%D \module %D [ file=supp-pdf, %D version=2004.12.16, %D title=\CONTEXT\ Support Macros, %D subtitle=\METAPOST\ to \PDF\ conversion, %D author=Hans Hagen \& others (see text), %D date=\currentdate, %D copyright=\PRAGMA] %C %C This module is part of the \CONTEXT\ macro||package and is %C therefore copyrighted by \PRAGMA. See mreadme.pdf for %C details. %D These macros are written as generic as possible. Some %D general support macro's are loaded from a small module %D especially made for non \CONTEXT\ use. In this module I %D use a matrix transformation macro written by Tanmoy %D Bhattacharya. Thanks to extensive testing by Sebastian %D Ratz I was able to complete this module within reasonable %D time. This module has support for \METAPOST\ extensions %D built in. %D %D Daniel H. Luecking came up with a better (more precise) %D transformation method. You can recognize his comment by %D his initials. (We keep the old code around because it's a %D nice illustration on how a module like this evolves.) % Beware, we cannot use \zeropoint here since it may be % defined in the range \dimen 0 - 20 which we happen to use % as scratch registers; inside context we may consider % using dedicated registers. % fails: ( ) vect10 9.96265 fshow % \040\040\040\040 % % some optimizations are possible, but we need to be generic %D First we take care of non||\CONTEXT\ use (newlines needed, %D expandafter not -). \ifx \undefined \writestatus \input supp-mis.tex \relax \fi \ifx \undefined \convertMPtoPDF \else \expandafter \endinput \fi \ifx \undefined \StartLatexHack \else \StartLatexHack \fi %D This module handles some \PDF\ conversion and insertions %D topics. By default, the macros use the \PDFTEX\ primitive %D \type{\pdfliteral} when available. Since \PDFTEX\ is now the %D default engine for \TEX\ distributions, we need a more complex %D test. \writestatus{loading}{Context Support Macros / PDF (2004.03.26)} \unprotect \ifx\PDFcode\undefined \ifx\pdfliteral\undefined \def\PDFcode#1{\special{PDF: #1}} \else\ifx\pdfoutput\undefined \def\PDFcode#1{\special{PDF: #1}} \else\ifcase\pdfoutput \def\PDFcode#1{\special{PDF: #1}} \else % pdftex as well as in pdf mode \let\PDFcode\pdfliteral \fi\fi\fi \else % we probably use context \fi %D First we define a handy constant: \bgroup \catcode`\%=\@@other \xdef\letterpercent{\string%} \egroup %D \macros %D {pdfimage,pdfimages,pdfclippedimage} %D %D Starting with pdftex version 14, images are included more %D natural to the form embedding. This enables alternative %D images to be embedded. %D %D \starttyping %D \pdfimage {file} %D \pdfimages {high res file} {low res file} %D \stoptyping %D %D The first one replaces the pre||version||14 original, %D while the latter provides alternative images. %D %D The next macro is dedicated to Maarten Gelderman, who %D needed to paste prepared \PDF\ pages into conference %D proceedings. %D %D \starttyping %D \pdfclippedimage {file} {l} {r} {t} {b} %D \stoptyping % \ifx\pdftexversion\undefined % no \m!systems for the sake of latex % \writestatus{systems}{Why not use pdf(e)TeX binaries?} % \protect % \expandafter\endinput % \fi \ifx\pdftexversion\undefined \else \ifnum\pdftexversion>13 \def\pdfimage#1#% {\dopdfimage{#1}} \def\dopdfimage#1#2% {\immediate\pdfximage#1{#2}% \pdfrefximage\pdflastximage} \def\pdfimages#1#% {\dopdfimages{#1}} \def\dopdfimages#1#2#3% {\immediate\pdfximage#1{#2}% \immediate\pdfobj {[ << /Image \the\pdflastximage\space0 R /DefaultForPrinting true >> ]}% \immediate\pdfximage#1 attr {/Alternates \the\pdflastobj\space0 R}{#3}% \pdfrefximage\pdflastximage} \def\pdfclippedimage#1#% specs {file}{left}{right}{top}{bottom} {\dopdfclippedimage{#1}} \def\dopdfclippedimage#1#2#3#4#5#6% {\bgroup \pdfximage#1{#2}% \setbox\scratchbox\hbox {\pdfrefximage\pdflastximage}% \hsize\wd\scratchbox \advance\hsize -#3 \advance\hsize -#4 \vsize\ht\scratchbox \advance\vsize -#5 \advance\vsize -#6 \setbox\scratchbox\vbox to \vsize {\vskip-#5\hbox to \hsize{\hskip-#3\box\scratchbox\hss}}% \pdfxform\scratchbox \pdfrefxform\pdflastxform \egroup} \fi \fi %D If you want to save a few hash entries, you may prefer the %D less readable alternatives, like: %D %D \starttyping %D \def\pdfimage#1#% This one is less readable but needs no additional %D {\bgroup % hash entry for the second stage macro. %D \def\pdfimage##1% %D {\immediate\pdfximage##1{#2}% %D \pdfrefximage\pdflastximage\egroup}} %D \stoptyping %D For old times sake we provide a few aliases. These will be %D removed some day. \ifx\pdftexversion\undefined \else \ifnum\pdftexversion>13 \let\pdfform =\pdfxform \let\pdflastform=\pdflastxform \let\pdfrefform =\pdfrefxform \fi \fi %D \macros %D {convertPDFtoPDF} %D %D \PDFTEX\ supports verbatim inclusion of \PDF\ code. The %D following macro takes care of inserting externally defined %D illustrations in \PDF\ format. According to a suggestion %D Tanmoy Bhattacharya posted to the \PDFTEX\ mailing list, we %D first skip lines until \type{stream} is reached and then %D copy lines until \type{endstream} is encountered. This %D scheme only works with vectorized graphics in which no %D indirect references to objects are used. Bitmaps also don't %D work. Interpreting their specifications is beyond the %D current implementation. %D %D \starttyping %D \convertPDFtoPDF %D {filename} %D {x scale} {y scale} %D {x offset } {y offset} %D {width} {height} %D \stoptyping %D %D When the scales are set to~1, the last last four values %D are the same as the bounding box, e.g. %D %D \starttyping %D \convertPDFtoPDF{mp-pra-1.pdf} {1} {1}{-1bp}{-1bp}{398bp}{398bp} %D \convertPDFtoPDF{mp-pra-1.pdf}{.5}{.5} {0bp} {0bp}{199bp}{199bp} %D \stoptyping %D %D Keep in mind, that this kind of copying only works for %D pure and valid pdf code (without fonts). %D The scanning and copying is straightforward and quite fast. %D To speed up things we use two constants. \def\@@PDFstream@@ {stream} \def\@@PDFendstream@@ {endstream} %D \macros %D {PDFmediaboxprefered} %D %D If needed, the macros can scan for the mediabox that %D specifies the dimensions and offsets of the graphic. When %D we say: %D %D \starttyping %D \PDFmediaboxpreferedtrue %D \stoptyping %D %D the mediabox present in the file superseded the user %D specified, already scaled and calculated offset and %D dimensions. Beware: the user supplied values are not the %D bounding box ones! %D This code has become obsolete. % \newif\ifPDFmediaboxprefered \def\setPDFboundingbox#1#2#3#4#5#6% {\dimen0=#1\dimen0=#5\dimen0 \ScaledPointsToBigPoints{\number\dimen0}\PDFxoffset \dimen0=#3\dimen0=#5\dimen0 \xdef\PDFwidth{\the\dimen0}% \dimen0=#2\dimen0=#6\dimen0 \ScaledPointsToBigPoints{\number\dimen0}\PDFyoffset \dimen0=#4\dimen0=#6\dimen0 \xdef\PDFheight{\the\dimen0}% \global\let\PDFxoffset\PDFxoffset \global\let\PDFyoffset\PDFyoffset} \def\setPDFmediabox#1[#2 #3 #4 #5]#6\done {\dimen2=#2bp\dimen2=-\dimen2 % \dimen2=-#2bp also works since tex handles -- \dimen4=#3bp\dimen4=-\dimen4 % \dimen4=-#3bp also works since tex handles -- \dimen6=#4bp\advance\dimen6 \dimen2 \dimen8=#5bp\advance\dimen8 \dimen4 \setPDFboundingbox{\dimen2}{\dimen4}{\dimen6}{\dimen8}\PDFxscale\PDFyscale} % \def\checkPDFmediabox#1/MediaBox#2#3\done % {\ifx#2\relax \else % \message{mediabox}% % \setPDFmediabox#2#3\done % \fi} %D We use the general macro \type{\doprocessfile} and feed this %D with a line handling macro that changes it's behavior when %D the stream operators are encountered. % \def\handlePDFline % {\ifx\@@PDFstream@@\fileline % \let\doprocessPDFline\copyPDFobject % \startPDFtoPDF % \else\ifPDFmediaboxprefered % \expandafter\checkPDFmediabox\fileline/MediaBox\relax\done % \fi\fi} % \def\copyPDFobject % {\ifx\@@PDFendstream@@\fileline % \ifPDFmediaboxprefered % \let\doprocessPDFline\findPDFmediabox % \else % \let\doprocessPDFline\relax % \fi % \else % \advance\scratchcounter 1 % \PDFcode{\fileline}% % \fi} % \def\findPDFmediabox % {\expandafter\checkPDFmediabox\fileline/MediaBox\relax\done} %D The main conversion macro wraps the \PDF\ codes in a box %D that is output as an object. The graphics are embedded %D in~\type{q} and~\type{Q} and are scaled and positioned using %D one transform call (\type{cm}). This saves some additional %D scaling. %D \starttyping %D \def\startPDFtoPDF% %D {\setbox0=\vbox\bgroup %D \message{[PDF to PDF \PDFfilename}% %D \forgetall %D \scratchcounter=0 %D \let\stopPDFtoPDF=\dostopPDFtoPDF} %D %D \def\dostopPDFtoPDF% %D {\ifnum\scratchcounter<0 \scratchcounter=1 \fi %D \message{(\the\scratchcounter\space lines)]}% %D \egroup %D \wd0=\PDFwidth %D \vbox to \PDFheight %D {\forgetall %D \vfill %D \PDFcode{q}% %D \PDFcode{1 0 0 1 \PDFxoffset\space \PDFyoffset\space cm}% %D \PDFcode{\PDFxscale\space 0 0 \PDFyscale\space 0 0 cm}% %D \box0 %D \PDFcode{Q}}} %D %D \def\stopPDFtoPDF% %D {\message{[PDF to PDF \PDFfilename\space not found]}} %D %D \def\convertPDFtoPDF#1#2#3#4#5#6#7% %D {\bgroup %D \def\PDFfilename{#1}% %D \def\PDFxscale {#2}% %D \def\PDFyscale {#3}% %D \setPDFboundingbox{#4}{#5}{#6}{#7}{1}{1}% %D \uncatcodespecials %D \endlinechar=-1 %D \let\doprocessPDFline=\handlePDFline %D \doprocessfile\scratchread\PDFfilename\doprocessPDFline %D \stopPDFtoPDF %D \egroup} \def\convertPDFtoPDF#1#2#3#4#5#6#7% {\message{[PDF to PDF use \string\PDFcode instead]}% \vbox{use the direct method instead}} %D \macros %D {dogetPDFmediabox} %D %D The next macro can be used to find the mediabox of a \PDF\ %D illustration. %D %D \starttyping %D \dogetPDFmediabox %D {filename} %D {new dimen}{new dimen}{new dimen}{new dimen} %D \stoptyping %D %D Beware of dimen clashes: this macro uses the 5~default %D scratch registers! When no file or mediabox is found, the %D dimensions are zeroed. \def\dogetPDFmediabox#1#2#3#4#5% {\bgroup \def\PDFxscale{1}% \def\PDFyscale{1}% \uncatcodespecials \endlinechar\minusone \def\checkPDFtypepage##1/Type /Page##2##3\done% {\ifx##2\relax \else\if##2s% accept /Page and /Pages \let\doprocessPDFline\findPDFmediabox \else \let\doprocessPDFline\findPDFmediabox \fi\fi}% \def\findPDFtypepage {\expandafter\checkPDFtypepage\fileline/Type /Page\relax\done}% \def\checkPDFmediabox##1/MediaBox##2##3\done% {\ifx##2\relax \else \setPDFmediabox##2##3\done \fileprocessedtrue \fi}% \def\findPDFmediabox {\expandafter\checkPDFmediabox\fileline/MediaBox\relax\done}% \let\doprocessPDFline\findPDFtypepage \doprocessfile\scratchread{#1}\doprocessPDFline \egroup \ifx\PDFxoffset\undefined #2\zeropoint #3\zeropoint #4\zeropoint #5\zeropoint \else #2=\PDFxoffset\onebasepoint #3=\PDFyoffset\onebasepoint #4=\PDFwidth #5=\PDFheight \fi} %D \macros %D {convertMPtoPDF} %D %D The next set of macros implements \METAPOST\ to \PDF\ %D conversion. Because we want to test as fast as possible, we %D first define the \POSTSCRIPT\ operators that \METAPOST\ %D uses. We don't define irrelevant ones, because these are %D skipped anyway. %D The converter can be made a bit faster by replacing the %D two test macros (the ones with the many \type {\if's}) by %D a call to named branch macros (something \typ {\getvalue %D {xPSmoveto}}. For everyday documents with relatively %D small graphics the gain in speed can be neglected. \def \PScurveto {curveto} \def \PSlineto {lineto} \def \PSmoveto {moveto} \def \PSshowpage {showpage} \def \PSnewpath {newpath} \def \PSfshow {fshow} \def \PSclosepath {closepath} \def \PSfill {fill} \def \PSstroke {stroke} \def \PSclip {clip} \def \PSrlineto {rlineto} \def \PSsetlinejoin {setlinejoin} \def \PSsetlinecap {setlinecap} \def \PSsetmiterlimit {setmiterlimit} \def \PSsetgray {setgray} \def \PSsetrgbcolor {setrgbcolor} \def \PSsetcmykcolor {setcmykcolor} \def \PSsetdash {setdash} \def \PSgsave {gsave} \def \PSgrestore {grestore} \def \PStranslate {translate} \def \PSscale {scale} \def \PSconcat {concat} \def \PSdtransform {dtransform} \def \PSsetlinewidth {setlinewidth} \def \PSpop {pop} \def \PSnfont {nfont} % was needed for TUG98 proceedings \def \PSspecial {special} % extensions to MetaPost %D A previous version set \type {%} to ignore, which %D simplified the following definitions. At the start of %D conversion the percent character was made active again. %D Because the whole graphic is one paragraph (there are no %D empty lines) this does not give the desired effect. This %D went unnoticed untill Scott Pakin sent me a test file %D percent characters in a string. So, from now on we have %D to prefix the following strings with percentages. \edef \PSBoundingBox {\letterpercent\letterpercent BoundingBox:} \edef \PSHiResBoundingBox {\letterpercent\letterpercent HiResBoundingBox:} \edef \PSExactBoundingBox {\letterpercent\letterpercent ExactBoundingBox:} \edef \PSMetaPostSpecial {\letterpercent\letterpercent MetaPostSpecial:} \edef \PSMetaPostSpecials {\letterpercent\letterpercent MetaPostSpecials:} \edef \PSPage {\letterpercent\letterpercent Page:} %D By the way, the \type {setcmykcolor} operator is not %D output by \METAPOST\ but can result from converting the %D \cap{RGB} color specifications, as implemented in %D \type{supp-mps}. %D In \POSTSCRIPT\ arguments precede the operators. Due to the %D fact that in some translations we need access to those %D arguments, and also because sometimes we have to skip them, %D we stack them up. The stack is one||dimensional for non path %D operators and two||dimensional for operators inside a path. %D This is because we have to save the whole path for %D (optional) postprocessing. Values are pushed onto the stack %D by: %D %D \starttyping %D \setMPargument {value} %D \stoptyping %D %D They can be retrieved by the short named macros: %D %D \starttyping %D \gMPa {number} %D \gMPs {number} %D \stoptyping %D %D When scanning a path specification, we also save the %D operator, using %D %D \starttyping %D \setMPkeyword {n} %D \stoptyping %D %D The path drawing operators are coded for speed: \type{clip}, %D \type{stroke}, \type{fill} and \type{fillstroke} become %D 1, 2, 3 and~4. %D %D When processing the path this code can be retrieved %D using %D %D \starttyping %D \getMPkeyword % {n} %D \stoptyping %D %D When setting an argument, the exact position on the stack %D depends on the current value of the \COUNTERS\ %D \type{\nofMPsegments} and \type{\nofMParguments}. \newcount\nofMPsegments \newcount\nofMParguments %D These variables hold the coordinates. The argument part of %D the stack is reset by: %D %D \starttyping %D \resetMPstack %D \stoptyping %D %D We use the prefix \type{@@MP} to keep the stack from %D conflicting with existing macros. To speed up things a bit %D more, we use the constant \type{\@@MP}. \def\@@MP{@@MP} \def\setMPargument% #1% {\advance\nofMParguments \plusone \expandafter\def \csname\@@MP\the\nofMPsegments\the\nofMParguments\endcsname} % {#1} \def\letMPargument {\advance\nofMParguments \plusone \expandafter\let \csname\@@MP\the\nofMPsegments\the\nofMParguments\endcsname} \def\setMPsequence#1 % {\advance\nofMParguments \plusone \expandafter\def \csname\@@MP\the\nofMPsegments\the\nofMParguments\endcsname{#1}% \handleMPsequence} \def\gMPa#1% {\csname\@@MP0\number#1\endcsname} \def\gMPs#1% {\csname\@@MP\the\nofMPsegments\number#1\endcsname} \def\dogMPa#1% {\@EAEAEA\do\csname\@@MP0\number#1\endcsname} \def\setMPkeyword#1 % {\expandafter\def\csname\@@MP\the\nofMPsegments0\endcsname{#1}% \advance\nofMPsegments \plusone \nofMParguments\zerocount} \def\getMPkeyword% #1% {\csname\@@MP\the\nofMPsegments0\endcsname} % {\csname\@@MP#10\endcsname} %D When we reset the stack, we can assume that all further %D comment is to be ignored and handled in strings. %D By redefining the reset macro after the first call, we %D save some run time. Only use this macro after all %D comments are processed and use the simple alternative %D when dealing with comments. \def\doresetMPstack {\nofMParguments\zerocount} \def\resetMPstack {\let\handleMPgraphic\handleMPendgraphic \let\resetMPstack\doresetMPstack \resetMPstack} %D The arguments are saved with the preceding command %D \type{\do}. By default this command expands to nothing, but %D when we deal with strings it's used to strip off the %D \type{(} and \type{)}. %D %D Strings are kind of tricky, because characters can be %D passed verbatim \type{(hello)}, by octal number %D \type{(\005)} or as command \type{(\()}. We therefore %D cannot simply ignore \type{(} and \type{)}, the way we do %D with \type{[} and \type{]}. Another complication is that %D strings may contain characters that normally have a %D special meaning in \TEX, like \type{$} and \type{{}}. %D %D A previous solution made \type{\} an active character and %D let it look ahead for a number or characters. We had to %D abandon this scheme because of the need for verbatim %D support. The next solution involved some \CATCODE\ %D trickery but works well. \def\octalMPcharacter#1#2#3% {\char'#1#2#3\relax} \bgroup \catcode`\|=\@@comment \catcode`\%=\@@active \catcode`\[=\@@active \catcode`\]=\@@active \catcode`\{=\@@active \catcode`\}=\@@active \catcode`B=\@@begingroup \catcode`E=\@@endgroup \gdef\ignoreMPspecials| B\let%\letterpercent| \def[BE| \def]BE| \def{BE| \def}BEE \gdef\obeyMPspecials| B\def%B\char 37\relax E| \def[B\char 91\relax E| \def]B\char 93\relax E| \def{B\char123\relax E| \def}B\char125\relax EE \gdef\setMPspecials| B\setnaturalcatcodes \catcode`\\=\@@escape \catcode`\%=\@@active \catcode`\[=\@@active \catcode`\]=\@@active \catcode`\{=\@@active \catcode`\}=\@@active \lccode`\-=0 | latex sets this to `\- \lccode`\%=`\% | otherwise it's seen as a number \def\(B\char40\relax E| \def\)B\char41\relax E| \def\\B\char92\relax E| \def\0B\octalMPcharacter0E| \def\1B\octalMPcharacter1E| \def\2B\octalMPcharacter2E| \def\3B\octalMPcharacter3E| \def\4B\octalMPcharacter4E| \def\5B\octalMPcharacter5E| \def\6B\octalMPcharacter6E| \def\7B\octalMPcharacter7E| \def\8B\octalMPcharacter8E| \def\9B\octalMPcharacter9EE \egroup %D We use the comment symbol as a sort of trigger. Beware! %D The whole graphic is seen as on eparagraph, which means %D that we cannot change the catcodes in between. \bgroup \catcode`\%=\@@active \gdef\startMPscanning{\let%=\startMPconversion} \egroup %D In earlier versions we used the sequence %D %D \starttyping %D \expandafter\handleMPsequence\input filename\relax %D \stoptyping %D %D Persistent problems in \LATEX\ however forced us to use a %D different scheme. Every \POSTSCRIPT\ file starts with a %D \type{%}, so we temporary make this an active character %D that starts the scanning and redefines itself. (The problem %D originates in the redefinition by \LATEX\ of the %D \type{\input} primitive.) \def\startMPconversion {\ignoreMPspecials \handleMPsequence} %D Here comes the main loop. Most arguments are numbers. This %D means that they can be recognized by their \type{\lccode}. %D This method saves a lot of processing time. We could %D speed up the conversion by handling the \type{path} %D seperately. \def\@EAEAEA{\expandafter\expandafter\expandafter} % to be sure \def\dohandleMPsequence#1% {\ifdone \ifcase\lccode`#1\relax \@EAEAEA\dohandleMPsequenceA \else \@EAEAEA\dohandleMPsequenceB \fi \else \@EA\dohandleMPsequenceC \fi#1} %\def\dohandleMPsequenceA#1 % % {\setMPargument{#1}% % \handleMPsequence} \let\dohandleMPsequenceA\setMPsequence \def\dohandleMPsequenceB#1 % {\edef\somestring{#1}% \ifx\somestring\PSmoveto \edef\lastMPmoveX{\gMPa1}% \edef\lastMPmoveY{\gMPa2}% \PDFcode{\!MPgMPa1 \!MPgMPa2 m}% \resetMPstack \else\ifx\somestring\PSnewpath \let\handleMPsequence\handleMPpath \else\ifx\somestring\PSgsave \PDFcode{q}% \resetMPstack \else\ifx\somestring\PSgrestore \PDFcode{Q}% \resetMPstack \else\ifx\somestring\PSdtransform % == setlinewidth \let\handleMPsequence\handleMPdtransform % after that we will encounter more tokens until setlinewidth+pop % or pop+setlinewidth which we catch next; we explicitly need to % reset the stack since [] n setdash may follow; a more clever % approach would be to read on till the condition is met, but it's % the only pop / setlinewidth we will encounter so ... \else\ifx\somestring\PSsetlinewidth % already handled in dtransform \resetMPstack \else\ifx\somestring\PSpop % already handled in dtransform \resetMPstack \else\ifx\somestring\PSconcat \PDFcode{\gMPa1 \gMPa2 \gMPa3 \gMPa4 \gMPa5 \gMPa6 cm}% \resetMPstack \else\ifx\somestring\PSsetrgbcolor \handleMPrgbcolor \resetMPstack \else\ifx\somestring\PSsetcmykcolor \handleMPcmykcolor \resetMPstack \else\ifx\somestring\PSsetgray \handleMPgraycolor \resetMPstack \else\ifx\somestring\PStranslate \PDFcode{1 0 0 1 \gMPa1 \gMPa2 cm}% \resetMPstack \else\ifx\somestring\PSsetdash \handleMPsetdash \resetMPstack \else\ifx\somestring\PSsetlinejoin \PDFcode{\gMPa1 j}% \resetMPstack \else\ifx\somestring\PSsetmiterlimit \PDFcode{\gMPa1 M}% \resetMPstack \else\ifx\somestring\PSfshow \PDFcode{n}% \handleMPfshow \resetMPstack \else\ifx\somestring\PSsetlinecap \PDFcode{\gMPa1 J}% \resetMPstack \else\ifx\somestring\PSrlineto \PDFcode{\!MP\lastMPmoveX\space\!MP\lastMPmoveY\space l S}% \resetMPstack \else\ifx\somestring\PSscale \PDFcode{\gMPa1 0 0 \gMPa2 0 0 cm}% \resetMPstack \else\ifx\somestring\PSspecial \handleMPspecialcommand \resetMPstack \else \handleMPgraphic% {#1}% \fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi \handleMPsequence} \def\dohandleMPsequenceC#1 % {\edef\somestring{#1}% \handleMPgraphic % {#1}% \handleMPsequence} %D Since colors are not sensitive to transformations, they %D are sometimes used for signaling. Therefore, we handle them %D separately. The next macro can be redefined if needed. \def\handleMPrgbcolor {\PDFcode{\!MPgMPa1 \!MPgMPa2 \!MPgMPa3 rg \!MPgMPa1 \!MPgMPa2 \!MPgMPa3 RG}} \def\handleMPcmykcolor {\PDFcode{\!MPgMPa1 \!MPgMPa2 \!MPgMPa3 \!MPgMPa4 k \!MPgMPa1 \!MPgMPa2 \!MPgMPa3 \!MPgMPa4 K}} \def\handleMPgraycolor {\PDFcode{\!MPgMPa1 g \!MPgMPa1 G}} \def\handleMPspotcolor {\PDFcode{0 g 0 G}} %D Beginning and ending the graphics is taken care of by the %D macro \type{\handleMPgraphic}, which is redefined when %D the first graphics operator is met. \def\handleMPendgraphic % #1% {\ifx\somestring\PSshowpage \let\handleMPsequence\finishMPgraphic \else \letMPargument\somestring % {#1}% \fi} \def\handleMPbegingraphic % #1% {\ifx\somestring\PSBoundingBox \def\handleMPsequence{\handleMPboundingbox1}% \else\ifx\somestring\PSHiResBoundingBox \def\handleMPsequence{\handleMPboundingbox2}% \else\ifx\somestring\PSExactBoundingBox \def\handleMPsequence{\handleMPboundingbox3}% \else\ifx\somestring\PSshowpage \let\handleMPsequence\finishMPgraphic \else\ifx\somestring\PSPage \let\handleMPsequence\handleMPpage \else\ifx\somestring\PSMetaPostSpecials \let\handleMPsequence\handleMPspecialscomment \else\ifx\somestring\PSMetaPostSpecial \let\handleMPsequence\handleMPspecialcomment \else \letMPargument\somestring % {#1}% \fi\fi\fi\fi\fi\fi\fi} \let\handleMPgraphic=\handleMPbegingraphic %D We check for three kind of bounding boxes: the normal one %D and two high precision ones: %D %D \starttyping %D BoundingBox: llx lly ucx ucy %D HiResBoundingBox: llx lly ucx ucy %D ExactBoundingBox: llx lly ucx ucy %D \stoptyping %D %D The original as well as the recalculated dimensions are %D saved for later use. \newif\ifskipemptyMPgraphic \skipemptyMPgraphicfalse \chardef\currentMPboundingbox=0 \def\handleMPboundingbox#1#2 #3 #4 #5 {\ifnum#1>\currentMPboundingbox \xdef\MPllx{#2}\xdef\MPlly{#3}% \xdef\MPurx{#4}\xdef\MPury{#5}% \dimen0=#2\onepoint \dimen0=-\MPxscale\dimen0 \dimen2=#3\onepoint \dimen2=-\MPyscale\dimen2 \xdef\MPxoffset{\withoutpt\the\dimen0}% \xdef\MPyoffset{\withoutpt\the\dimen2}% \dimen0=#2\onebasepoint \dimen0=-\dimen0 \dimen2=#3\onebasepoint \dimen2=-\dimen2 \advance\dimen0 #4\onebasepoint \dimen0=\MPxscale\dimen0 \xdef\MPwidth{\the\dimen0}% \advance\dimen2 #5\onebasepoint \xdef\MPyshift{\the\dimen2}% unscaled \dimen2=\MPyscale\dimen2 \xdef\MPheight{\the\dimen2}% \chardef\currentMPboundingbox#1\relax \fi \doresetMPstack \let\handleMPsequence\dohandleMPsequence \let\next\handleMPsequence \ifskipemptyMPgraphic \ifdim\MPheight=\zeropoint\relax\ifdim\MPwidth=\zeropoint\relax \def\next{\endinput\finishMPgraphic}% \fi\fi \fi \next} %D Unless defined otherwise, we simply ignore specialcomments. \def\handleMPspecialcomment {\doresetMPstack \let\handleMPsequence\dohandleMPsequence \handleMPsequence} \let\handleMPspecialscomment\handleMPspecialcomment %D We use the \type{page} comment as a signal that %D stackbuilding can be started. \def\handleMPpage #1 #2 {\doresetMPstack \donetrue \let\handleMPsequence\dohandleMPsequence \handleMPsequence} %D The same applies to the special extensions. \def\handleMPspecialcommand {\doresetMPstack \let\handleMPsequence\dohandleMPsequence \handleMPsequence} %D \METAPOST\ draws its dots by moving to a location and %D invoking \type{0 0 rlineto}. This operator is not %D available in \PDF. Our solution is straightforward: we draw %D a line from $(current\_x, current\_y)$ to itself. This %D means that the arguments of the preceding \type{moveto} have %D to be saved. \def\lastMPmoveX{0} \def\lastMPmoveY{0} %D These saved coordinates are also used when we handle the %D texts. Text handling proved to be a bit of a nuisance, but %D finally I saw the light. It proved that we also had to %D take care of \type{(split arguments)}. \def\setMPfshowfont#1#2% {\font\temp=#1\space at #2\relax\temp} \let\MPfshowcommand\empty \def\dohandleMPfshow {\bgroup \setbox\scratchbox\hbox {\obeyMPspecials \edef\size{\gMPa\nofMParguments}% \ifx\size\PSnfont % round font size (to pt) \advance\nofMParguments \minusone \expandafter\scratchdimen\gMPa\nofMParguments\onepoint\relax \ifdim\scratchdimen<\onepoint \def\size{1pt}% \else \advance\scratchdimen .5\onepoint \def\size##1.##2\relax{\def\size{##1pt}}% \expandafter\size\the\scratchdimen\relax \fi \else \edef\size{\size bp}% \fi \advance\nofMParguments \minusone %\font\temp=\gMPa\nofMParguments\space at \size \let\temp\relax % to be sure \setMPfshowfont{\gMPa\nofMParguments}\size \advance\nofMParguments \minusone \temp \MPfshowcommand {\ifnum\nofMParguments=\plusone \def\do(##1){##1}% \dogMPa1% \else % we need to catch ( a ) (a a a) (\123 \123 \123) etc \scratchcounter\plusone \def\dodo##1% Andreas Fieger's bug: (\304...) {\edef\!!stringa{##1\empty\empty}% and another one: ( 11) -> \ifx 11 \ifx\!!stringa\MPspacechar\MPspacechar\else\expandafter##1\fi}% \def\do(##1{\dodo{##1}}% \dogMPa\scratchcounter\MPspacechar \let\do\relax \loop \advance\scratchcounter \plusone \ifnum\scratchcounter<\nofMParguments\relax \gMPa\scratchcounter\MPspacechar \repeat \def\do##1){\dodo{##1}}% \dogMPa\scratchcounter \fi \unskip}}% % % this fails in some versions of pdftex % % \dimen0=\lastMPmoveY bp % \advance\dimen0 by \ht0 % \ScaledPointsToBigPoints{\number\dimen0}\lastMPmoveY % \PDFcode{n q 1 0 0 1 \lastMPmoveX\space\lastMPmoveY\space cm}% % \dimen0=\ht0 % \advance\dimen0 by \dp0 % \box0 % \vskip-\dimen0 % \PDFcode{Q}% % \egroup} % \setbox\scratchbox\hbox {\hskip\lastMPmoveX\onebasepoint\raise\lastMPmoveY\onebasepoint\box\scratchbox}% \ht\scratchbox\zeropoint \dp\scratchbox\zeropoint \wd\scratchbox\zeropoint \box\scratchbox \egroup} \let\handleMPfshow\dohandleMPfshow % so we can overload this one later %D You could consider the following definition to be the most %D natural one. % \def\MPspacechar{\space} % normal case \def\MPspacechar{\char32\relax} % old solution does not work with math %D However, the following implementation is more robust, since %D some fonts have funny visible spaces in the space slot. This %D gives a mismatch between the space that \METAPOST\ took into %D account and the \quote {natural} space. This only happens in %D labels, since \type {btex}||\type {etex} thingies don't have %D spaces. This phenomena showed up when preparing the %D \METAFUN\ manual, where Palatino fonts are used. We can %D safely assume that \METAPOST\ considers \type {\char32} to %D be the space. \def\MPspacechar{\setbox\scratchbox\hbox{\char32}\kern\wd\scratchbox} %D Well, this does not work with math fonts, so: \def\MPspacechar{\char32\relax} %D Most operators are just converted and keep their %D arguments. Dashes however need a bit different treatment, %D otherwise \PDF\ viewers complain loudly. Another %D complication is that one argument comes after the \type{]}. %D When reading the data, we simply ignore the array boundary %D characters. We save ourselves some redundant newlines and %D at the same time keep the output readable by packing the %D literals. \def\handleMPsetdash {\bgroup \def\somestring{[}% \scratchcounter\plusone \loop \ifnum\scratchcounter<\nofMParguments \edef\somestring{\somestring\space\gMPa\scratchcounter}% \advance\scratchcounter \plusone \repeat \edef\somestring{\somestring]\gMPa\scratchcounter\space d}% \PDFcode{\somestring}% \egroup} %D The \type{setlinewidth} commands looks a bit complicated. There are %D two alternatives, that result in a similar look in both %D $x$- and $y$-dorection. As John Hobby says: %D %D \startnarrower \switchtobodyfont[ss] %D \starttyping %D x 0 dtransform exch truncate exch idtransform pop setlinewidth %D 0 y dtransform truncate idtransform setlinewidth pop %D \stoptyping %D %D These are just fancy versions of \type{x setlinewidth} and %D \type{y setlinewidth}. The \type{x 0 ...} form is used if %D the path is {\em primarily vertical}. It rounds the width %D so that vertical lines come out an integer number of pixels %D wide in device space. The \type{0 y ...} form does the same %D for paths that are {\em primarily horizontal}. The reason %D why I did this is Knuth insists on getting exactly the %D widths \TEX\ intends for the horizontal and vertical rules %D in \type{btex...etex} output. (Note that PostScript scan %D conversion rules cause a horizontal or vertical line of %D integer width $n$ in device space to come out $n+1$ pixels %D wide, regardless of the phase relative to the pixel grid.) %D \stopnarrower %D %D The common operator in these sequences is \type{dtransform}, %D so we can use this one to trigger setting the linewidth. \def\handleMPdtransform {\ifdim\gMPa1\onepoint>\zeropoint \PDFcode{\gMPa1 w}% \def\next##1 ##2 ##3 ##4 ##5 ##6 {\handleMPsequence}% \else \PDFcode{\gMPa2 w}% \def\next##1 ##2 ##3 ##4 {\handleMPsequence}% \fi \let\handleMPsequence\dohandleMPsequence \resetMPstack \next} %D The most complicated command is \type{concat}. \METAPOST\ %D applies this operator to \type{stroke}. At that moment the %D points set by \type{curveto} and \type{moveto}, are already %D fixed. In \PDF\ however the \type{cm} operator affects the %D points as well as the pen (stroke). Like more \PDF\ %D operators, \type{cm} is defined in a bit ambiguous way. %D The only save route for non||circular penshapes, is saving %D the path, recalculating the points and applying the %D transformation matrix in such a way that we can be sure %D that its behavior is well defined. This comes down to %D inverting the path and applying \type{cm} to that path as %D well as the pen. This all means that we have to save the %D path. %D In \METAPOST\ there are three ways to handle a path $p$: %D %D \starttyping %D draw p; fill p; filldraw p; %D \stoptyping %D %D The last case outputs a \type{gsave fill grestore} before %D \type{stroke}. Handling the path outside the main loops %D saves about 40\% run time.\footnoot{We can save some more by %D following the \METAPOST\ output routine, but for the moment %D we keep things simple.} Switching between the main loop and %D the path loop is done by means of the recursely called %D macro \type{\handleMPsequence}. \def\handleMPpath {\chardef\finiMPpath0 \let\closeMPpath\relax \let\flushMPpath\flushnormalMPpath \resetMPstack \nofMPsegments\plusone \let\handleMPsequence\dohandleMPpath \dohandleMPpath} %D Most paths are drawn with simple round pens. Therefore we've %D split up the routine in two. \def\flushnormalMPsegment {\ifcase\getMPkeyword\relax \PDFcode{\!MPgMPs1 \!MPgMPs2 l}% \or \PDFcode{\!MPgMPs1 \!MPgMPs2 \!MPgMPs3 \!MPgMPs4 \!MPgMPs5 \!MPgMPs6 c}% \or \PDFcode{\!MP\lastMPmoveX\space\!MP\lastMPmoveY\space l S}% \or \edef\lastMPmoveX{\gMPs1}% evt \!MP here \edef\lastMPmoveY{\gMPs2}% \PDFcode{\!MP\lastMPmoveX\space \!MP\lastMPmoveY\space m}% \fi} \def\flushconcatMPsegment {\ifcase\getMPkeyword\relax \doMPconcat{\gMPs1}\a{\gMPs2}\b% \PDFcode{\!MP\a\space\!MP\b\space l}% \or \doMPconcat{\gMPs1}\a{\gMPs2}\b% \doMPconcat{\gMPs3}\c{\gMPs4}\d% \doMPconcat{\gMPs5}\e{\gMPs6}\f% \PDFcode{\!MP\a\space\!MP\b\space \!MP\c\space\!MP\d\space \!MP\e\space\!MP\f\space c}% \or \bgroup \noMPtranslate \doMPconcat\lastMPmoveX\a\lastMPmoveY\b% \PDFcode{\!MP\a\space\!MP\b\space l S}% \egroup \or \edef\lastMPmoveX{\gMPs1}% \edef\lastMPmoveY{\gMPs2}% \doMPconcat\lastMPmoveX\a\lastMPmoveY\b% \PDFcode{\!MP\a\space\!MP\b\space m}% \fi} % \def\flushnormalMPpath % {\scratchcounter\nofMPsegments % \nofMPsegments\plusone % \loop % \flushnormalMPsegment % \advance\nofMPsegments \plusone % \ifnum\nofMPsegments<\scratchcounter % \repeat} % % \def\flushconcatMPpath % {\scratchcounter\nofMPsegments % \nofMPsegments\plusone % \loop % \flushconcatMPsegment % \advance\nofMPsegments \plusone % \ifnum\nofMPsegments<\scratchcounter % \repeat} % % an alternative is presented below: (no \def assignment) \def\doflushsomeMPpath {\dodoflushsomeMPpath \advance\nofMPsegments \plusone \ifnum\nofMPsegments<\scratchcounter \expandafter\doflushsomeMPpath \fi} \def\flushsomeMPpath {\scratchcounter\nofMPsegments \nofMPsegments\plusone \doflushsomeMPpath} \def\flushnormalMPpath{\let\dodoflushsomeMPpath\flushnormalMPsegment\flushsomeMPpath} %OLD \def\flushconcatMPpath{\let\dodoflushsomeMPpath\flushconcatMPsegment\flushsomeMPpath} %NEW pre-calculate 1/D so it needn't be repeated for each control point. \def\flushconcatMPpath {\MPreciprocaldeterminant \let\dodoflushsomeMPpath\flushconcatMPsegment\flushsomeMPpath} %D The transformation of the coordinates is handled by one of %D the macros Tanmoy posted to the \PDFTEX\ mailing list. %D I rewrote and optimized the original macro to suit the other %D macros in this module. %D %D \starttyping %D \doMPconcat {x position} \xresult {y position} \yresult %D \stoptyping %D %D By setting the auxiliary \DIMENSIONS\ \type{\dimen0} upto %D \type{\dimen10} only once per path, we save over 20\% run %D time. Some more speed was gained by removing some parameter %D passing. These macros can be optimized a bit more by using %D more constants. There is however not much need for further %D optimization because penshapes usually are round and %D therefore need no transformation. Nevertheless we move the %D factor to the outer level and use a bit different \type{pt} %D removal macro. Although the values represent base points, %D we converted them to pure points, simply because those can %D be converted back. %OLD \mathchardef\MPconcatfactor=256 % beware don't remove spaces before it %OLD \def\doMPreducedimen#1 %OLD {\count0\MPconcatfactor %OLD \advance\dimen#1 \ifdim\dimen#1>\zeropoint .5\else -.5\fi\count0 %OLD \divide\dimen#1 \count0\relax} %OLD % too inaccurate (see old pragma logo) %OLD %OLD \def\doMPreducedimen#1 %OLD {\count0=\MPconcatfactor %OLD \divide\dimen#1 \count0\relax} %OLD \def\doMPreducedimen#1 %OLD {\advance\dimen#1 \ifdim\dimen#1>\zeropoint .5\else -.5\fi\MPconcatfactor %OLD \divide\dimen#1 \MPconcatfactor} %D The transformation code is rewritten by Daniel H. Luecking who %D describes his patch as follows: %D %D We would like to divide 1 by $X4, but all divisions are integer so %D for accuracy we want to convert to large integers and make sure the %D integer quotient has as many significant digits as possible. Thus we %D need to replace $1/X$ with $M/N$ where $N$ is as large as possible %D and $M/N$ is as large as possible. Also for simplicity $M$ should be %D a power of 2. So we make $M = 2^{30}$ \footnote{$2^{31} - 1$ is the %D largest legal integer. Using it (and simply ignoring the inaccuracy %D caused by $-1$) turns out to be at least as accurate in all cases, %D and more accurate in some.} (largest legal power of 2) and adjust %D $X4 downward (if necessary) to the the range $1-2^{16}$. This gives %D at least 15 significant binary digits, (almost as accurate as %D \METAPOST\ for numbers near 1) or almost 5 significant figures %D (decimal). \newcount\MPscratchCnt \newdimen\MPscratchDim % will be assigned global \def\MPadjustdimen % sets \MPscratchDim and \MPscratchCnt {\MPscratchCnt\zerocount \doMPadjustdimen} \def\doMPadjustdimen {\ifdim\MPscratchDim>\onepoint \divide\MPscratchDim 2 \advance\MPscratchCnt \plusone \expandafter\doMPadjustdimen \fi} %OLD \def\doMPexpanddimen#1 %OLD {\multiply\dimen#1 \MPconcatfactor\relax} %D DHL: When viewed as an integer, $1 \hbox{pt}=2^{16}$ so $2^{32}/X$ %D is the right way to do $(1 \hbox{pt})/(X \hbox{pt})$ and get the %D answer in points. But we are limited to $2^{30}/X$. However, we %D actually do $[ 2^{30} / (X/2^K) ]*2^{2-K}$ where $K$ is the number %D of halvings it takes to bring $X4 below $1 \hbox{pt}$. If $K$ is 0 %D or 1 we readjust by multiplying by 4 or 2, otherwise by halving %D $(K-2)$ times \type {\MPscratchCnt} holds the value of $K$ from %D \type {\MPadjustdimen}. \def\MPreadjustdimen % acts on \MPscratchDim and MPscratchCnt {\ifcase\MPscratchCnt \multiply\MPscratchDim 4 \or \multiply\MPscratchDim 2 \else \expandafter\doMPreadjustdimen \fi} \def\doMPreadjustdimen {\ifnum\MPscratchCnt > 2 \divide\MPscratchDim 2 \advance\MPscratchCnt \minusone \expandafter\doMPreadjustdimen \fi} \def\MPreciprocaldeterminant {\MPscratchDim\withoutpt\the\dimen0 \dimen6 % s_x*s_y \advance\MPscratchDim - \withoutpt\the\dimen2 \dimen4 % s_x*s_y - r_x*r_y \ifdim\MPscratchDim<\zeropoint % we need a positive dimension \MPscratchDim-\MPscratchDim % for \MPadjustdimen \doMPreciprocal \MPscratchDim-\MPscratchDim \else \doMPreciprocal \fi} \newcount\MPnumerator \MPnumerator = 1073741824 % 2^{30} % todo: dimexpr \def\doMPreciprocal % replace \MPscratchDim with its reciprocal {\ifdim\MPscratchDim=\onepoint \else \MPadjustdimen \scratchcounter\MPnumerator \divide\scratchcounter\MPscratchDim \MPscratchDim1\scratchcounter % 1 needed ! \MPreadjustdimen \fi} %OLD \def\presetMPconcat %OLD {\dimen 0=\gMPs1\onepoint \doMPreducedimen 0 % r_x %OLD \dimen 2=\gMPs2\onepoint \doMPreducedimen 2 % s_x %OLD \dimen 4=\gMPs3\onepoint \doMPreducedimen 4 % s_y %OLD \dimen 6=\gMPs4\onepoint \doMPreducedimen 6 % r_y %OLD \dimen 8=\gMPs5\onepoint \doMPreducedimen 8 % t_x %OLD \dimen10=\gMPs6\onepoint \doMPreducedimen10 } % t_y %OLD %OLD \def\presetMPscale %OLD {\dimen 0=\gMPs1\onepoint \doMPreducedimen 0 %OLD \dimen 2 \zeropoint %OLD \dimen 4 \zeropoint %OLD \dimen 6=\gMPs2\onepoint \doMPreducedimen 6 %OLD \dimen 8 \zeropoint %OLD \dimen10 \zeropoint} \def\presetMPconcat {\dimen 0=\gMPs1\onepoint % s_x \dimen 2=\gMPs2\onepoint % r_x \dimen 4=\gMPs3\onepoint % r_y \dimen 6=\gMPs4\onepoint % s_y \dimen 8=\gMPs5\onepoint % t_x \dimen10=\gMPs6\onepoint} % t_y \def\presetMPscale {\dimen 0=\gMPs1\onepoint \dimen 2 \zeropoint \dimen 4 \zeropoint \dimen 6=\gMPs2\onepoint \dimen 8 \zeropoint \dimen10 \zeropoint} \def\noMPtranslate % use this one grouped {\dimen 8 \zeropoint % t_x \dimen10 \zeropoint} % t_y %D \starttyping %D \def\doMPconcat#1#2#3#4% %D {\dimen12=#1 pt \doMPreducedimen12 % p_x %D \dimen14=#3 pt \doMPreducedimen14 % p_y %D % %D \dimen16 \dimen 0 %D \multiply \dimen16 \dimen 6 %D \dimen20 \dimen 2 %D \multiply \dimen20 \dimen 4 %D \advance \dimen16 -\dimen20 %D % %D \dimen18 \dimen12 %D \multiply \dimen18 \dimen 6 %D \dimen20 \dimen14 %D \multiply \dimen20 \dimen 4 %D \advance \dimen18 -\dimen20 %D \dimen20 \dimen 4 %D \multiply \dimen20 \dimen10 %D \advance \dimen18 \dimen20 %D \dimen20 \dimen 6 %D \multiply \dimen20 \dimen 8 %D \advance \dimen18 -\dimen20 %D % %D \multiply \dimen12 -\dimen 2 %D \multiply \dimen14 \dimen 0 %D \advance \dimen12 \dimen14 %D \dimen20 \dimen 2 %D \multiply \dimen20 \dimen 8 %D \advance \dimen12 \dimen20 %D \dimen20 \dimen 0 %D \multiply \dimen20 \dimen10 %D \advance \dimen12 -\dimen20 %D % %D \doMPreducedimen16 %D \divide \dimen18 \dimen16 \doMPexpanddimen18 %D \divide \dimen12 \dimen16 \doMPexpanddimen12 %D % %D \edef#2{\withoutpt\the\dimen18}% % p_x^\prime %D \edef#4{\withoutpt\the\dimen12}} % p_y^\prime %D \stoptyping %D The following optimization resulted from some tests by %D and email exchanges with Sanjoy Mahajan. %D %D \starttyping %D \def\doMPconcat#1#2#3#4% %D {\dimen12=#1 pt \doMPreducedimen12 % p_x %D \dimen14=#3 pt \doMPreducedimen14 % p_y %D % %D \dimen16 \dimen 0 %D \multiply \dimen16 \dimen 6 %D \dimen20 \dimen 2 %D \multiply \dimen20 \dimen 4 %D \advance \dimen16 -\dimen20 %D % %D \dimen18 \dimen12 %D \multiply \dimen18 \dimen 6 %D \dimen20 \dimen14 %D \multiply \dimen20 \dimen 4 %D \advance \dimen18 -\dimen20 %D \dimen20 \dimen 4 %D \multiply \dimen20 \dimen10 %D \advance \dimen18 \dimen20 %D \dimen20 \dimen 6 %D \multiply \dimen20 \dimen 8 %D \advance \dimen18 -\dimen20 %D % %D \multiply \dimen12 -\dimen 2 %D \multiply \dimen14 \dimen 0 %D \advance \dimen12 \dimen14 %D \dimen20 \dimen 2 %D \multiply \dimen20 \dimen 8 %D \advance \dimen12 \dimen20 %D \dimen20 \dimen 0 %D \multiply \dimen20 \dimen10 %D \advance \dimen12 -\dimen20 %D % %D %\ifdim\dimen16>1pt % oeps, can be < 1pt too %D \ifdim\dimen16=1pt \else %D \ifdim\dimen16>\MPconcatfactor pt %D \doMPreducedimen16 %D \divide \dimen18 \dimen16 \doMPexpanddimen18 %D \divide \dimen12 \dimen16 \doMPexpanddimen12 %D \else %D \divide \dimen18 \dimen16 \doMPexpanddimen18 \doMPexpanddimen18 %D \divide \dimen12 \dimen16 \doMPexpanddimen12 \doMPexpanddimen12 %D \fi %D \fi %D % %D \edef#2{\withoutpt\the\dimen18}% % p_x^\prime %D \edef#4{\withoutpt\the\dimen12}} % p_y^\prime %D \stoptyping %D %D But, this one is still too inaccurate, so we now have: % \def\doMPconcat#1#2#3#4% % {\dimen12=#1pt % p_x % \dimen14=#3pt % p_y % % % % we should test for >-1024 too, but for the moment take the gamble % \chardef\MPfactor1\ifdim\dimen12<1024pt \ifdim\dimen14<1024pt 6\fi\fi % % % \multiply\dimen12 \MPfactor % \multiply\dimen14 \MPfactor % % % \doMPreducedimen12 % \doMPreducedimen14 % % % \dimen16 \dimen 0 % \multiply \dimen16 \dimen 6 % \dimen20 \dimen 2 % \multiply \dimen20 \dimen 4 % \advance \dimen16 -\dimen20 % % % \dimen18 \dimen12 % \multiply \dimen18 \dimen 6 % \dimen20 \dimen14 % \multiply \dimen20 \dimen 4 % \advance \dimen18 -\dimen20 % \dimen20 \dimen 4 % \multiply \dimen20 \dimen10 % \advance \dimen18 \dimen20 % \dimen20 \dimen 6 % \multiply \dimen20 \dimen 8 % \advance \dimen18 -\dimen20 % % % \multiply \dimen12 -\dimen 2 % \multiply \dimen14 \dimen 0 % \advance \dimen12 \dimen14 % \dimen20 \dimen 2 % \multiply \dimen20 \dimen 8 % \advance \dimen12 \dimen20 % \dimen20 \dimen 0 % \multiply \dimen20 \dimen10 % \advance \dimen12 -\dimen20 % % % \ifdim\dimen16=1pt \else % \ifdim\dimen16>\MPconcatfactor pt % \doMPreducedimen16 % \divide \dimen18 \dimen16 \doMPexpanddimen18 % \divide \dimen12 \dimen16 \doMPexpanddimen12 % \else % \divide \dimen18 \dimen16 \doMPexpanddimen18 \doMPexpanddimen18 % \divide \dimen12 \dimen16 \doMPexpanddimen12 \doMPexpanddimen12 % \fi % \fi % % % \divide\dimen18 \MPfactor % \divide\dimen12 \MPfactor % % % \edef#2{\withoutpt\the\dimen18}% % p_x^\prime % \edef#4{\withoutpt\the\dimen12}} % p_y^\prime %D We cannot use \type {\beginETEX} here since in plain we %D get \type {\outer} problems, sigh. %OLD \beginTEX %OLD %OLD \def\MPcriteriumA {512pt} % scale %OLD \def\MPcriteriumB {2pt} % scale %OLD %OLD \endTEX %OLD %OLD \ifx\MPcriteriumA\undefined %OLD %OLD \newdimen\MPcriteriumA \MPcriteriumA=512pt %OLD \newdimen\MPcriteriumB \MPcriteriumB= 2pt %OLD %OLD \fi %OLD \def\doMPconcat#1#2#3#4% %OLD {\dimen12=#1pt % p_x %OLD \dimen14=#3pt % p_y %OLD % %OLD \chardef\MPfactor\zerocount %OLD \ifdim\dimen4<\MPcriteriumB\ifdim\dimen4>-\MPcriteriumB %OLD \ifdim\dimen6<\MPcriteriumB\ifdim\dimen6>-\MPcriteriumB %OLD \ifdim\dimen8<\MPcriteriumB\ifdim\dimen8>-\MPcriteriumB %OLD \ifdim\dimen10<\MPcriteriumB\ifdim\dimen10>-\MPcriteriumB %OLD \chardef\MPfactor\plusone %OLD \fi\fi %OLD \fi\fi %OLD \fi\fi %OLD \fi\fi %OLD \ifcase\MPfactor % spurious 0 removed %OLD \chardef\MPfactor\plusone %OLD \ifdim\dimen12<\MPcriteriumA\ifdim\dimen12>-\MPcriteriumA %OLD \ifdim\dimen14<\MPcriteriumA\ifdim\dimen14>-\MPcriteriumA %OLD \chardef\MPfactor16 %OLD \fi\fi %OLD \fi\fi %OLD \fi %OLD % %OLD \multiply\dimen12 \MPfactor %OLD \multiply\dimen14 \MPfactor %OLD % %OLD \doMPreducedimen12 %OLD \doMPreducedimen14 %OLD % %OLD \dimen16 \dimen 0 %OLD \multiply \dimen16 \dimen 6 %OLD \dimen20 \dimen 2 %OLD \multiply \dimen20 \dimen 4 %OLD \advance \dimen16 -\dimen20 %OLD % %OLD \dimen18 \dimen12 %OLD \multiply \dimen18 \dimen 6 %OLD \dimen20 \dimen14 %OLD \multiply \dimen20 \dimen 4 %OLD \advance \dimen18 -\dimen20 %OLD \dimen20 \dimen 4 %OLD \multiply \dimen20 \dimen10 %OLD \advance \dimen18 \dimen20 %OLD \dimen20 \dimen 6 %OLD \multiply \dimen20 \dimen 8 %OLD \advance \dimen18 -\dimen20 %OLD % %OLD \multiply \dimen12 -\dimen 2 %OLD \multiply \dimen14 \dimen 0 %OLD \advance \dimen12 \dimen14 %OLD \dimen20 \dimen 2 %OLD \multiply \dimen20 \dimen 8 %OLD \advance \dimen12 \dimen20 %OLD \dimen20 \dimen 0 %OLD \multiply \dimen20 \dimen10 %OLD \advance \dimen12 -\dimen20 %OLD % %OLD \ifdim\dimen16=\onepoint \else %OLD \ifdim\dimen16>\MPconcatfactor \onepoint \relax %OLD \doMPreducedimen16 %OLD \divide \dimen18 \dimen16 \doMPexpanddimen18 %OLD \divide \dimen12 \dimen16 \doMPexpanddimen12 %OLD \else %OLD \divide \dimen18 \dimen16 \doMPexpanddimen18 \doMPexpanddimen18 %OLD \divide \dimen12 \dimen16 \doMPexpanddimen12 \doMPexpanddimen12 %OLD \fi %OLD \fi %OLD % %OLD \divide\dimen18 \MPfactor %OLD \divide\dimen12 \MPfactor %OLD % %OLD \edef#2{\withoutpt\the\dimen18}% % p_x^\prime %OLD \edef#4{\withoutpt\the\dimen12}} % p_y^\prime %D DHL: Ideally, $r_x$, $r_y$, $s_x4, $s_y$ should be in macros, not %D dimensions (they are scalar quantities after all, not lengths). I %D suppose the authors decided to do calculations with integer %D arithmetic instead of using real factors because it's faster. %D However, the actual macros test slower, possibly because I've %D omitted three nested loops. In my test files, my approach is more %D accurate. It is also far simpler and overflow does not seem to be a %D significant concern. The scale factors written by Metapost are (?) %D always $<=1$ (it scales coordinates internally) and coordinates are %D always likely to be less than \type {\maxdimen}. %D %D If this should ever cause problems, the scale factors can be reduced. \def\doMPconcat#1#2#3#4% {\dimen12=#1pt % p_x % #1\onepoint \dimen14=#3pt % p_y % #3\onepoint \advance\dimen12 -\dimen8 % p_x - t_x \advance\dimen14 -\dimen10 % p_y - t_y \dimen18=\withoutpt\the\dimen6 \dimen12 % s_y(p_x - t_x) \advance\dimen18 -\withoutpt\the\dimen4 \dimen14 % - r_y(p_y-t_y) \dimen14=\withoutpt\the\dimen0 \dimen14 % s_x(p_y-t_y) \advance\dimen14 -\withoutpt\the\dimen2 \dimen12 % - r_x(p_x-t_x) % \MPscratchDim contains precomputed 1/D: \dimen18=\withoutpt\the\MPscratchDim \dimen18 \dimen14=\withoutpt\the\MPscratchDim \dimen14 \edef#2{\withoutpt\the\dimen18}% % p_x^\prime \edef#4{\withoutpt\the\dimen14}} % p_y^\prime %D One reason for Daniel to write this patch was that at small sizes %D the accuracy was less than optimal. Here is a test that demonstrates %D that his alternative is pretty good: %D %D \startlinecorrection %D \startMPcode %D for i = 5cm,1cm,5mm,1mm,.5mm,.1mm,.01mm : %D draw fullcircle scaled i withpen pencircle xscaled (i/10) yscaled (i/20) rotated 45 ; %D endfor ; %D \stopMPcode %D \stoplinecorrection %D The following explanation of the conversion process was %D posted to the \PDFTEX\ mailing list by Tanmoy. The original %D macro was part of a set of macro's that included sinus and %D cosinus calculations as well as scaling and translating. The %D \METAPOST\ to \PDF\ conversion however only needs %D transformation. %M \start \switchtobodyfont [ss] %D Given a point $(U_x, U_y)$ in user coordinates, the business %D of \POSTSCRIPT\ is to convert it to device space. Let us say %D that the device space coordinates are $(D_x, D_y)$. Then, in %D \POSTSCRIPT\ $(D_x, D_y)$ can be written in terms of %D $(U_x, U_y)$ in matrix notation, either as %D %D \placeformula %D \startformula %D \pmatrix{D_x&D_y&1\cr} = \pmatrix{U_x&U_y&1\cr} %D \pmatrix{s_x&r_x&0\cr %D r_y&s_y&0\cr %D t_x&t_y&1\cr} %D \stopformula %D %D or %D %D \placeformula %D \startformula %D \pmatrix{D_x\cr D_y\cr 1} = \pmatrix{s_x&r_y&t_x\cr %D r_x&s_y&t_y\cr %D 0 &0 &1 \cr} %D \pmatrix{U_x\cr %D U_y\cr %D 1 \cr} %D \stopformula %D %D both of which is a shorthand for the same set of equations: %D %D \placeformula %D \startformula %D D_x = s_x U_x + r_y U_y + t_x %D \stopformula %D %D \placeformula %D \startformula %D D_y = r_x U_x + s_y U_y + t_y %D \stopformula %D %D which define what is called an `affine transformation'. %D %D \POSTSCRIPT\ represents the `transformation matrix' as a %D six element matrix instead of a $3\times 3$ array because %D three of the elements are always~0, 0 and~1. Thus the above %D transformation is written in postscript as $[s_x\, r_x\, %D r_y\, s_y\, t_x\, t_y]$. However, when doing any %D calculations, it is useful to go back to the original %D matrix notation (whichever: I will use the second) and %D continue from there. %D %D As an example, if the current transformation matrix is %D $[s_x\, r_x\, r_y\, s_y\, t_x\, t_y]$ and you say \typ{[a b %D c d e f] concat}, this means: %D %D \startnarrower %D Take the user space coordinates and transform them to an %D intermediate set of coordinates using array $[a\, b\, c\, d\, %D e\, f]$ as the transformation matrix. %D %D Take the intermediate set of coordinates and change them to %D device coordinates using array $[s_x\, r_x\, r_y\, s_y\, t_x\, t_y]$ %D as the transformation matrix. %D \stopnarrower %D %D Well, what is the net effect? In matrix notation, it is %D %D \placeformula %D \startformula %D \pmatrix{I_x\cr I_y\cr 1\cr} = \pmatrix{a&c&e\cr %D b&d&f\cr %D 0&0&1\cr} %D \pmatrix{U_x\cr %D U_y\cr %D 1 \cr} %D \stopformula %D %D \placeformula %D \startformula %D \pmatrix{D_y\cr D_y\cr 1\cr} = \pmatrix{s_x&r_y&t_x\cr %D r_x&s_y&t_y\cr %D 0 &0 &1 \cr} %D \pmatrix{I_x\cr %D I_y\cr %D 1 \cr} %D \stopformula %D %D where $(I_x, I_y)$ is the intermediate coordinate. %D %D Now, the beauty of the matrix notation is that when there is %D a chain of such matrix equations, one can always compose %D them into one matrix equation using the standard matrix %D composition law. The composite matrix from two matrices can %D be derived very easily: the element in the $i$\high{th} %D horizontal row and $j$\high{th} vertical column is %D calculated by`multiplying' the $i$\high{th} row of the first %D matrix and the $j$\high{th} column of the second matrix (and %D summing over the elements). Thus, in the above: %D %D \placeformula %D \startformula %D \pmatrix{D_x\cr D_y\cr 1} = \pmatrix{s_x^\prime&r_y^\prime&t_x^\prime\cr %D r_x^\prime&s_y^\prime&t_y^\prime\cr %D 0 &0 &0 \cr} %D \pmatrix{U_x\cr %D U_y\cr %D 1 \cr} %D \stopformula %D %D with %D %D \placeformula %D \startformula %D \eqalign %D {s_x^\prime & = s_x a + r_y b \cr %D r_x^\prime & = r_x a + s_y b \cr %D r_y^\prime & = s_x c + r_y d \cr %D s_y^\prime & = r_x c + s_y d \cr %D t_x^\prime & = s_x e + r_y f + t_x \cr %D t_y^\prime & = r_x e + s_y f + t_y \cr} %D \stopformula %D In fact, the same rule is true not only when one is going %D from user coordinates to device coordinates, but whenever %D one is composing two `transformations' together %D (transformations are `associative'). Note that the formula %D is not symmetric: you have to keep track of which %D transformation existed before (i.e.\ the equivalent of %D $[s_x\, r_x\, r_y\, s_y\, t_x\, t_y]$) and which was %D specified later (i.e.\ the equivalent of $[a\, b\, c\, d\, %D e\, f]$). Note also that the language can be rather %D confusing: the one specified later `acts earlier', %D converting the user space coordinates to intermediate %D coordinates, which are then acted upon by the pre||existing %D transformation. The important point is that order of %D transformation matrices cannot be flipped (transformations %D are not `commutative'). %D %D Now what does it mean to move a transformation matrix %D before a drawing? What it means is that given a point %D $(P_x, P_y)$ we need a different set of coordinates %D $(P_x^\prime, P_y^\prime)$ such that if the transformation %D acts on $(P_x^\prime, P_y^\prime)$, they produce $(P_x, %D P_y)$. That is we need to solve the set of equations: %D %D \placeformula %D \startformula %D \pmatrix{P_x\cr P_y\cr 1\cr} = \pmatrix{s_x&r_y&t_x\cr %D r_x&s_y&t_y\cr %D 0 &0 &1 \cr} %D \pmatrix{P_x^\prime\cr %D P_y^\prime\cr %D 1 \cr} %D \stopformula %D %D Again matrix notation comes in handy (i.e. someone has %D already solved the problem for us): we need the inverse %D transformation matrix. The inverse transformation matrix can %D be calculated very easily: %D %D \placeformula %D \startformula %D \pmatrix{P_x^\prime\cr P_y^\prime\cr 1\cr} = %D \pmatrix{s_x^\prime&r_y^\prime&t_x^\prime\cr %D r_x^\prime&s_y^\prime&t_y^\prime\cr %D 0 &0 &1 \cr} %D \pmatrix{P_x\cr %D P_y\cr %D 1 \cr} %D \stopformula %D %D where, the inverse transformation matrix is given by %D %D \placeformula %D \startformula %D \eqalign %D {D & = s_x s_y - r_x r_y \cr %D s_x^\prime & = s_y / D \cr %D s_y^\prime & = s_x / D \cr %D r_x^\prime & = - r_x / D \cr %D r_y^\prime & = - r_y / D \cr %D t_x^\prime & = ( - s_y t_x + r_y t_y ) / D \cr %D t_y^\prime & = ( r_x t_x - s_x t_y ) / D \cr} %D \stopformula %D %D And you can see that when expanded out, this does %D give the formulas: %D %D \placeformula %D \startformula %D P_x^\prime = { { s_y(p_x-t_x) + r_y(t_y-p_y) } \over %D { s_x s_y-r_x r_y } } %D \stopformula %D %D \placeformula %D \startformula %D P_y^\prime = { { s_x(p_y-t_y) + r_x(t_x-p_x) } \over %D { s_x*s_y-r_x*r_y } } %D \stopformula %D %D The code works by representing a real number by converting %D it to a dimension to be put into a \DIMENSION\ register: 2.3 would %D be represented as 2.3pt for example. In this scheme, %D multiplying two numbers involves multiplying the \DIMENSION\ %D registers and dividing by 65536. Accuracy demands that the %D division be done as late as possible, but overflow %D considerations need early division. %D %D Division involves dividing the two \DIMENSION\ registers and %D multiplying the result by 65536. Again, accuracy would %D demand that the numerator be multiplied (and|/|or the %D denominator divided) early: but that can lead to overflow %D which needs to be avoided. %D %D If nothing is known about the numbers to start with (in %D concat), I have chosen to divide the 65536 as a 256 in each %D operand. However, in the series calculating the sine and %D cosine, I know that the terms are small (because I never %D have an angle greater than 45 degrees), so I chose to %D apportion the factor in a different way. %M \stop %D The path is output using the values saved on the stack. If %D needed, all coordinates are recalculated. \def\finishMPpath {\PDFcode{\ifcase\finiMPpath W n\or S\or f\or B\fi}} \def\processMPpath {\checkMPpath \ifcase\nofMPsegments\else \flushMPpath \closeMPpath \finishMPpath \fi \let\handleMPsequence\dohandleMPsequence \resetMPstack \nofMPsegments\zerocount \handleMPsequence} %D The following \METAPOST\ code is quite valid but, when %D processed and converted to \PDF, will make a file %D unprintable on a Hewlett Packard printer (from Acrobat %D $v<=5$). Who is to blame, the driver of the OS layer in %D between, is hard to determine, so we add an additional %D check. %D %D \starttyping %D clip currentpicture to origin -- cycle ; %D setbounds currentpicture to fullsquare scaled 5cm ; %D \stoptyping \def\checkMPpath {\ifcase\finiMPpath \ifnum\nofMPsegments<3 % n is one ahead \message{omitting zero clip path}% \nofMPsegments\zerocount \fi \fi} %D In \PDF\ the \type{cm} operator must precede the path %D specification. We therefore can output the \type{cm} at %D the moment we encounter it. \def\handleMPpathconcat {\presetMPconcat \PDFcode{\gMPs1 \gMPs2 \gMPs3 \gMPs4 \gMPs5 \gMPs6 cm}% \resetMPstack} \def\handleMPpathscale {\presetMPscale \PDFcode{\gMPs1 0 0 \gMPs2 0 0 cm}% \resetMPstack} %D This macro interprets the path and saves it as compact as %D possible. \def\dohandleMPpath#1% {\ifcase\lccode`#1\relax \@EA\dohandleMPpathA \else \@EA\dohandleMPpathB \fi#1} %\def\dohandleMPpathA#1 % % {\setMPargument{#1}% % \handleMPsequence} \let\dohandleMPpathA\setMPsequence \def\dohandleMPpathB#1 % {\def\somestring{#1}% \ifx\somestring\PSlineto \setMPkeyword0 \else\ifx\somestring\PScurveto \setMPkeyword1 \else\ifx\somestring\PSrlineto \setMPkeyword2 \else\ifx\somestring\PSmoveto \setMPkeyword3 \else\ifx\somestring\PSclip % \chardef\finiMPpath0 % already \let\handleMPsequence\processMPpath \else\ifx\somestring\PSgsave \chardef\finiMPpath3 \else\ifx\somestring\PSgrestore \else\ifx\somestring\PSfill \ifcase\finiMPpath \chardef\finiMPpath2 \let\handleMPsequence\processMPpath \fi \else\ifx\somestring\PSstroke \ifcase\finiMPpath \chardef\finiMPpath1 \fi \let\handleMPsequence\processMPpath \else\ifx\somestring\PSclosepath \def\closeMPpath{\PDFcode{h}}% \else\ifx\somestring\PSconcat \let\flushMPpath\flushconcatMPpath \handleMPpathconcat \else\ifx\somestring\PSscale \let\flushMPpath\flushconcatMPpath \handleMPpathscale \fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi\fi \handleMPsequence} %D The main conversion command is: %D %D \starttyping %D \convertMPtoPDF {filename} {x scale} {y scale} %D \stoptyping %D %D The dimensions are derived from the bounding box. So we %D only have to say: %D %D \starttyping %D \convertMPtoPDF{mp-pra-1.eps}{1}{1} %D \convertMPtoPDF{mp-pra-1.eps}{.5}{.5} %D \stoptyping %D \macros %D {makeMPintoPDFobject,lastPDFMPobject} %D %D For experts there are a few more options. When attributes %D are to be added, the code must be embedded in an object %D accompanied with the appropriate directives. One can %D influence this process with \type {\makeMPintoPDFobject}. %D %D This option defaults to~0, because \CONTEXT\ takes care %D of objects at another level, which saves some bytes. %D %D \starttabulate[|l|l|p|] %D \NC 0 \NC never \NC don't use an object \NC\NR %D \NC 1 \NC always \NC always use an object \NC\NR %D \NC 2 \NC optional \NC use object when needed \NC\NR %D \stoptabulate %D %D The last object number used is avaliable in the macro %D \type {\lastPDFMPobject}. \ifx\makeMPintoPDFobject\undefined \chardef\makeMPintoPDFobject=0 \fi \def\lastPDFMPobject{0} %D The additional code needed can be made available in the %D (global) macro \type {\currentPDFresources}. \let\currentPDFresources\empty \newtoks\everyMPtoPDFconversion \def\convertMPtoPDF % #1#2#3% {\bgroup \ifx\pdfdecimaldigits\undefined\else \pdfdecimaldigits=5 \fi % new \setbox\scratchbox\vbox\bgroup \forgetall \offinterlineskip \startMPresources \doprocessMPtoPDFfile} % %D The next one is kind of private and probably will become obsolete): \def\processMPtoPDFfile % file xscale yscale {\bgroup \let\finishMPgraphic\egroup \doprocessMPtoPDFfile} \def\doprocessMPtoPDFfile#1#2#3% file xscale yscale {% the following line is needed for latex where onepoint is not % onepoint but a number (maxdimen); some day i'll make a latex % variant of this file so that i no longer have to deal with such % issues; then i'll also speed up this module using a few context % tricks % \let\onepoint\onerealpoint % \setMPspecials \the\everyMPtoPDFconversion \catcode`\^^M=\@@endofline \startMPscanning \let\do\empty \xdef\MPxscale{#2}% \xdef\MPyscale{#3}% \xdef\MPxoffset{0}% \xdef\MPyoffset{0}% \xdef\MPyshift{\zeropoint}% \donefalse \let\handleMPsequence\dohandleMPsequence \message{[MP to PDF]}% was: [MP to PDF #1] but there is a (#1) anyway \input#1\relax} % strange rounding/clip in pdftex/viewer % % \def\finishMPgraphic % {\stopMPresources % \egroup % \ifx\pdftexversion\undefined\else\ifnum\pdftexversion<14 % for the moment % \chardef\makeMPintoPDFobject=0 % \fi\fi % \ifcase\makeMPintoPDFobject\or\or\ifx\currentPDFresources\empty\else % \chardef\makeMPintoPDFobject=1 % \fi\fi % \setbox\scratchbox=\vbox % {\forgetall % \hbox % {\PDFcode{q \MPxscale\space 0 0 \MPyscale\space \MPxoffset\space \MPyoffset\space cm}% % \lower\MPyshift\box\scratchbox % unscaled shift % \PDFcode{Q}}}% % \ht\scratchbox\MPheight % \wd\scratchbox\MPwidth % \dp\scratchbox\zeropoint % \ifcase\makeMPintoPDFobject % \box\scratchbox % \or % \immediate\pdfxform resources{\currentPDFresources}\scratchbox % \xdef\lastPDFMPobject{\the\pdflastxform}% % \pdfrefxform\lastPDFMPobject % \global\let\currentPDFresources\empty % \else % \box\scratchbox % \fi % \egroup} % % funny clip in viewer % % \setbox\scratchbox=\vbox % {\forgetall % \dimen0=\MPllx bp % \dimen2=\MPlly bp % \setbox\scratchbox=\hbox{\hskip-\dimen0\raise-\dimen2\box\scratchbox}% % \ht\scratchbox=\zeropoint % \dp\scratchbox=\zeropoint % \wd\scratchbox=\zeropoint % \hbox % {\PDFcode{q \MPxscale\space 0 0 \MPyscale\space 0 0 cm}% % \lower\MPshift\box\scratchbox % \PDFcode{Q}}}% % \let\PDFMPformoffset\zeropoint \def\PDFMPformoffset {\ifx\objectoffset\undefined\zeropoint\else\objectoffset\fi} \def\finishMPgraphic {\stopMPresources \egroup \setbox\scratchbox\vbox {\forgetall \hbox {\PDFcode{q \MPxscale\space 0 0 \MPyscale\space \MPxoffset\space \MPyoffset\space cm}% \lower\MPyshift\box\scratchbox % unscaled shift \PDFcode{Q}}}% \ht\scratchbox\MPheight \wd\scratchbox\MPwidth \dp\scratchbox\zeropoint\relax \dopackageMPgraphic\scratchbox \egroup \endinput} %D Alternative for \PDFTEX. We cannot come up with something more contexy %D because this module is also used in \LATEX. \def\dopackageMPgraphic#1% #1 = boxregister {%\ifx\pdfxform\undefined % \chardef\makeMPintoPDFobject\zerocount % no pdftex at all %\else\ifx\pdftexversion\undefined % \chardef\makeMPintoPDFobject\zerocount % no pdftex at all %\else\ifnum\pdftexversion<14 % \chardef\makeMPintoPDFobject\zerocount % no resource support %\else % % keep the default value %\fi\fi\fi \ifcase\makeMPintoPDFobject\or\or\ifx\currentPDFresources\empty\else % an existing value of 2 signals object support (set elsewhere) \chardef\makeMPintoPDFobject\plusone \fi\fi \ifcase\makeMPintoPDFobject \box#1% \or \scratchdimen\PDFMPformoffset\relax \ifdim\scratchdimen>\zeropoint % compensate for error \setbox#1\vbox spread 2\scratchdimen {\forgetall\vss\hbox spread 2\scratchdimen{\hss\box#1\hss}\vss}% \fi \setMPPDFobject{\currentPDFresources}{#1}% \ifdim\scratchdimen>\zeropoint % compensate for error \vbox to \MPheight {\forgetall\vss\hbox to \MPwidth{\hss\getMPPDFobject\hss}\vss}% \else \getMPPDFobject \fi \global\let\currentPDFresources\empty \else \box#1% \fi} \def\setMPPDFobject#1#2% resources boxnumber {\ifx\pdfxform\undefined \def\getMPPDFobject{\box#2}% \else\ifx\pdftexversion\undefined \def\getMPPDFobject{\box#2}% \else\ifnum\pdftexversion<14 \def\getMPPDFobject{\box#2}% \else \immediate\pdfxform resources{#1}#2% \edef\getMPPDFobject{\noexpand\pdfrefxform\the\pdflastxform}% \fi\fi\fi} \let\getMPPDFobject\relax %D \macros %D {deleteMPgraphic, %D startMPresources, %D stopMPresources} %D %D Here are a few hooks for \CONTEXT\ specific things. \ifx\deleteMPgraphic\undefined \def\deleteMPgraphic#1{} \fi \ifx\startMPresources\undefined \let\startMPresources\relax \let\stopMPresources\relax \fi %D \macros %D {twodigitMPoutput} %D %D We can limit the precision to two digits after the comma %D by saying: %D %D \starttyping %D \twodigitMPoutput %D \stoptyping %D %D This option only works in \CONTEXT\ combined with \ETEX. \def\twodigitMPoutput {\let\!MP \twodigitrounding \def\!MPgMPs##1{\twodigitrounding{\gMPs##1}}% \def\!MPgMPa##1{\twodigitrounding{\gMPa##1}}} \let\!MP \empty \let\!MPgMPa\gMPa \let\!MPgMPs\gMPs %D This kind of conversion is possible because \METAPOST\ %D does all the calculations. Converting other \POSTSCRIPT\ %D files would drive both me and \TEX\ crazy. \ifx\undefined\StopLatexHack \else \StopLatexHack \fi \protect \endinput