Leftharpoon  := (reverse Rightharpoon) xscaled -1;
  Leftharpoon.clear := (reverse Rightharpoon.clear) xscaled -1;
  Arrowhead := Leftharpoon & Rightharpoon;
  Arrowhead.clear := Leftharpoon.clear & Rightharpoon.clear & cycle;
  if hfilled:
    Arrowhead := Arrowhead--cycle;
    Rightharpoon := Rightharpoon--(0,-1)--cycle;
    Leftharpoon := Leftharpoon--(0,-1)--cycle;
  fi
enddef;
headshape (1,1,false);

%    \end{macrocode}
%
% \DescribeRoutine{ahead}
% This command draws an arrowhead. Current code for arrows no longer uses
% it. \gbc{front} and \gbc{back} are in device coordinates. They are the
% point of the arrowhead (\gbc{front}) and the point such that \gbc{front
% - back} is as long as the arrowhead and points in the direction of the
% arrow. We use the ratio \gbc{hwr} to compute the other two corners. So
% \gbc{side} is the vector from \gbc{back} to one of the corners \gbc{p1},
% and the other corner is on the other side. \gbc{f} is the path of the
% arrowhead.
%
% If \gbc{filled} is true we close the curve draw it and fill it,
% otherwise we just draw it. (To only fill it, make the pen width $0$.)
% \gbc{clr} is the color used to draw or fill it.
%
% For backward compatibility we define \gbc{head}. In \MF{} \gbc{head}
% didn't have a color parameter, while in \MP{} it has always had one,
% in retrospect, this was not a good idea, and I should have followed
% the pattern of other macros. However, \gbc{head} was never a user-level
% macro and it didn't seem to mattered at the time.
%    \begin{macrocode}
%<MF>def head =  ahead (headcolor)  enddef;
%<MP>def head =  ahead  enddef;

vardef ahead (expr clr, front, back, hwr, tens, filled) =
  settension (_tn) tens;  fixtension (_tn);
  if front <> back:
    setpair (side) (hwr/2) * ((front-back) rotated 90);
    setpath (f) (back + side)..tension _tn..
       {front-back}front{back-front}..tension _tn..(back - side);
    if clearhead:
      safeunfill (back - side)--(front-side)--(front+side)--
        (back+side) & f & cycle;
      colorsafedraw (background) (back - side)--(front-side)--
        (front+side)--(back+side) & f & cycle;
    fi
    if filled:
      f := f--cycle;
      colorsafefill (clr) f;
    fi
    colorsafedraw (clr) f;
  fi
enddef;

%    \end{macrocode}
%
% It is a fact of life that, unless the path to which the head is added is
% a straight line, the arrowhead may appear to point in the wrong
% direction. But I know of no automatic way of making it always look
% correct. Therefore \grafbase{} and \mfpic{} have provided a means to
% micro-adjust the head.
%
% The various arrow creation commands take a path expression \gbc{f} in
% graph coordinates, puts a head/tail/decoration on it and returns
% \gbc{f}. There are also four parameters determining (i)~the color,
% (ii)~the size, (iii)~a rotation adjustment, and (iv)~a position
% adjustment.
%
% \DescribeRoutine{headpath}
% \RoutineIndex{colorheadpath}
% \RoutineIndex{headpathx}
% \RoutineIndex{colorheadpathx}
% \gbc{headpath} calls \gbc{Gheadpath}, a more general command that takes
% a boolean expression and a shape (path) suffix as arguments. It
% supplies \mfc{false} for the boolean and \gbc{Arrowhead} for the shape.
% The \gbc{headpathx} version differs only in that the boolean is
% \mfc{true}. The \gbc{color...} versions call \gbc{colorGheadpath} in the
% same way, but require that a color parameter follow.
%    \begin{macrocode}
def headpath =  Gheadpath (false) (Arrowhead)  enddef;
def headpathx =  Gheadpath (true) (Arrowhead)  enddef;

def colorheadpath =  colorGheadpath (false) (Arrowhead)  enddef;
def colorheadpathx =  colorGheadpath (true) (Arrowhead)  enddef;

%    \end{macrocode}
%
% \DescribeRoutine{Gheadpath}
% \RoutineIndex{colorGheadpath}
% \RoutineIndex{Gheadpathx}
% \RoutineIndex{colorGheadpathx}
% For general arrowhead shapes we require two paths; one giving the shape
% of the head and the other the shape that is cleared when the boolean
% parameter \gbc{trim} is true. We pass this information by name with a
% suffix parameter \gbc{ah} that names the head shape. Then \gbc{ah.clear}
% names the cleared region. If \gbc{ah} is a cycle, the head is filled,
% otherwise it is drawn with the current \gbc{drawpen}.
% If one wants it drawn and center erased, one could place first the
% solid (filled) version with color \mfc{background}, then the outline
% (drawn only) version.
%
% We also need to know which point on the path is the tip, and
% \gbc{ah.tip} provides that. If \gbc{ah.clear} is undefined, the clearing
% is silently skipped. If \gbc{ah.tip} is unknown, it is taken to be
% \mfc{(0,0)}. The head shape is assumed to be initially defined
% pointing \mfc{up} (to match predefined shapes like \gbc{Triangle} and
% \gbc{Diamond}).
%
% The standard symbols for \gbc{plot} have both the clearing path and
% the tip defined. Thus one can produce \gbc{Diamond}-tipped arrows. The
% old arrow heads are given by \gbc{Arrowhead}, which is redefined with
% every call to \gbc{headshape}.
%
% If the \gbc{rot} and \gbc{pos} parameters are zero, the head is placed
% with its tip at the end of the path, pointing in the direction of the
% path at that point. Otherwise, the head is rotated around the tip by the
% amount \gbc{rot} and \emph{then} shifted backward along its new
% direction by the amount \gbc{pos}.
%
% We have a version that takes a color parameter. The simpler version
% simply calls it with \gbc{clr} equal to \gbc{headcolor}.
%    \begin{macrocode}
def Gheadpath (expr trim) (suffix ah) =
  colorGheadpath (trim) (ah) (headcolor)
enddef;
vardef colorGheadpath
(expr trim) (suffix ah) (expr clr, sc, rot, pos) expr f =
  if (sc <> 0) and (known ah) and (path ah):
    convertpath (_g) f;
    setpair (_P) predirection[length _g] (_g);
    if _P <> origin:
      _P := _P rotated rot;
      setnumeric (_ang) anglefromto (up, _P);
      _P := pnt[length _g] (_g) - pos * _P;
      setpair (_tip) if known ah.tip:  ah.tip  else:  origin fi;

      if trim:
        if known ah.clear:
          safeunfill (ah.clear shifted - _tip)
            scaled sc
            rotated _ang
            shifted _P;
        fi
        setnumeric (_ys) max(bp, penwd, last_dot_size);
        safeunfill cut_path
%<MP>          xscaled sc yscaled _ys
%<MF>          xscaled ceiling sc yscaled ceiling _ys
          rotated _ang shifted _P;
      fi
      if cycle ah:  colorsafefill  else:  colorsafedraw  fi  (clr)
        (ah shifted -_tip)
          scaled sc
          rotated _ang
          shifted _P;
    fi
  fi
  f
enddef;


%    \end{macrocode}
%
% \DescribeVariable{cut_path}
% Additional clearing path, almost the same as plain.mf's \gbc{cut_} (no
% \gbc{cut_} in plain.mp) but rotated, and scaled differently. The odd
% scaling is so that if yscaled by the diameter of a dot, and the dot
% happens to be digitized to a square shape, then the \gbc{cut_path},
% centered at the center of the dot and rotated 45 degrees, will encompass
% the whole square (theoretically).
%    \begin{macrocode}
path cut_path;  cut_path := (.5,0)--(.5,.71)--(-.5,.71)--(-.5,0)--cycle;

%    \end{macrocode}
%
% \DescribeRoutine{tailpath}
% \RoutineIndex{colortailpath}
% The macro \gbc{tailpath} places a tail at the start of a path. It is
% almost like \gbc{Gheadpath} except there is no clearing done and the tip
% is at the start (point 0) of the path. Also, the position parameter
% \gbc{pos} is a forward shift.
%
% \DescribeRoutine{midpath}
% \RoutineIndex{colormidpath}
% The macro \gbc{midpath} is just like \gbc{tailpath} except it puts the
% given shape somewhere in the middle of the path. Its position parameter
% indicates the fraction of the length of the path where the shape is to
% be placed. This works best in two cases: the shape has a definite
% direction (like the \gbc{Arrowhead}) and the tip is placed at the given
% position, or the shape has a center of symmetry and that is placed at
% the given position. We obtain this in most cases by shifting $(0,0)$ to
% that position. The standard arrowhead has its tip at this point, and the
% standard symbols (with the exception of \gbc{Circle}) have their center
% of symmetry there.
%    \begin{macrocode}
def tailpath (suffix sh) =  colortailpath (sh) (headcolor)  enddef;
vardef colortailpath (suffix sh) (expr clr, sc, rot, pos) expr f =
  if (sc <> 0) and (known sh) and (path sh):
    convertpath (_g) f;
    setpair(_P) postdirection0 (_g);
    if _P <> origin:
      _P := _P rotated rot;
      if cycle sh:  colorsafefill  else:  colorsafedraw  fi  (clr)
        (sh if known sh.tip:  shifted -sh.tip fi)
          scaled sc
          rotated anglefromto (up, _P)
          shifted (pnt0 (_g) + pos * _P);
    fi
  fi
  f
enddef;

def midpath (suffix sh) =  colormidpath (sh) (headcolor)  enddef;
vardef colormidpath (suffix sh) (expr clr, sc, rot, pos) expr f =
  if (sc <> 0) and (known sh) and (path sh):
    convertpath (_g) f;
    setnumeric (_t) pathtime[pos] (_g);
    setpair (_P) postdirection[_t] (_g);
    if _P <> origin:
      _P := _P rotated rot;
      if cycle sh:  colorsafefill  else:  colorsafedraw  fi  (clr)
        sh scaled sc
           rotated anglefromto (up, _P)
           shifted (pnt[_t] (_g));
    fi
  fi
  f
enddef;

%    \end{macrocode}
%
% \subsection{Randomizing a path}
%
% In order to randomly change a path, we need to randomly change its
% points and its controls. If we just apply independent random shifts to
% every point and control point, it could happen that the direction from a
% point to a control changes dramatically, introducing a wild change even
% with a small shift (if point and control were very close to begin with).
% Also, this method would almost guarantee that a smooth path would
% randomize into one with all corners. Our solution to these problems is
% in the following paragraph.
%
% If $z\sb0$ and $z\sb3$ are the start and end points, with controls
% $z\sb1$ and $z\sb2$, then we randomize $z\sb0$ and $z\sb3$ using a
% random shift with size supplied as a parameter. If there was a
% preceeding segment, its ending angle and the angle of $z\sb1-z\sb0$
% determine an angle difference which we multiply by a random factor.
% This determines the direction to the new control point. If there was no
% preceeding segment we rotate $z\sb1-z\sb0$ a random amount. Finally we
% randomly scale $|z\sb1-z\sb0|$.
%
% The following `\gbc{deviate}s' are analogous to \MF{}'s
% \mfc{uniformdeviate}.
% \DescribeRoutine{signeddeviate}
% The first, \gbc{signeddeviate X}, produces a random number uniformly
% distributed in $(-X, X)$. The second,
% \DescribeRoutine{scaledeviate}
% \gbc{scaledeviate (W, A)}, produces a pair in a particular direction
% with length distributed in $(2^{-w}, 2^w)$.
% \DescribeRoutine{polardeviate}
% The third, \gbc{polardeviate R} produces a pair whose polar coordinates
% are separately uniformly distributed, the radius over the interval $(0,
% R)$ the angle over $(0,360)$.
% \DescribeRoutine{xydeviate}
% The last, \gbc{xydeviate (X,Y)}, produces a pair uniformly distributed
% over the rectangle with corners at $(-X,-Y)$ and $(X,Y)$.
%
% \DescribeRoutine{randompair}
% Finally, \gbc{randompair} runs \gbc{polardeviate} if \gbc{X} is
% numeric and \gbc{pairdeviate} if it is a pair.
%    \begin{macrocode}
vardef signeddeviate primary X =
  (uniformdeviate 1)[-X,X]
enddef;
vardef scaledeviate (expr W, A) =
  2 ** (signeddeviate W) * dir A
enddef;
vardef polardeviate primary R =
  (uniformdeviate abs(R)) * dir uniformdeviate 360
enddef;
vardef xydeviate primary Z =
  (signeddeviate (xpart Z), signeddeviate (ypart Z))
enddef;

vardef randompair (expr maxshift) =
  if numeric maxshift:  polardeviate (maxshift)
  elseif pair maxshift:  xydeviate (maxshift)
  else:  (0,0)
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{randompath}
% The \gbc{randompath} macro returns a path formed from \gbc{f} by
% shifting each of its point with independent instances of
% \gbc{randompair (maxshift)}. Its control points are also modified, but
% the algorithm is not as simple.
%
% Let $X$ be one of the points of \gbc{f}, with precontrol $X-U$ and
% postcontrol $X+V$. We create random $X'$, $U'$ and $V'$ as follows.
%
% Let $S$ be the pair that results from \gbc{randompair(maxshift)}, let
% $w$ be the value of \gbc{weirdness}, let $\eta\sb j(w)$ be the value of
% the $j$th instance of \gbc{signeddeviate ($w$)}.
% Then $X' = X + S$, $U'$ is $U$ rotated $\beta=30\eta\sb1(w)$ and scaled
% $\sigma = 2^{\eta\sb2(w)}$. Let $\alpha$ be the angle between the two
% vectors $U$ and $V$. Then $V'$ is $\sigma V$ rotated to make the angle
% between $U'$ and $V'$ equal to $\alpha 2^{\eta\sb3(w)}$.
% In the new path, the point is $X'$ with precontrol $X'-U'$ and
% postcontrol $X' + V'$. Note that if the path is smooth at $X$, then
% $\alpha = 0$ and the new angle is also $0$.
%
% \DescribeRoutine{randomlines}
% This is a simpler version that simply shifts the nodes and connects
% the results with straight lines. It is intended to be applied to
% polyline paths.
%
% \DescribeRoutine{detrivialized}
% We start with a routine that strips out trivial segments from a path.
% This makes some loops a lot easier. We wouldn't want to differently
% shift the two (equal) endpoints of a trivial segment, making it
% nontrivial.
%    \begin{macrocode}
vardef detrivialized expr f =
  save g;  path p, g[];  g := 0;
  for k = 1 upto length f:
    p := subpath (k-1,k) of f;
    if not trivial p:  g[incr g] := p;  fi
  endfor
  if g = 0:  onepointpath (cycle f, pnt0(f))
  else:  g1 for k = 2 upto g:  &g[k] endfor if cycle f:  &cycle  fi
  fi
enddef;

vardef randompath (expr maxshift, weirdness) expr f =
  save g, n;  path g;
  g := detrivialized f;
  n := length g;
  if n = 0:
    f shifted randompair (maxshift)
  else:
    save X, U, V;
    pair X[], U[], V[];
    if cycle g:  n := n - 1;  fi
    for k = 0 upto n:
      X[k] := pnt[k](g);
      U[k] := X[k] - pre[k](g);
      V[k] := post[k](g) - X[k];
    endfor
    save A, B;
    for k := 0 upto n:
      X[k] := X[k] shifted randompair (maxshift);
      A := anglefromto (U[k],V[k]);
      B := signeddeviate (30weirdness);
      U[k] := X[k] - (U[k] zscaled scaledeviate (weirdness,B));
      B := B - A + A * (2 ** signeddeviate weirdness);
      V[k] := X[k] + (V[k] zscaled scaledeviate (weirdness,B));
    endfor
    X0 for k = 1 upto n:
        .. controls V[k-1] and U[k] .. X[k]
       endfor
    if cycle g:
        .. controls V[n] and U0 .. cycle
    fi
  fi
enddef;

vardef randomlines (expr maxshift) expr f =
  save g, n;  path g;
  g := detrivialized f;
  n := length g;
  if n = 0:
    f shifted randompair (maxshift)
  else:
    if cycle g:  n := n - 1;  fi
    (pnt0(g) shifted randompair (maxshift))
      for k = 1 upto n:
        -- (pnt[k](g) shifted randompair (maxshift))
      endfor
      if cycle g:
        -- cycle
      fi
  fi
enddef;

%    \end{macrocode}
%
% \subsection{Parallelling a path}
%
% \DescribeRoutine{parasegment}
% This creates a path parallel to a given cubic Bezier segment \gbc{f}.
% It should be called by a command (such as \gbc{parapath}) that makes
% sure \gbc{f} is nontrivial (meaning the directions are non-zero). It
% splits the segment into subsegments for accuracy. Its arguments are the
% distance the original path is shifted, the number of subsegments to
% split into, and the path.
%    \begin{macrocode}
vardef parasegment (expr d, segs, f) =
  if d = 0:  f
  else:
    save u, v, t;  pair u[], v[];
    for n = 0 upto segs:
      t := n/segs;
      u[n] := postdirection [t] (f);
      v[n] := pnt[t] (f) + (u[n] zscaled (0,d));
    endfor
    v0{u0}
    for n = 1 upto segs:  ...v[n]{u[n]} endfor
  fi
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{parapath}
% Attempt to parallel one path with another at a distance \gbc{d}. The
% algorithm is as follows:
% \begin{enumerate}
% \item Create an array of paths that parallel each segment of \gbc{f}.
% \item Redo the array by adding a half circle at each end. The purpose
%       of the half circle is to force consecutive elements of the array
%       to intersect (I hope).
% \item Do something like \mfc{buildcycle}: where consecutive array
%       elements intersect, strip the the elements to the subpaths between
%       intersection points, and join them directly.
% \item If \gbc{f} is a cycle, process the joining at the endpoints in
%       the same way.
% \end{enumerate}
% We use \gbc{force_equal_ends} to force exact equality of endpoints so
% we can join segments with \mfc{\&}. The reason for prefering the
% \mfc{\&} join is to permit better performance of macros (such as this
% one) that take a path apart into segments. If we used `\mfc{..}' instead
% then the returned path would have a great many added segments, nearly
% all of which are nearly trivial. For similar reasons we drop trivial
% segments of \gbc{f}.
%    \begin{macrocode}
vardef parapath (expr d) expr f =
  if d = 0:
    f
  else:
    save a, g, h, p, q, s, t, u, v, w;
    path g[], h, p[], q[];
    numeric a, s, t;
    pair u, v, w, w[];
    s := emax (3, emin (segment_split, ceiling (max_points/5/length f)));
    p := 0;
    for i = 1 upto length f:
      h := subpath (i-1, i) of f;
      if not trivial h:
        q[incr p] := h;
        p[p] := parasegment (d, s, h);
      fi
    endfor
    if p = 0:
      f
    else:
      a := if d>0: - fi 180;
      h := p1;
      for i = 1 upto p-1:
        u := predirection 1 (q[i]);
        v := postdirection 0 (q[i+1]);
        w1 := pnt 1  (q[i])  - (u zscaled (0,d));
        w2 := pnt 0 (q[i+1]) - (v zscaled (0,d));
        w3 := pnt [infinity] (h);
        w4 := pnt 0 (p[i+1]);
        g0 := arcpps(w3, w1, a);
        g1 := h & g0;
        g2 := arcpps(w2, w4, a) & p[i+1];
        if (p[i] & g0) intersects reverse g2:
          s := length g2 - _Ytime;
          t := length h - length p[i] + _Xtime;
          g1  := subpath (0, t) of g1;
          g2 := subpath (s, length g2) of g2;
          force_equal_ends (g1, g2);
          h := g1 & g2;
        else:
          h := h .. p[i+1];
        fi
      endfor

      if cycle f:
        u := predirection 1 (q[p]);
        v := postdirection 0 (q[1]);
        w1 := pnt 1  (q[p])  - (u zscaled (0,d));
        w2 := pnt 0 (q[1]) - (v zscaled (0,d));
        w3 := pnt [infinity] (h);
        w4 := pnt 0 (p[1]);
        g3 := arcpps(w3, w1, a);
        g0 := arcpps(w2, w4, a);
        g1 := g0 & h & g3;
        g2 := g0 & p[1];
        if (p[p] & g3) intersects reverse g2:
          s := length g2 - _Ytime;
          t := length g0 + length h - length p[p] + _Xtime;
          g1 := subpath (s, t) of g1;
          force_equal_ends (g1, g1);
          h := g1 & cycle;
        else:
          h := h..cycle;
        fi
      fi
      h
    fi
  fi
enddef;

vardef turnangle@# (expr f) =
  anglefromto(predirection@# (f), postdirection@#(f))
enddef;

%    \end{macrocode}
%
% \section{Miscellaneous}\label{misc}
%
% \subsection{Implementation of \mfpic{}'s \cs{plotdata} command}%
% \label{plotdata}
%
% In \mfpic, the \cs{plotdata} command draws several curves with one
% command. The curves are drawn with changeable methods of rendering.
% There are three schemes. The first draws the curves with different dash
% patterns. Another scheme is to plot the curves with different symbols.
% Still another is to use different colors (\MP{} only).
%
% We implement the changing of patterns (symbols, colors) by defining
% arrays of such things and changing the index into the array. For
% example, when the user has selected dashes, the first curve is
% \gbc{gendashed} with the pattern \gbc{dashtype0}, the next with
% \gbc{dashtype1}, etc.
%
% \DescribeRoutine{setdatadashes}
% We have this method for users to select their own dash patterns. The
% \gbc{setdatadashes} command requires a list of suffixes previously
% defined by the \gbc{dashpat} command. Since a dash pattern need only
% be an array, we check if it is one and, if it is, we copy it to the
% next \gbc{dashtype[n]}. We actually copy it to a temporary array and
% make sure there are at least two patterns before we overwrite
% \gbc{dashtype}.
%
% \DescribeRoutine{getdashpat}
% We have removed this mod-ing operation from \TeX, where it is
% cumbersome, to \MF, where it is trivial.
%    \begin{macrocode}
def setdatadashes (text lst) =
  save __type;  __type := 0;
  forsuffixes _itm = lst:
    if knownnumericarray _itm :
      copyarray (_itm) (__type[__type]);
      next __type;
    else:  GBwarn "Improper dash pattern in setdatadashes.";
    fi
  endfor
  if __type > 1:
    save dashtype;  dashtype := __type;
    for _j = 0 upto dashtype - 1:
      copyarray (__type[_j]) (dashtype[_j]);
    endfor
  else:
    SetdataWarn "dashes";
  fi
enddef;
def getdashpat expr n =  dashtype[n mod dashtype]  enddef;

def SetdataWarn expr s =
  GBwarn "command setdata"& s &"() failed. Previous values retained.";
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{defaultdashes}
% These are the default dash patterns. Their setting is done by a macro
% so the user may easily restore them. The spaces are apparently larger
% than the dashes, but taking the thickness of the pen into account
% (\mfc{.5bp}) the dashes will appear about \mfc{.5bp} larger than stated
% and the spaces about \mfc{.5bp} smaller (unless the user inexplicably
% sets \mfc{linecap} to \mfc{butt}).
%    \begin{macrocode}
numeric Solid, Simpledash, Simpledot, Dotdash, Dotdashdot, Dotdashdash;
dashpat (Solid) (0);
dashpat (Simple_dash) (3bp, 4bp);
dashpat (Simple_dot) (0, 4bp);
dashpat (Dot_dash) (0, 4bp, 3bp, 4bp);
dashpat (Dot_dash_dot) (0, 4bp, 3bp, 4bp, 0, 4bp);
dashpat (Dot_dash_dash) (0, 4bp, 3bp, 4bp, 3bp, 4bp);

numeric dashtype, dashtype[], dashtype[][];
def defaultdashes =
  setdatadashes (Solid, Simple_dash, Simple_dot,
    Dot_dash, Dot_dash_dot, Dot_dash_dash);
enddef;
defaultdashes;

%    \end{macrocode}
%
% \DescribeRoutine{setdatasymbols}
% This can be used to define the sequence of point plotting styles for
% \mfpic's \cs{plotdata} command. It is quite similar to
% \gbc{setdatadashes} above and \gbc{setdatacolors} below.
%
% \DescribeRoutine{getsymbol}
% This is similar to \gbc{getdashpat}. In fact we could write a
% single macro to do both, but I think we get a more readable \mfpic{}
% output file if we have separate commands.
%    \begin{macrocode}
def setdatasymbols (text lst) =
  save __type;  path __type[];
  __type := 0;
  for _itm = lst:
    if (known _itm) and (path _itm):
      __type[__type] := _itm;
      next __type;
    else:
      GBwarn "Improper symbol in setdatasymbols().";
    fi
  endfor
  if __type > 1:
    save pointtype;  pointtype := __type;
    path pointtype[];
    for _j = 0 upto pointtype - 1:
      pointtype[_j] := __type[_j];
    endfor
  else:
    SetdataWarn "symbols";
  fi
enddef;
def getsymbol expr n :=  pointtype[n mod pointtype]  enddef;

%    \end{macrocode}
%
% Before we can set the default symbols we need to define some. They need
% to be paths. The ones below named with `\gbc{Solid}' are closed paths.
% Since the drawing commands that use them feed the path to \gbc{setdot},
% they end up filled if they are cyclic, merely drawn if not.
%
% All are intended to have roughly the area (when area makes sense) of a
% circle with diameter 1. The scaling factors are the square root of the
% ratios of the areas.
%
% Associated with each is a another path with the same basename and the
% suffix \gbc{clear} and a pair with the suffix \gbc{tip}. Moreover,
% they are (mostly) symmetric about the $y$-axis pointing (where that
% makes sense) in the direction \mfc{up}. The purpose of all this is so
% that they can be plugged into code for adding arrowheads/tails of
% different shapes.\VariableIndex{Triangle}\VariableIndex{Square}
% \VariableIndex{Circle}\VariableIndex{Diamond}\VariableIndex{Star}
% \VariableIndex{Plus}\VariableIndex{Cross}\VariableIndex{Asterisk}
% \VariableIndex{Crossbar}\VariableIndex{Leftbar}\VariableIndex{Rightbar}
% \VariableIndex{Righthook}\VariableIndex{Lefthook}
% \VariableIndex{SolidTriangle}\VariableIndex{SolidSquare}
% \VariableIndex{SolidCircle}\VariableIndex{SolidDiamond}
% \VariableIndex{SolidStar}
%    \begin{macrocode}
def DeclareGBSymbols (text S) =
  forsuffixes _itm = S:
    path _itm;
    path _itm.clear;
    pair _itm.tip;
  endfor
enddef;
DeclareGBSymbols(
  Triangle, Square, Circle, Diamond, Star, Plus, Cross,
  Asterisk, Crossbar, Leftbar, Rightbar, Righthook,
  Lefthook, SolidTriangle, SolidSquare, SolidCircle,
  SolidDiamond, SolidStar
);

vardef undo_cycle expr f =  subpath (0, length f) of f  enddef;

SolidTriangle := (up--(dir 210)--(dir -30)--cycle) scaled .78;
Triangle := undo_cycle SolidTriangle;

Triangle.clear := SolidTriangle.clear :=
  ((dir -30)--(cosd 30,1)--(cosd 210,1)--(dir 210)--up--cycle)
  scaled .78;

SolidSquare := (up--(-1,1)--(-1,-1)--(1,-1)--(1,1)--cycle) scaled .443;
Square := undo_cycle SolidSquare;

SolidCircle := fullcircle rotated 90;
Circle := undo_cycle SolidCircle;
Circle.clear := SolidCircle.clear :=
  halfcircle--(-.5,.5)--(.5,.5)--cycle;

SolidDiamond := (up--left--down--right--cycle)
                scaled .522 yscaled 1.44;
Diamond := undo_cycle SolidDiamond;
Diamond.clear := SolidDiamond.clear :=
  (right--(1,1)--(-1,1)--left--up--cycle) scaled .522 yscaled 1.44;

Plus := ((0,0)--up--down--(0,0)--left--right) scaled .65;
Plus.clear := (right--(1,1)--(-1,1)--(left)--cycle) scaled .65;

Cross := ((0,0)--(dir 45)--(dir -135)--(0,0)--(dir -45)--(dir 135))
         scaled .65;
Cross.clear := ((0,0)--(dir -45)--dir(45)--(dir 135)--(dir -135)--cycle)
               scaled .65;

Asterisk := ((0,0)--up--down--(0,0)--(dir 30)--(dir -150)
    --(0,0)--(dir -30)--(dir 150)) scaled .6;
Asterisk.clear := ((0,0)--(dir -30)--(cosd 30,1)--(cosd 150,1)
                  --(dir -150)--cycle) scaled .6;

Crossbar := ((0,0)--left--right) scaled .65;
Crossbar.clear := rect (right,(-1,.5)) scaled .65;

Leftbar := ((0,0)--left);
Rightbar := ((0,0)--right);
Leftbar.clear := rect((0,0),(-1,.5));
Rightbar.clear := rect((0,0),(1,.5));

Righthook := arcpps((0,0),(1,0),180);
Lefthook  := Righthook xscaled -1;
Righthook.clear := Righthook--cycle;
Lefthook.clear := Lefthook--cycle;

%    \end{macrocode}
%
% We do some computations to find the vertices of an n-pointed
% star. We assume that \mfc{A1} is \mfc{up} and the line from there
% to \mfc{A[1 + m]} determines one side of the top point of the star.
% We must have $2 \le {}$\gbc{m}${}\le{}$\gbc{n}${}-2$. The rest of the
% vertices are determined by symmetry. \gbc{Star} is made with \gbc{n=5}
% and \gbc{m=2}. We store the points in an array so we can use them for
% the \gbc{Star.clear} path.
%    \begin{macrocode}
vardef mkstar (expr n, m) (suffix A) =
  save ang;  ang := 360/n;
  A1 := up;  A3 := up rotated ang;
  A2 = (whatever)[A1, A1 rotated ( ang*m)];
  A2 = (whatever)[A3, A3 rotated (-ang*m)];
  for i = 4 upto 2n:
    A[i] := A[i-2] rotated ang;
  endfor
  A := 2n;
  mkpoly (true, A)
enddef;

save _A;  pair _A[];
SolidStar := mkstar (5, 2, _A) scaled .84;
Star := undo_cycle SolidStar;
Star.clear := polyline (true)
  (_A9, _A10, _A1, _A2, _A3, (xpart _A3, 1), (xpart _A9, 1)) scaled .84;
SolidStar.clear := Star.clear;

forsuffixes S =
 Triangle, Square, Circle, Diamond, Star, Plus, Cross,
 Asterisk, Crossbar, Leftbar, Rightbar, Righthook,
 Lefthook, SolidTriangle, SolidSquare, SolidCircle,
 SolidDiamond, SolidStar :
  S.tip := point 0 of S;
endfor

%    \end{macrocode}
%
% \DescribeRoutine{gcd}
% I thought I was going to use \gbc{gcd} for the \gbc{mkstar} routine
% above, but went another way. Still, it might have a future use. Once we
% have it,
% \DescribeRoutine{lcm}
% \gbc{lcm} is a snap. Since \gbc{gcd} always returns a positive result,
% \gbc{lcm} satisfies the usual rule for signs of products.
%    \begin{macrocode}
vardef gcd (expr n, m) =
  save a, b, r;
  a := emax (abs(m), abs(n));
  b := emin (abs(m), abs(n));
  if b > 0:
    forever:
      r := a mod b;
      exitif r < 1;
      a := b;  b := r;
    endfor
    b
  else:
    a
  fi
enddef;

vardef lcm (expr n, m) =
  n/gcd(n, m)*m
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{defaultsymbols}
% The command for restoring the default symbols.
%    \begin{macrocode}
numeric pointtype;  path pointtype[];
def defaultsymbols =
  setdatasymbols( Circle, Cross, SolidDiamond, Square, Plus,
    Triangle, SolidCircle, Star, SolidTriangle);
enddef;
defaultsymbols;

%    \end{macrocode}
%
% \DescribeRoutine{setdatacolors}
% Finally, for \MP, we do a similar pair of commands for setting
% the colors for the \cs{plotdata} command, and for
% \DescribeRoutine{getcolor}
% getting the next one. The odd indirection (\gbc{colortype[]} is an array
% of strings, the names of variables having color values) is because \MP{}
% now has three different data types for colors. Arrays must be all one
% type.
%
% \DescribeRoutine{defaultcolors}
% These default colors were tested on screen and on an inkjet printer.
% The adjustments away from pure colors is based on a compromise between
% those experiments.
%    \begin{macrocode}
%<*MP>
def setdatacolors (text lst) =
  setnumeric (__type) 0;
  % First, just count and store the known colors in the list
  for _itm = lst:
    if knowncolor _itm :
      if __type = 0 :  def _datacolors =  _itm  enddef;
      else:
        expandafter  def
        expandafter _datacolors
        expandafter =  _datacolors, _itm enddef;
      fi
      next __type;
    else:  GBwarn "Improper color in setdatacolors().";
    fi
  endfor
  if __type > 1:
    save colortype, _tmpstr;
    colortype := 0;
    % colortype[] is an array of strings:
    string colortype[], _tmpstr;
    for _itm = _datacolors: %
    % Each string is the name of some color variable
      _tmpstr := "colortype_" & GBromannumeral(colortype);
      setcolor (scantokens(_tmpstr)) _itm;
      colortype[colortype] := _tmpstr;
      next colortype;
    endfor
  else:
    SetdataWarn "colors";
  fi
enddef;
def getcolor expr n = (scantokens (colortype[n mod colortype])) enddef;

numeric colortype;  string colortype[];

setcolor (dRed)     (1, 0, 0);
setcolor (dBlue)    (.2,.2,1);
setcolor (dOrange)  (1,.34,0);
setcolor (dGreen)   (0,.80,0);
setcolor (dBlack)     cmykblack;
if has_cmyk :
  setcolor (dCyan)    cyan;
  setcolor (dMagenta) magenta;
  setcolor (dYellow)  yellow;
else: % rgb colors seem to be lighter than the cmyk equivalents.
  setcolor (dCyan)    cmyk(.85,0,0,.15);
  setcolor (dMagenta) cmyk(0,.85,0,.15);
  setcolor (dYellow)  cmyk(0,0,.85,.15);
fi
def defaultcolors =
  setdatacolors(dBlack, dRed, dBlue, dOrange,
    dGreen, dMagenta, dCyan, dYellow);
enddef;
defaultcolors;

%</MP>
%    \end{macrocode}
%
%
% \subsection{Pie Charts and Bar Charts}\label{charts}
%
% \DescribeRoutine{computepie}
% The \gbc{computepie} command calculates the wedges of a pie from the text
% parameter \gbc{data}. It should be a list of positive numerics, and the
% result will be one wedge for each datum, the area of the wedge being
% proportional to the corresponding datum. The wedge for each datum has
% its point at \gbc{cent} and the wedge for the first datum begins at
% angle \gbc{ang}. Each wedge is clockwise from the preceding one if
% \gbc{sign = -1}, otherwise anticlockwise. The radius of the pie is
% \gbc{rad}.
%
% \DescribeRoutine{piechart}
% This calls \gbc{computepie} to calculate the angles and store that in
% the array \gbc{_dat}, then \gbc{mkpiewedges} creates the actual user
% level arrays. This separation allows us (in a future enhancement) to
% easily handle named piecharts, so there can be more than one defined at
% a time.
%    \begin{macrocode}
def computepie (suffix dat) (expr sign, ang, cent, rad) (text data) =
begingroup
  save _tot, _max, _toobig;
  _max := 0;  dat := 0;
  for _val = data:
    dat[incr dat] := _val;
    _max := emax (_max, _val);
  endfor
  if dat=0:  GBwarn  "piechart attempted with empty list.";
    _toobig := 1;
  else:
    _toobig := infinity/dat;
  fi
  if _max > _toobig:
    for _idx = 1 upto dat:
      dat[_idx] := dat[_idx]/_toobig;
    endfor
  fi
  for _idx = 2 upto dat:
    dat[_idx] := dat[_idx - 1] + dat[_idx];
  endfor
  _tot := dat[dat];
  for _idx = dat downto 2:
    dat[_idx] := ang + sign*dat[_idx-1]/_tot*360;
  endfor
  dat1 := ang; dat[dat + 1] := ang + 360sign;
endgroup
enddef;

def piechart (expr sign, ang, cent, rad) (text data) =
  save _dat;
  computepie (_dat) (sign, ang, cent, rad) (data);
  mkpiewedges (_dat, cent, rad);
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{mkpiewedges}
% The wedges (closed sectors) are stored in the array \gbc{piewedge[\,]}
% with the numeric \gbc{piewedge} holding the number of wedges. The center
% is saved in \gbc{piecenter}, the directions of the wedges (the bisecting
% rays) are stored in \gbc{piedirection[\,]}, the starting angles of the
% wedges in \gbc{pieangle[\,]}
%    \begin{macrocode}
def mkpiewedges (suffix dat) (expr cent, rad) =
  numeric piewedge, piedirection, pieangle, pieangle[];
  pair piecenter, piedirection[];
  path piewedge[];
  piecenter := cent;
  piedirection := pieangle := piewedge := dat;
  for _idx = 1 upto dat:
    pieangle[_idx] := dat[_idx];
    piewedge[_idx] := sector (piecenter, rad, dat[_idx], dat[_idx+1]);
    piedirection[_idx] := dir(0.5[ dat[_idx], dat[_idx+1] ]);
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{namedpiechart}
% This is a future enhancement. It takes a suffix, the name of the chart
% which will be the base name of the various arrays and key values, and
% will be a numeric equal to the number of wedges. The first part is
% identical to that of \gbc{piechart} above. Note that some elements
% of the chart are not (yet) directly accessible. Those needed by mfpic
% are: the paths, the directions and the center. The rest would be easy
% to add.
%    \begin{macrocode}
def namedpiechart (suffix nm) (expr sign, ang, cent, rad) (text data) =
  save _dat;
  computepie (_dat) (sign, ang, cent, rad) (data);
  setnumeric (nm) _dat;
  pair nm.center, nm.direction[];
  path nm.wedge[];
  nm.center := cent;
  for _idx = 1 upto _dat:
    nm.wedge[_idx] := sector (cent, rad, _dat[_idx], _dat[_idx+1]);
    nm.direction[_idx] := dir(0.5[ _dat[_idx], _dat[_idx+1] ]);
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{barchart}
% I was told that there are better ways (than piecharts) to represent
% quantitative data. Perhaps bar charts are better. \gbc{barchart}
% calculates the bars from the text parameter, \gbc{data}. These bars are
% vertical if \gbc{vert} is true, otherwise horizontal.
%
% \gbc{firstbar} is the location (on the appropriate axis) of the start of
% the first bar. \gbc{sep} is the separation between bar centers. \gbc{r}
% is the ratio of the width of the bars to their separation.
%
% After the calculations, the array of paths \gbc{chartbar[\,]} holds the
% rectangles, \gbc{barend[\,]} holds their rightmost or topmost
% coordinates (which is just the items in \gbc{data} or their yparts),
% \gbc{barbegin[\,]} holds their leftmost or bottommost coordinates (either
% 0 or the xparts of the data), \gbc{barstart[\,]} holds the appropriate
% coordinate of the leading edge of the bar, and \gbc{barwd = r*sep}.
%
% If the data are pair data, this command uses the xpart as the beginning
% of the bar and the ypart as the end. Thus Gantt diagrams can be
% created. We keep \gbc{barlength} for backward compatibility (formerly
% all data had to be numeric and bars went from 0 to \gbc{barlength[\,]}).
% \gbc{barlength[\,]} was made available to help place some label or symbol
% at the end of a bar and existing code might break if we omit it.
%    \begin{macrocode}
def barchart (expr firstbar, sep, r, vert)(text data) =
  numeric barbegin, barbegin[],
          barend, barend[],
          barlength, barlength[],
          barstart, barstart[],
          chartbar, barwd;
  path chartbar[];
  chartbar := 0;  barwd := r*sep;
  for _itm = data:
    barend[incr chartbar] := if pair _itm:  ypart _itm  else:  _itm  fi;
    barbegin[chartbar]    := if pair _itm:  xpart _itm  else:  0  fi;
  endfor
  barbegin := barend := barlength := barstart := chartbar;
  for _nn = 1 upto chartbar:
    barstart[_nn]  := firstbar + sep*(_nn-1);
    barlength[_nn] := barend[_nn];
    chartbar[_nn]  :=  rect ((barbegin[_nn], 0), ( barend[_nn], barwd))
          shifted (0, barstart[_nn]) if vert:  xyswap  fi;
  endfor
enddef;

%    \end{macrocode}
%
% \DescribeRoutine{namedbarchart}
% This is a future enhancement. It takes a suffix, the name of the chart,
% which will be the base name of the path array. Note that the various key
% elements (ends of the bar, etc.) are not (yet) directly accessible as in
% the above version, but they would be easy to add.
%    \begin{macrocode}
def namedbarchart (suffix nm) (expr first, sep, r, vert) (text data) =
  save nm;
begingroup
  save _bb, _ee, _ww;
  path nm.bar[];
  nm := 0;  _ww := r*sep;
  for _itm = data:
    _ee := if pair _itm:  ypart _itm  else:  _itm  fi;
    _bb := if pair _itm:  xpart _itm  else:  0  fi;
    nm.bar[incr nm] :=  rect ((_bb, 0), ( _ee, _ww) )
          shifted (0, first + sep*(nm-1)) if vert:  xyswap fi;
  endfor
endgroup
enddef;

%    \end{macrocode}
%
%^^A Overlays - taken from MFbook, p 295.  (Bruce Leban)
%
% \subsection{Overlays}\label{overlays}
%
% This final code predates me. When I inherited \mfpic{} it contained no
% use of \gbc{keepit}. For the \MP{} version I just tried to make sure
% everything was defined in \MP{} or \file{plain.mp} and otherwise left it
% alone. One might presumably issue \gbc{keepit} periodically, clearing
% \mfc{currentpicture} after adding it onto \gbc{totalpicture}. This would
% save some memory because manipulating \mfc{currentpicture} often
% requires a couple of copies to be around simultaneously. Right now we
% treat currentpicture as the place to save things and manipulate local
% picture variables.
%    \begin{macrocode}
picture totalpicture;
boolean totalnull, currentnull;
def clearit =
  currentpicture := totalpicture := nullpicture;
  currentnull := totalnull := true;
enddef;

def keepit =
  addto totalpicture also currentpicture;
%<MF>  mono (totalpicture);
  currentpicture := nullpicture;
  totalnull := totalnull or currentnull;
  currentnull := true;
enddef;

def addto_currentpicture =
  currentnull := false;
  addto currentpicture
enddef;

def mergeit (text do) =
  if totalnull:
    do currentpicture
  elseif currentnull:
    do totalpicture
  else:
    begingroup
      save _v_;  picture _v_;
      _v_ := currentpicture;
      addto _v_ also totalpicture;
      do _v_
    endgroup
  fi
enddef;

%    \end{macrocode}
% This implements \mfpic{}'s \cs{stopshipping} and \cs{resumeshipping}
% commands. It used to be that those commands wrote a redefinition of
% \gbc{shipit}, now we write a boolean assignment and the \cs{shipit}
% command tests it.
%    \begin{macrocode}
boolean noship;  noship := false;
def shipit =  if noship:  else:  mergeit (shipout)  fi  enddef;

%<*MF>
def showit_ =  mergeit (show_)  enddef;
def show_ suffix v =  display v inwindow currentwindow  enddef;

%</MF>
%    \end{macrocode}
%
% Here we initialize \gbc{gcode} (which current versions of mfpic do not
% use) for hacked \mfpic{} files that require it. In \MP{} we load the
% color definitions. And thats all.
%    \begin{macrocode}
numeric gcode;  gcode := 0;

%<MF>%  end grafbase.mf
%<MP>input dvipsnam.mp;
%<MP>%  end grafbase.mp
%</MF|MP>
%    \end{macrocode}
%
% \subsection{Dvips names for colors}\label{dvipsnam}
%
% In order to make \file{dvipsnam.mp} useful outside grafbase, we give
% here a definition for \gbc{cmyk} when \gbc{grafbaseversion} is unknown.
%    \begin{macrocode}
%<*dvips>
if unknown grafbaseversion:
  if unknown mpversion:
    let cmykcolor=color;
    vardef cmyk (expr c, m, y, k) =
      (max(1-c-k,0), max(1-m-k,0), max(1-y-k,0))
    enddef;
  else:
    vardef cmyk (expr c, m, y, k) =  (c, m, y, k)  enddef;
  fi
fi

%    \end{macrocode}
%
% The following code was borrowed from the the standard \LaTeX{} graphics
% package (\file{dvipsname.def} by David Carlisle and Sebastian Rahtz). In
% fact it was mostly generated automatically by some editor macros that
% replaced \prog{graphics} package code with the \grafbase{} code.
%
%^^A This file may be distributed under the terms of the LaTeX Project Public
%^^A License, as described in \file{lppl.txt} in the base LaTeX
%^^A distribution, either version 1.0 or, at your option, any later version.
%
% Declare all the dvips color names to be color variables, and define
% them as in \file{dvipsnam.def}:
%    \begin{macrocode}
cmykcolor Apricot, Aquamarine, Bittersweet, Black, Blue, BlueGreen,
BlueViolet, BrickRed, Brown, BurntOrange, CadetBlue, CarnationPink,
Cerulean, CornflowerBlue, Cyan, Dandelion, DarkOrchid, Emerald,
ForestGreen, Fuchsia, Goldenrod, Gray, Green, GreenYellow, JungleGreen,
Lavender, LimeGreen, Magenta, Mahogany, Maroon, Melon, MidnightBlue,
Mulberry, NavyBlue, OliveGreen, Orange, OrangeRed, Orchid, Peach,
Periwinkle, PineGreen, Plum, ProcessBlue, Purple, RawSienna, Red,
RedOrange, RedViolet, Rhodamine, RoyalBlue, RoyalPurple, RubineRed,
Salmon, SeaGreen, Sepia, SkyBlue, SpringGreen, Tan, TealBlue, Thistle,
Turquoise, Violet, VioletRed, White, WildStrawberry, Yellow,
YellowGreen, YellowOrange;

Apricot        := cmyk(0,0.32,0.52,0);
Aquamarine     := cmyk(0.82,0,0.30,0);
Bittersweet    := cmyk(0,0.75,1,0.24);
Black          := cmyk(0,0,0,1);
Blue           := cmyk(1,1,0,0);
BlueGreen      := cmyk(0.85,0,0.33,0);
BlueViolet     := cmyk(0.86,0.91,0,0.04);
BrickRed       := cmyk(0,0.89,0.94,0.28);
Brown          := cmyk(0,0.81,1,0.60);
BurntOrange    := cmyk(0,0.51,1,0);
CadetBlue      := cmyk(0.62,0.57,0.23,0);
CarnationPink  := cmyk(0,0.63,0,0);
Cerulean       := cmyk(0.94,0.11,0,0);
CornflowerBlue := cmyk(0.65,0.13,0,0);
Cyan           := cmyk(1,0,0,0);
Dandelion      := cmyk(0,0.29,0.84,0);
DarkOrchid     := cmyk(0.40,0.80,0.20,0);
Emerald        := cmyk(1,0,0.50,0);
ForestGreen    := cmyk(0.91,0,0.88,0.12);
Fuchsia        := cmyk(0.47,0.91,0,0.08);
Goldenrod      := cmyk(0,0.10,0.84,0);
Gray           := cmyk(0,0,0,0.50);
Green          := cmyk(1,0,1,0);
GreenYellow    := cmyk(0.15,0,0.69,0);
JungleGreen    := cmyk(0.99,0,0.52,0);
Lavender       := cmyk(0,0.48,0,0);
LimeGreen      := cmyk(0.50,0,1,0);
Magenta        := cmyk(0,1,0,0);
Mahogany       := cmyk(0,0.85,0.87,0.35);
Maroon         := cmyk(0,0.87,0.68,0.32);
Melon          := cmyk(0,0.46,0.50,0);
MidnightBlue   := cmyk(0.98,0.13,0,0.43);
Mulberry       := cmyk(0.34,0.90,0,0.02);
NavyBlue       := cmyk(0.94,0.54,0,0);
OliveGreen     := cmyk(0.64,0,0.95,0.40);
Orange         := cmyk(0,0.61,0.87,0);
OrangeRed      := cmyk(0,1,0.50,0);
Orchid         := cmyk(0.32,0.64,0,0);
Peach          := cmyk(0,0.50,0.70,0);
Periwinkle     := cmyk(0.57,0.55,0,0);
PineGreen      := cmyk(0.92,0,0.59,0.25);
Plum           := cmyk(0.50,1,0,0);
ProcessBlue    := cmyk(0.96,0,0,0);
Purple         := cmyk(0.45,0.86,0,0);
RawSienna      := cmyk(0,0.72,1,0.45);
Red            := cmyk(0,1,1,0);
RedOrange      := cmyk(0,0.77,0.87,0);
RedViolet      := cmyk(0.07,0.90,0,0.34);
Rhodamine      := cmyk(0,0.82,0,0);
RoyalBlue      := cmyk(1,0.50,0,0);
RoyalPurple    := cmyk(0.75,0.90,0,0);
RubineRed      := cmyk(0,1,0.13,0);
Salmon         := cmyk(0,0.53,0.38,0);
SeaGreen       := cmyk(0.69,0,0.50,0);
Sepia          := cmyk(0,0.83,1,0.70);
SkyBlue        := cmyk(0.62,0,0.12,0);
SpringGreen    := cmyk(0.26,0,0.76,0);
Tan            := cmyk(0.14,0.42,0.56,0);
TealBlue       := cmyk(0.86,0,0.34,0.02);
Thistle        := cmyk(0.12,0.59,0,0);
Turquoise      := cmyk(0.85,0,0.20,0);
Violet         := cmyk(0.79,0.88,0,0);
VioletRed      := cmyk(0,0.81,0,0);
White          := cmyk(0,0,0,0);
WildStrawberry := cmyk(0,0.96,0.39,0);
Yellow         := cmyk(0,0,1,0);
YellowGreen    := cmyk(0.44,0,0.74,0);
YellowOrange   := cmyk(0,0.42,1,0);

% End of file `dvipsnam.mp'.
%</dvips>
%    \end{macrocode}
% \clearpage
%\Finale