summaryrefslogtreecommitdiff
path: root/macros/latex/contrib/dialogl/dialogl.dtx
diff options
context:
space:
mode:
authorNorbert Preining <norbert@preining.info>2019-09-02 13:46:59 +0900
committerNorbert Preining <norbert@preining.info>2019-09-02 13:46:59 +0900
commite0c6872cf40896c7be36b11dcc744620f10adf1d (patch)
tree60335e10d2f4354b0674ec22d7b53f0f8abee672 /macros/latex/contrib/dialogl/dialogl.dtx
Initial commit
Diffstat (limited to 'macros/latex/contrib/dialogl/dialogl.dtx')
-rw-r--r--macros/latex/contrib/dialogl/dialogl.dtx1429
1 files changed, 1429 insertions, 0 deletions
diff --git a/macros/latex/contrib/dialogl/dialogl.dtx b/macros/latex/contrib/dialogl/dialogl.dtx
new file mode 100644
index 0000000000..105a2494dd
--- /dev/null
+++ b/macros/latex/contrib/dialogl/dialogl.dtx
@@ -0,0 +1,1429 @@
+%%% ====================================================================
+%%% @LaTeX-style-file{
+%%% filename = "dialogl.dtx",
+%%% version = "1.99a",
+%%% date = "2013/01/24",
+%%% author = "Michael Downes",
+%%% copyright = "This file is part of the dialogl package, released
+%%% under the LPPL; see dialogl.ins for details."
+%%% keywords = "TeX, dialog",
+%%% supported = "no",
+%%% abstract = "This file provides macros for writing
+%%% messages and menus on screen, and reading user responses. It
+%%% can be used with LaTeX as a documentstyle option, or in
+%%% other forms of TeX by a standard \input statement.",
+%%% }
+%%% ====================================================================
+%
+% \iffalse
+%<*driver>
+\input{dia-driv.tex}
+%</driver>
+% \fi
+%
+% \section{History}
+% This file, \fn{dialog.sty}, was born out of a utility called
+% \fn{listout.tex} that I wrote for my personal use. The purpose of
+% \fn{listout.tex} was to facilitate printing out plain text
+% files\Dash electronic mail, program source files in various
+% programming languages, and, foremost, \tex/ macro files and log
+% files. An important part of my \tex/ programming practice is to
+% print out a macro file on paper and read it through, marking
+% corrections along the way, then use the marked copy as a script
+% for editing the file. (For one thing, this allows me to analyze
+% and mark corrections while riding the bus to work, or
+% sitting out in the back yard to supervise the kids.) The output I
+% normally desired was two `pages' per sheet of U.S. letter-size
+% paper printed landscape, in order to conserve paper.
+%
+% Once created, \fn{listout.tex} quickly became my favorite means
+% of printing out plain text files, not to mention an indispensable
+% tool in my debugging toolbox: I turn on \cw{tracingmacros} and
+% \cw{tracingcommands}, then print out the resulting log file so
+% that I can see several hundred lines of the log at once (by
+% spreading out two or three pages on my desk with 100+ lines per
+% page); then I trace through, cross things out, label other
+% things, draw arrows, and so forth.
+%
+% I soon added a filename prompting loop to make it convenient to
+% print multiple files in a single run. In the process of
+% perfecting this simple prompting routine\Dash over two or three
+% years\Dash and adding the ability to optionally specify things like
+% number of columns at run time, eventually I wrote so much
+% dialog-related macro code that it became clear this code should
+% be moved out of \fn{listout.tex} into its own module. The result
+% was \fn{dialog.sty}.
+%
+% Before getting into the macro definitions and technical
+% commentary, here are descriptions from the user's
+% perspective of the functions defined in this file.
+%
+% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Message-sending functions}
+%
+%\begin{usage}
+%\mesj{<text>}
+%\end{usage}
+% Sends the message verbatim: category 12 for all special
+% characters except braces, tab characters, and carriage returns:
+%
+%^^V { } ^^I ^^M
+%
+% Naturally, the catcode changes are effective only if \cw{mesj} is
+% used directly, not inside a macro argument or definition
+% replacement text.
+%
+% Multiple spaces in the argument of \cw{mesj} print as multiple
+% spaces on screen. A tab character produces always eight spaces;
+% `smart' handling of tabs is more complicated than I would care to
+% attempt.
+%
+% Line breaks in the argument of \cw{mesj} will produce line breaks
+% on screen. That is, you don't need to enter a special sequence
+% such as \ctrl{J}\qc{\%} to get line breaks. See the technical
+% commentary for \cw{mesjsetup} for details. Even though curly
+% braces are left with their normal catcodes, they can be printed
+% in a message without any problem, if they occur in balanced
+% pairs. If not, the message should be sent using \cw{xmesj}
+% instead of \cw{mesj}.
+%
+% Because of its careful handling of the message text, \cw{mesj} is
+% extremely easy to use. The only thing you have to worry about is
+% having properly matched braces. Beyond that, you simply type
+% everything exactly as you want it to appear on screen.
+%
+%\begin{usage}
+%\xmesj{<text>}
+%\end{usage}
+% This is like \cw{mesj} but expands embedded control sequences
+% instead of printing them verbatim. All special characters have
+% category 12 except backslash, percent, braces, tab, return, and
+% ampersand:
+%
+%^^V \ % { } ^^I ^^M &
+%
+% The first four have normal \tex/ catcodes to make it possible to
+% use most normal \tex/ commands, and comments, in the message
+% text. \ctrl{I} and \ctrl{M} are catcode 13 and behave as
+% described for \cw{mesj}. The \qc{\&} is a special convenience, an
+% abbreviation for \cw{noexpand}, to use for controlling expansion
+% inside the message text.
+%
+% Doubled backslash \cs{\\} in the argument will produce a single
+% category 12 backslash character\Dash thus, \verb'\\xxx' can be used
+% instead of \cw{string}\cw{xxx} or \cw{noexpand}\cw{xxx} (notice
+% that this works even for outer things like \cw{bye} or
+% \cw{newif}). Similarly \cs{\%}, \cs{\{}, \cs{\}} and \cs{\&}
+% produce the corresponding single characters.
+%
+% Category 12 space means that you cannot write something like
+%
+%^^V \ifvmode h\else v\fi rule
+%
+% in the argument of \cw{xmesj} without getting a space after the
+% \cw{ifvmode}, \cw{else}, and \cw{fi}.%^^A
+%^^A
+%\footnote{Well, actually, you could replace each space by
+%\qc{\%}\meta{newline} to get rid of it. But that makes the message
+%text harder to read for the programmer.}
+%^^A
+% Since occasionally this may be troublesome, \cs{\.} is defined
+% inside the argument of \cw{xmesj} to be a `control word
+% terminator': If the expansion of \cw{foo} is \verb'abc', then
+% \verb'\foo\.xyz' produces \verb'abcxyz' on screen (as opposed to
+% \verb'\foo xyz' which would produce \verb'abc xyz'). Thus the
+% above conditional could be written as
+%
+%^^V \ifvmode\.h\else\.v\fi\.rule
+%
+% Even though the catcode changes done by \cw{xmesj} setup have no
+% effect if \cw{xmesj} is used inside an argument or definition
+% replacement text, I find it convenient occasionally to use
+% \cw{xmesj} in those contexts, in order to get other aspects of the
+% \cw{xmesj} setup. For instance, if you need to embed a message
+% that contains a percent sign inside a definition, you can write
+%
+%^^V \def\foo{...
+%^^V \xmesj{... this is a percent
+%^^V sign: \% (sans backslash) ...}
+%^^V ...}
+%
+% To further support such uses of \cw{xmesj}, the following changes
+% are also done by \cw{xmesj} setup: the backslash-space control
+% symbol {\tt\bslash\char32} is made equivalent to \cw{space};
+% \cs{\^^J} and \cs{\^^M} are defined to produce a \cw{newlinechar};
+% and active tilde \qc{\~} will produce a category-12 tilde.
+%
+% Among other things, this setup makes it easier to obtain
+% newlines and multiple spaces in an embedded message. For example,
+% in the following definition the message will have a line break on
+% screen for each backslash at the end of a line, and the third
+% line will be indented four spaces.
+%
+%^^V \def\bar{...
+%^^V \xmesj{First line\
+%^^V Second line\
+%^^V \ \ \ \ Indented line\
+%^^V Last line}%
+%^^V ...}
+%
+% The alternative of defining a separate message function
+% \cw{barfoo} with \cw{f[x]mesj} and calling \cw{barfoo} inside of
+% \cw{bar} would allow more natural entry of the newlines and the
+% multiple spaces, but would be slightly more expensive in string
+% pool and hash table usage.
+%
+%\begin{usage}
+%\promptmesj{<text>}
+%\promptxmesj{<text>}
+%\end{usage}
+% These are like \cw{mesj}, \cw{xmesj} but use \cw{message} rather
+% than \cw{immediate}\cw{write}\verb'16' internally, thus if the
+% following operation is a \cw{read}, the user will see the cursor
+% on screen at the end of the last line, as may be desired when
+% prompting for a short reply, rather than at the beginning of the
+% next line. The character \qc{\!} is preempted internally for
+% newlinechar, for these two functions only,
+% which means that it cannot be actually printed in
+% the message text. Use of a visible character such as \qc{\!},
+% rather than the normal \cw{newlinechar} \ctrl{J}, is necessary
+% for robustness because of the fact that the \cw{message}
+% primitive was unable to use an `invisible' character (outside the
+% range 32--126) for newlines up until \tex/ version 3.1415, which
+% some users do not yet have (at the time of this writing\Dash July
+% 1994).
+%
+%\begin{usage}
+%\storemesj\foo{<text>}
+%\storexmesj\foo{<text>}
+%\end{usage}
+% These functions are similar to \cw{mesj}, \cw{xmesj} but store the
+% given text in the control sequence \cw{foo} instead of immediately
+% sending the message. Standard \tex/ parameter syntax can be used
+% to make \cw{foo} a function with arguments, e.g. after
+%
+%^^V \storemesj\foo#1{...#1...}
+%
+% then you can later write
+%
+%^^V \message{\foo{\the\hsize}}
+%
+% and get the current value of \cw{hsize} into the middle of the
+% message text. Consequently also in the x-version \cw{storexmesj} a
+% category-12 \qc{\#} character can be obtained with \cs{\#}.
+%
+%\begin{usage}
+%\fmesj\foobar#1#2...{...#1...#2...}
+%\end{usage}
+% Defines \cw{foobar} as a function that will take the given
+% arguments, sow them into the message text \verb"{...}", and send
+% the message. In the message text all special characters are
+% category 12 except for braces, \qc{\#}, tab, and carriage return.
+%
+% If an unmatched brace or a \qc{\#} must be printed in the message
+% text \cw{fxmesj} must be used instead. (\arg{#} could be used to
+% insert a single category-6 \qc{\#} token into the message text, and
+% \tex/ would print it without an error, but both \cw{message} and
+% \cw{write} would print it as two \arg{#} characters, even though
+% it's only a single token internally.)
+%
+%\begin{usage}
+%\fxmesj\foobar#1#2...{...#1...#2...}
+%\end{usage}
+% Combination of \cw{xmesj} and \cw{fmesj}. Defines \cw{foobar} like
+% \cw{fmesj}, but with full expansion of the replacement text and
+% with normal category codes for backslash, percent, braces, and hash
+% \qc{\#}. The control symbols \cs{\\} \cs{\%} \cs{\{} \cs{\}}
+% \cs{\&} and \cs{\.} can be used as in \cw{xmesj}, with also \cs{\#}
+% for printing a \qc{\#} character of category 12.
+%
+% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Reading functions}
+%
+%\begin{usage}
+%\readline{<default>}\answer
+%\end{usage}
+% This reads a line of input from the user into the macro
+% \cw{answer}. (The macro name can be anything chosen by the
+% programmer, not just \cw{answer}.) Before reading, all special
+% characters are deactivated, so that the primitive \cw{read} will
+% not choke if the user happens to enter something like \cw{newif}
+% or {\sc control-l} or \qc{\}}. Depending on the operating
+% system, certain characters\Dash e.g., {\sc control-c}, {\sc
+% control-z}, {\sc control-d}, {\sc control-h}\Dash might have
+% special effects instead of being entered into the replacement
+% text of \cw{answer}, regardless of the catcode changes. To
+% take the most obvious example, under most
+% operating systems, typing {\sc control-h} (the Rubout or
+% Backward-Delete key) will delete the previous character from the
+% user's response, instead of entering an \ascii/ character 8
+% into \cw{answer}.
+%
+% There is one significant exception from the catcode changes that
+% are done for \cw{readline}: spaces and tabs retain their normal
+% catcode of 10, so that multiple spaces in an answer will be
+% reduced to a single space, and macros with normal space-delimited
+% arguments will work when applied to the answer. (I can't think of
+% any likely scenario where category 12 for spaces would be
+% useful.) Also, the catcode of \ctrl{M} is set to 9 (ignore) so
+% that an empty line\Dash meaning that the user just pressed the
+% carriage return/enter key\Dash will result in an empty \cw{answer}.
+% If the answer is empty, the given default string will be
+% substituted. The default string can be empty.
+%
+%\begin{usage}
+%\xreadline{<default>}\answer
+%\end{usage}
+% Like \cw{readline} but the answer is read as executable tokens;
+% the usual catcodes of the \tex/ special characters remain in
+% effect while reading the answer. A few common outer things
+% (\cw{bye}, \cs{\+}, \cw{newif}, \ctrl{L}, among others) are
+% neutralized before the \cw{read} is done, but the user can still
+% cause problems by entering some other outer control sequence or
+% unbalanced braces. I doubt there's any bulletproof solution, if
+% the tokens are to remain executable, short of the usual last
+% resort: reading the answer using \cw{readline}, writing it to a
+% file, then inputting the file.
+%
+%\begin{usage}
+%\readchar{<default>}\answer
+%\end{usage}
+% This is like \cw{readline} but it reduces the answer to its first
+% character. \meta{default} is either a single character or empty.
+%
+%\begin{usage}
+%\readChar{<default>}\answer
+%\end{usage}
+% This is like \cw{readchar} and also uppercases the answer.
+%
+%\begin{usage}
+%\changecase\uppercase\answer
+%\end{usage}
+% The function \cw{changecase} redefines its second argument, which
+% must be a macro, to contain the same text as before, but
+% uppercased or lowercased according to the first argument. Thus
+% \cw{readChar}\verb'{Q}'\cw{answer} is equivalent to
+%
+%^^V \readchar{q}\answer
+%^^V \changecase\uppercase\answer
+%
+% It might sometimes be desirable to force lower case before using
+% a file name given by the user, for example.
+%
+% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Checking functions}
+%
+%\begin{usage}
+%\checkinteger\reply\tempcount
+%\end{usage}
+% To read in and check an answer that is supposed to be an integer,
+% use \cw{readline}{}\cw{reply} and then apply \cw{checkinteger} to
+% the \cw{reply}, supplying a count register, not necessarily named
+% \cw{tempcount}, wherein \cw{checkinteger} will leave the validated
+% integer. If \cw{reply} does not contain a valid integer the
+% returned value will be \verb'-'\cw{maxdimen}.
+%
+% At the present time only decimal digits are handled;
+% some valid \tex/ numbers such as \verb'"AB', \verb'`\@',
+% \cw{number}\cw{prevgraf}, or a count register name, will not be
+% recognized by \cw{checkinteger}. There seems to be no bulletproof
+% way to allow these possibilities.
+%
+% Tests that hide \cw{checkinteger} under the hood, such as a
+% \cw{nonnegativeinteger} test, are not provided because as often as
+% not the number being prompted for will have to be tested to see if
+% it falls inside a more specific range, such as 0\dash 255 for an
+% 8-bit number or 1\dash 31 for a date, and it seems common sense to
+% omit overhead if it would usually be redundant. It's easy enough to
+% define such a test for yourself, if you want one.
+%
+%\begin{usage}
+%\checkdimen\reply\tempdim
+%\end{usage}
+% Analog of \cw{checkinteger} for dimension values. If \cw{reply}
+% does not contain a valid dimension the value returned in
+% \cw{tempdim} will be \verb'-'\cw{maxdimen}.
+%
+% Only explicit dimensions with decimal digits, optional decimal
+% point and more decimal digits, followed by explicit units
+% \verb'pt' \verb'cm' \verb'in' or whatever are checked for; some
+% valid \tex/ dimensions such as \cw{parindent},
+% \verb'.3'\cw{baselineskip}, or \cw{fontdimen}\verb'5'\cw{font}
+% will not be recognized by \cw{checkdimen}.
+%
+% \subsection*{What good is all this?}
+% What good is all this stuff, practically speaking?\Dash you may
+% ask. Well, a typical application might be something like: At the
+% beginning of a document, prompt interactively to find out if the
+% user wants to print on A4 or US letter-size paper, or change the
+% top or left margin. Such a query could be done like this:
+%
+%^^V \promptxmesj{
+%^^V Do you want to print on A4 or US letter paper?
+%^^V Enter u or U for US letter, anything else for A4: }
+%^^V \readChar{A}\reply % default = A4
+%^^V \if U\reply \textheight=11in \textwidth=8.5in
+%^^V \else \textheight=297mm \textwidth=210mm \fi
+%^^V % Subtract space for 1-inch margins
+%^^V \addtolength{\textheight}{-2in}
+%^^V \addtolength{\textwidth}{-2in}
+%^^V
+%^^V \promptxmesj{
+%^^V Left margin setting? [Return = keep current value,
+%^^V \the\oddsidemargin]: }
+%^^V \readline{\the\oddsidemargin}\reply
+%^^V \checkdimen\reply{\dimen0}
+%^^V \ifdim\dimen0>-\maxdimen
+%^^V \setlength\oddsidemargin{\dimen0}%
+%^^V \xmesj{OK, using new left margin of %
+%^^V \the\oddsidemargin.}
+%^^V \else
+%^^V \xmesj{Sorry, I don't understand %
+%^^V that reply: `\reply'.\
+%^^V Using default value: \the\oddsidemargin.}
+%^^V \fi
+%
+% Although \latex/'s \cw{typeout} and \cw{typein} functions can
+% be used for this same task, they are rather more awkward, and
+% checking the margin value for validity would be quite difficult.
+%
+% \StopEventually{}
+%
+% \section{Implementation}
+% Standard package identification:
+% \begin{macrocode}
+%<*2e>
+\NeedsTeXFormat{LaTeX2e}
+\ProvidesPackage{dialog}[1994/11/08 v0.9y]
+%</2e>
+% \end{macrocode}
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Preliminaries}
+%
+% \begin{macrocode}
+%<*2e>
+\RequirePackage{grabhedr}
+%</2e>
+% \end{macrocode}
+%
+% If \fn{grabhedr.sty} is not already loaded, load it now. The
+% \cw{trap.input} function is explained in \fn{grabhedr.doc}.
+% \begin{macrocode}
+%<*209>
+\csname trap.input\endcsname
+\input grabhedr.sty \relax
+\fileversiondate{dialog.sty}{0.9y}{1994/11/08}%
+%</209>
+% \end{macrocode}
+%
+% The functions \cw{localcatcodes} and \cw{restore\-catcodes} are
+% defined in \fn{grabhedr.sty}. We use them to save and restore
+% catcodes of any special characters needed in this file whose
+% current catcodes might not be what we want them to be. Saving and
+% restoring catcode of at-sign \qc{\@} makes this file work equally
+% well as a \latex/ documentstyle option or as a simple input file in
+% other contexts. The double quote character \qc{\"} might be active
+% for German and other languages. Saving and restoring tilde \qc{\~},
+% hash \qc{\#}, caret \qc{\^}, and left quote \qc{\`} catcodes is
+% normally redundant but reduces the number of assumptions we must
+% rely on. (The following catcodes are assumed: \qc{\\} 0, \qc{\{} 1,
+% \qc{\}} 2, \qc{\%} 14, {\tt a}\dash {\tt z} {\tt A}\dash {\tt Z}
+% 11, {\tt 0}\dash {\tt 9} \qc{\.} \qc{\-} 12. Also note that
+% \cw{endlinechar} is assumed to have a non-null value.)
+% \begin{macrocode}
+%% The line break is significant here:
+\localcatcodes{\@{11}\ {10}\
+{5}\~{13}\"{12}\#{6}\^{7}\`{12}}
+% \end{macrocode}
+%
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Definitions}
+%
+% \begin{macro}{\otherchars}
+%
+% For deactivating characters with special catcodes during
+% \cw{readline} we use, instead of \cw{dospecials}, a more
+% bulletproof (albeit slower) combination of \cw{otherchars},
+% \cw{controlchars}, and \cw{highchars} that covers all characters in
+% the range 0\dash 255 except letters and digits. Handling the
+% characters above 127 triples the overhead done for each read
+% operation or message definition but seems mandatory for maximum
+% robustness.^^A
+%^^A%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \footnote{If you are using \fn{dialog.sty} functions on a slow
+% computer, you might want to try setting \cw{highchars} = empty to
+% see if that helps the speed.}
+%^^A%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%
+% \cw{otherchars} includes the thirty-three nonalphanumeric visible
+% characters (counting space as visible). It is intended as an
+% executable list like \cw{dospecials} but, as an exercise in
+% memory conservation, it is constructed without the \cw{do}s. For
+% the usual application of changing catcodes, the list can still be
+% executed nicely as shown below. Also, if we arrange to
+% make sure that each character token gets category 12, it's not
+% necessary to use control symbols such as \cs{\%} in place of
+% those few special characters that would otherwise be difficult to
+% place inside of a definition. This avoids a problem that would
+% otherwise arise if we included \cs{\+} in the list and tried to
+% process the list with a typical definition of \cs{do}: \cs{\+} is
+% `outer' in plain \tex/ and would cause an error message when
+% \cw{do} attempted to read it as an argument. (As a matter of fact
+% the catcode changes below show a different way around that
+% problem, but a list of category-12 character tokens is a fun
+% thing to have around anyway.)
+%
+% \begin{macrocode}
+\begingroup
+% \end{macrocode}
+% First we start a group to localize \cw{catcode} changes. Then we
+% change all relevant catcodes to 12 except for backslash, open
+% brace, and close brace, which can be handled by judicious
+% application of \cw{escapechar}, \cw{string}, \cw{edef}, and
+% \cw{xdef}. By defining \cw{do} in a slightly backward way, so that
+% it doesn't take an argument, we don't need to worry about the
+% presence of \cs{\+} in the list of control symbols. Notice the
+% absence of \cs{\`} from the list of control symbols; it was already
+% catcoded to 12 in the \cw{localcatcodes} declaration at the
+% beginning of this file\Dash otherwise it would be troublesome to
+% make the definition of \cw{do} bulletproof (consider the
+% possibilities that \qc{\`} might have catcode 0, 5, 9, or 14).
+%
+% \begin{macrocode}
+\def\do{12 \catcode`}
+\catcode`\~\do\!\do\@\do\#\do\$\do\^\do\&
+\do\*\do\(\do\)\do\-\do\_\do\=\do\[\do\]
+\do\;\do\:\do\'\do\"\do\<\do\>\do\,\do\.
+\do\/\do\?\do\|12\relax
+% \end{macrocode}
+% To handle backslash and braces, we define \cs{\\},
+% \cs{\{}, and \cs{\}} to produce the corresponding category-12
+% character tokens. Setting \cw{escapechar} to $-1$ means that
+% \cw{string} will omit the leading backslash that it would
+% otherwise produce when applied to a control sequence.
+% \begin{macrocode}
+\escapechar -1
+\edef\\{\string\\}
+\edef\{{\string\{}\edef\}{\string\}}
+% \end{macrocode}
+% Space and percent are done last. Then, with almost all the
+% special characters now category 12, it's rather easy to define
+% \cw{otherchars}.
+% \begin{macrocode}
+\catcode`\ =12\catcode`\%=12
+\xdef\otherchars
+{ !"#$%&'()*+,-./:;<=>?[\\]^_`\{|\}~}
+\endgroup % ^ ^ ^
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\controlchars}
+%
+% \cw{controlchars} is another list for the control characters
+% \ascii/ 0\dash 31 and 127. The construction of this list is similar
+% to the construction of \cw{otherchars}. We need to turn off
+% \cw{endlinechar} because the catcode of \ctrl{M} is going to be
+% changed. The \ctrl{L} inside the \cw{gdef} is not a problem (as it
+% might have been, due to the usual outerness of \ctrl{L}) because
+% the catcode is changed from 13 to 12 before that point.
+%
+% \begin{macrocode}
+\begingroup
+\endlinechar = -1
+\def\do{12 \catcode`}
+\catcode`\^^@\do\^^A\do\^^B\do\^^C
+\do\^^D\do\^^E\do\^^F\do\^^G\do\^^H\do\^^I
+\do\^^J\do\^^K\do\^^L\do\^^M\do\^^N\do\^^O
+\do\^^P\do\^^Q\do\^^R\do\^^S\do\^^T\do\^^U
+\do\^^V\do\^^W\do\^^X\do\^^Y\do\^^Z\do\^^[
+\do\^^\\do\^^]\do\^^^\do\^^_\do\^^? 12\relax
+%
+\gdef\controlchars{^^@^^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^^[^^\^^]^^^^^_^^?}
+\endgroup
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\highchars}
+% And finally, the list \cw{highchars} contains characters
+% 128\dash 255, the ones that have the eighth bit set.
+%
+% \begin{macrocode}
+\begingroup
+\def\do{12 \catcode`}
+\catcode`\^^80\do\^^81\do\^^82\do\^^83\do\^^84
+\do\^^85\do\^^86\do\^^87\do\^^88\do\^^89\do\^^8a
+\do\^^8b\do\^^8c\do\^^8d\do\^^8e\do\^^8f
+\do\^^90\do\^^91\do\^^92\do\^^93\do\^^94\do\^^95
+% \end{macrocode}
+%\verbdots\iffalse
+\do\^^96\do\^^97\do\^^98\do\^^99\do\^^9a\do\^^9b
+\do\^^9c\do\^^9d\do\^^9e\do\^^9f
+\do\^^a0\do\^^a1\do\^^a2\do\^^a3\do\^^a4\do\^^a5
+\do\^^a6\do\^^a7\do\^^a8\do\^^a9\do\^^aa\do\^^ab
+\do\^^ac\do\^^ad\do\^^ae\do\^^af
+\do\^^b0\do\^^b1\do\^^b2\do\^^b3\do\^^b4\do\^^b5
+\do\^^b6\do\^^b7\do\^^b8\do\^^b9\do\^^ba\do\^^bb
+\do\^^bc\do\^^bd\do\^^be\do\^^bf
+\do\^^c0\do\^^c1\do\^^c2\do\^^c3\do\^^c4\do\^^c5
+\do\^^c6\do\^^c7\do\^^c8\do\^^c9\do\^^ca\do\^^cb
+\do\^^cc\do\^^cd\do\^^ce\do\^^cf
+\do\^^d0\do\^^d1\do\^^d2\do\^^d3\do\^^d4\do\^^d5
+\do\^^d6\do\^^d7\do\^^d8\do\^^d9\do\^^da\do\^^db
+\do\^^dc\do\^^dd\do\^^de\do\^^df
+\do\^^e0\do\^^e1\do\^^e2\do\^^e3\do\^^e4\do\^^e5
+\do\^^e6\do\^^e7\do\^^e8\do\^^e9\do\^^ea\do\^^eb
+\do\^^ec\do\^^ed\do\^^ee\do\^^ef
+\do\^^f0\do\^^f1\do\^^f2\do\^^f3\do\^^f4\do\^^f5
+\do\^^f6\do\^^f7\do\^^f8\do\^^f9\do\^^fa\do\^^fb
+%\fi
+% \begin{macrocode}
+\do\^^fc\do\^^fd\do\^^fe\do\^^ff 12\relax
+%
+\gdef\highchars{%
+^^80^^81^^82^^83^^84^^85^^86^^87^^88%
+^^89^^8a^^8b^^8c^^8d^^8e^^8f%
+^^90^^91^^92^^93^^94^^95^^96^^97^^98%
+% \end{macrocode}
+%\verbdots\iffalse
+^^99^^9a^^9b^^9c^^9d^^9e^^9f%
+^^a0^^a1^^a2^^a3^^a4^^a5^^a6^^a7^^a8%
+^^a9^^aa^^ab^^ac^^ad^^ae^^af%
+^^b0^^b1^^b2^^b3^^b4^^b5^^b6^^b7^^b8%
+^^b9^^ba^^bb^^bc^^bd^^be^^bf%
+^^c0^^c1^^c2^^c3^^c4^^c5^^c6^^c7^^c8%
+^^c9^^ca^^cb^^cc^^cd^^ce^^cf%
+^^d0^^d1^^d2^^d3^^d4^^d5^^d6^^d7^^d8%
+^^d9^^da^^db^^dc^^dd^^de^^df%
+^^e0^^e1^^e2^^e3^^e4^^e5^^e6^^e7^^e8%
+^^e9^^ea^^eb^^ec^^ed^^ee^^ef%
+^^f0^^f1^^f2^^f3^^f4^^f5^^f6^^f7^^f8%
+%\fi
+% \begin{macrocode}
+^^f9^^fa^^fb^^fc^^fd^^fe^^ff}
+\endgroup
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\actively}
+%
+% The function \cw{actively} makes a given character active and
+% carries out the assignment given as the first argument. The
+% assignment can be embedded in the replacement text of a macro
+% definition without requiring any special setup to produce an active
+% character in the replacement text. The argument should be a control
+% symbol, e.g. \cw{@} or \cs{\#} or \cs{\^^M}, rather than a single
+% character. (Except that \qc{\+} is safer than \cs{\+} in
+% \plaintex/.) If the assignment is a definition (\cw{def},
+% \cw{edef}, \cw{gdef}, \cw{xdef}) it is allowed to take arguments in
+% the normal \tex/ way. Prefixes such as \cw{global}, \cw{long}, or
+% \cw{outer} must go inside the first argument rather than before
+% \cw{actively}.
+%
+% Usage:
+%\begin{usage}
+%\actively\def\?{<replacement text>}
+%\actively\def\%#1#2{<replacement text>}
+%\actively{\global\let}\^^@=\space
+%\end{usage}
+% One place where this function can be put to good use is in
+% making \ctrl{M} active in order to get special action at the end
+% of each line of input. The usual way of going about this would be
+% to write
+%
+%^^V \def\par{something}\obeylines
+%
+% \noindent which is a puzzling construction to the \tex/ novice
+% who doesn't know what \cw{obeylines} does with \cw{par}. The same
+% effect could be gotten a little more transparently with
+%
+%^^V \actively\def\^^M{something}
+%
+% In the definition of \cw{actively} we use the unique properties
+% of \cw{lowercase} to create an active character with the right
+% character code, overlapping with a \cw{begingroup} \cw{endgroup}
+% structure that localizes the necessary lc-code change.
+% \begin{macrocode}
+\def\actively#1#2{\catcode`#2\active
+ \begingroup \lccode`\~=`#2\relax
+ \lowercase{\endgroup#1~}}
+% \end{macrocode}
+% \end{macro}
+%
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%
+% \begin{macro}{\mesjsetup}
+%
+% The \cw{mesjsetup} function starts a group to localize catcode
+% changes. The group will be closed eventually by a separate
+% function that does the actual sending or stores the message text
+% for later retrieval.
+%
+% We want to change the catcode of each character in the three
+% lists \cw{otherchars}, \cw{controlchars}, and \cw{highchars} to
+% 12. After giving \cw{do} a recursive definition, we apply it to
+% each of the three lists, adding a suitable element at the end of
+% the list to make the recursion stop there. This allows
+% leaving out the \cw{do} tokens from the character lists, without
+% incurring the cost of an if test at each recursion step.
+% \begin{macrocode}
+\def\mesjsetup{\begingroup \count@=12
+ \def\do##1{\catcode`##1\count@ \do}%
+% \end{macrocode}
+% The abbreviation \cw{xp@} = \cw{expandafter} is from
+% \fn{grabhedr.sty}.
+% \begin{macrocode}
+ \xp@\do\otherchars{a11 \@gobbletwo}%
+ \xp@\do\controlchars{a11 \@gobbletwo}%
+ \xp@\do\highchars{a11 \@gobbletwo}%
+% \end{macrocode}
+% Make the tab character produce eight spaces:
+% \begin{macrocode}
+ \actively\edef\^^I{ \space\space\space
+ \space\space\space\space}%
+% \end{macrocode}
+% The convenient treatment of newlines in the argument of \cw{mesj}
+% (every line break produces a line break on screen) is achieved by
+% making the \ctrl{M} character active and defining it to produce a
+% category-12 \ctrl{J} character. Although for \cw{mesj} it would
+% have sufficed to make \ctrl{M} category 12 and locally set
+% \cw{newlinechar} = \ctrl{M} while sending the message, it turns out
+% to be useful for other functions to have the \ctrl{M} character
+% active, so that it can be remapped to an arbitrary function for
+% handling new lines (e.g., perhaps adding extra spaces at the
+% beginning of each line). And if \cw{mesj} treats \ctrl{M} the same,
+% we can arrange for it to share the setup routines needed for the
+% other functions.
+% \begin{macrocode}
+ \endlinechar=`\^^M\actively\let\^^M=\relax
+ \catcode`\{=1 \catcode`\}=2 }
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\sendmesj}
+%
+% In \cw{sendmesj} we go to a little extra trouble to make sure
+% \ctrl{M} produces a newline character, no matter what the value of
+% \cw{newlinechar} might be in the surrounding environment. The
+% impending \cw{endgroup} will restore \cw{newlinechar} to its
+% previous value. One reason for using \ctrl{J} (instead of, say,
+% \ctrl{M} directly) is to allow e.g. \cw{mesj}\verb'{xxx^^Jxxxx}' to
+% be written inside a definition, as is sometimes convenient. This
+% would be difficult with \ctrl{M} instead of \ctrl{J} because of
+% catcodes.
+%
+% \begin{macrocode}
+\def\sendmesj{\newlinechar`\^^J%
+ \actively\def\^^M{^^J}%
+ \immediate\write\sixt@@n{\mesjtext}\endgroup}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\mesj}
+%
+% Given the support functions defined above, the definition of
+% \cw{mesj} is easy: Use \cw{mesjsetup} to clear all special
+% catcodes, then set up \cw{sendmesj} to be triggered by the next
+% assignment, then read the following balanced-braces group into
+% \cw{mesjtext}. As soon as the definition is completed, \tex/ will
+% execute \cw{sendmesj}, which will send the text and close the
+% group that was started in \cw{mesjsetup} to localize the catcode
+% changes.
+%
+% \begin{macrocode}
+\def\mesj{\mesjsetup \afterassignment\sendmesj
+ \def\mesjtext}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\sendprompt}
+%
+% The \cw{sendprompt} function is just like \cw{sendmesj} except
+% that it uses \cw{message} instead of \cw{write}, as might be
+% desired when prompting for user input, so that the on-screen
+% cursor stays on the same line as the prompt instead of hopping
+% down to the beginning of the next line. In order for newlines to
+% work with \cw{message} we must use a visible character instead of
+% \ctrl{J}. When everyone has \tex/ version 3.1415 or later this
+% will no longer be true. The choice of \qc{\!} might be construed
+% (if you wish) as editorial comment that \qc{\!} should not be
+% shouted at the user in a prompt.
+%
+% \begin{macrocode}
+\def\sendprompt{%
+ \newlinechar`\!\relax \actively\def\^^M{!}%
+ \message{\mesjtext}\endgroup}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\promptmesj}
+%
+% This function is like \cw{mesj} but uses \cw{sendprompt} instead of
+% \cw{sendmesj}.
+%
+% \begin{macrocode}
+\def\promptmesj{\mesjsetup
+ \afterassignment\sendprompt \def\mesjtext}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\storemesj}
+%
+% Arg \arg{1} of \cw{storemesj} is the control sequence under which
+% the message text is to be stored.
+% \begin{macrocode}
+\def\storemesj#1{\mesjsetup
+ \catcode`\#=6 % to allow arguments if needed
+ \afterassignment\endgroup
+ \long\gdef#1}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\fmesj}
+%
+% While \cw{storemesj}\cw{foo}\verb"{...}" is more or less the
+% same as \cw{def}\cw{foo}\verb"{...}" with special catcode
+% changes, \cw{fmesj}\cw{foo}\verb"{...}" corresponds to
+% \cw{def}\cw{foo}\qc{\{}\cw{mesj}\verb"{...}}", that is, after
+% \cw{fmesj}\cw{foo} the function \cw{foo} can be executed directly
+% to send the message. Thus \cw{storemesj} is typically used for
+% storing pieces of messages, while \cw{fmesj} is used for storing
+% entire messages.
+%
+% To read the parameter text \arg{2}, we use the peculiar
+% \verb'#{' feature of \tex/ to read everything up to the opening
+% brace.
+% \begin{macrocode}
+\def\fmesj#1#2#{\mesjsetup
+ \catcode`\#=6 % restore to normal
+% \end{macrocode}
+% The parameter text \arg{2} must be stored in a token register
+% rather than a macro to avoid problems with \qc{\#} characters.
+% The \cw{long} prefix is just to admit the (unlikely) possibility
+% of using \cw{fmesj} to define something such as an error message
+% saying `You can't use \arg{1} here' where one of the
+% possibilities for \arg{1} is \cw{string}\cw{par}.
+% \begin{macrocode}
+ \toks@{\long\gdef#1#2}%
+% \end{macrocode}
+% Define \cw{@tempa} to put together the first two arguments and
+% [pseudo]argument \arg{3} and make the definition of \arg{1}.
+% \begin{macrocode}
+ \def\@tempa{%
+ \edef\@tempa{%
+ \the\toks@{%
+% \end{macrocode}
+% The abbreviation \cw{nx@} = \cw{noexpand} is from
+% \fn{grabhedr.sty}.
+% \begin{macrocode}
+ \begingroup\def\nx@\mesjtext{\the\toks2 }%
+ \nx@\sendmesj}%
+ }%
+ \@tempa
+ \endgroup % Turn off the \mesjsetup catcodes
+ }%
+ \afterassignment\@tempa
+ \toks2=}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\xmesjsetup}
+%
+% \cw{xmesjsetup} is like \cw{mesjsetup} except it prepares to allow
+% control sequence tokens and normal comments in the message text.
+% For \tex/nicians' convenience certain other features are
+% thrown in.
+%
+% Here, unlike the setup for \cw{xreadline}, I don't bother to
+% remove the outerness of \cw{bye}, \cw{newif}, etc., because I
+% presume the arguments of \cw{xmesj}, \cw{fxmesj},
+% \cw{storexmesj}, \cw{fxmenu}, etc.\ are more likely to be written
+% by a \tex/nician than by an average end user, whereas
+% \cw{xreadline} is designed to handle arbitrary input from
+% arbitrary users.
+%
+% \begin{macrocode}
+\def\xmesjsetup{\mesjsetup
+% \end{macrocode}
+% Throw in pseudo braces just in case we are inside an \cw{halign}
+% with \cs{\\} let equal to \cw{cr} at the time when \cw{xmesjsetup}
+% is called. (As might happen in \amstex/.)
+% \begin{macrocode}
+ \iffalse{\fi
+ \catcode`\\=0 \catcode`\%=14
+% \end{macrocode}
+% Define {\def~{\spacefactor1500 \space}%
+% \cs{\%}~\cs{\\}~\cs{\{}~\cs{\}}~\cs{\&}} to produce the
+% corresponding single characters, category 12. The \cw{lowercase}
+% trick here allows these definitions to be nonglobal.
+% \begin{macrocode}
+ \begingroup \lccode`\0=`\\\lccode`\1=`\{%
+ \lccode`\2=`\}\lccode`\3=`\%%
+ \lowercase{\endgroup \def\\{0}\def\{{1}%
+ \def\}{2}\def\%{3}}%
+ \iffalse}\fi
+ \edef\&{\string &}%
+% \end{macrocode}
+% Let \qc{\&} \qc{\=} \cw{noexpand} for expansion control inside the
+% argument text; let active \ctrl{M} \qc{\=} \cw{relax} so that
+% newlines will remain inert during the expansion.
+% \begin{macrocode}
+ \actively\let\&=\noexpand
+ \actively\let\^^M=\relax
+% \end{macrocode}
+% Define \cs{\.} to be a noop, for terminating a control word when
+% it is followed by letters and no space is wanted.
+% \begin{macrocode}
+ \def\.{}%
+% \end{macrocode}
+% Support for use of \cw{xmesj} inside a definition replacement
+% text or macro argument: control-space
+% \verb*"\ " = \cw{space}, tilde \qc{\~} prints as itself,
+% \cs{\^^M} (i.e., a lone backslash at the end of a line) will
+% produce a newline, also \cs{\^^J}, while finally \cw{par} \qc{\=}
+% blank line translates to two newlines.
+% \begin{macrocode}
+ \def\ { }\edef~{\string ~}%
+% \end{macrocode}
+% Define \cs{\^^M} to produce an active \ctrl{M} character, which (we
+% hope) will be suitably defined to produce a newline or whatever.
+% \begin{macrocode}
+ \begingroup \lccode`\~=`\^^M%
+ \lowercase{\endgroup \def\^^M{~}}%
+ \let\^^J\^^M \def\par{\^^M\^^M}%
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\xmesj}
+%
+% \cw{xmesj} uses \cw{xmesjsetup} and \cw{edef}.
+% \begin{macrocode}
+\def\xmesj{\xmesjsetup \afterassignment\sendmesj
+ \edef\mesjtext}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\promptxmesj}
+%
+% \cw{promptxmesj} is analogous to \cw{promptmesj}, but with
+% expansion.
+% \begin{macrocode}
+\def\promptxmesj{\xmesjsetup
+ \afterassignment\sendprompt \edef\mesjtext}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\storexmesj}
+%
+% And \cw{storexmesj} is like \cw{storemesj}, with expansion. Since
+% we allow arguments for the function being defined, we also must
+% define \cs{\#} to produce a single category-12 \qc{\#} character so
+% that there will be a way to print \qc{\#} in the message text.
+% \begin{macrocode}
+\def\storexmesj#1#2#{\xmesjsetup
+ \catcode`\#=6 % to allow arguments if needed
+ \edef\#{\string##}%
+ \afterassignment\endgroup
+ \long\xdef#1#2}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\fxmesj}
+%
+% And \cw{fxmesj} is the expansive analog of \cw{fmesj}.
+% \begin{macrocode}
+\def\fxmesj#1#2#{\xmesjsetup
+ \catcode`\#=6 % restore to normal
+ \edef\#{\string##}%
+ \toks@{\long\xdef#1#2}%
+ \def\@tempa{%
+ \edef\@tempa{%
+ \the\toks@{\begingroup
+ \def\nx@\nx@\nx@\mesjtext{\the\toks\tw@}%
+ \nx@\nx@\nx@\sendmesj}}%
+ \@tempa % execute the constructed xdef
+ \endgroup % restore normal catcodes
+ }%
+ \afterassignment\@tempa
+ \toks\tw@=}
+% \end{macrocode}
+% \end{macro}
+%
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% \section{Reading functions}
+%
+% \begin{macro}{\readline}
+% The \cw{readline} function gets one line of input from the user.
+% Arguments are: \arg{1} default to be used if the user response is
+% empty (i.e., if the user just pressed the return/enter key),
+% \arg{2} macro to receive the input.
+%
+% \begin{macrocode}
+\def\readline#1#2{%
+ \begingroup \count@ 12 %
+ \def\do##1{\catcode`##1\count@ \do}%
+ \xp@\do\otherchars{a11 \@gobbletwo}%
+ \xp@\do\controlchars{a11 \@gobbletwo}%
+ \xp@\do\highchars{a11 \@gobbletwo}%
+% \end{macrocode}
+% Make spaces and tabs normal instead of category 12.
+% \begin{macrocode}
+ \catcode`\ =10 \catcode`\^^I=10 %
+ \catcode`\^^M=9 % ignore
+% \end{macrocode}
+% Reset end-of-line char to normal, just in case.
+% \begin{macrocode}
+ \endlinechar`\^^M
+% \end{macrocode}
+% We go to a little trouble to avoid \cw{gdef}-ing \arg{2}, in order
+% to prevent save stack buildup if the user of \cw{readline} carries
+% on unaware doing local redefinitions of \arg{2} after the initial
+% read.
+% \begin{macrocode}
+ \read\m@ne to#2%
+ \edef#2{\def\nx@#2{#2}}%
+ \xp@\endgroup #2%
+ \ifx\@empty#2\def#2{#1}\fi
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\xreadline}
+%
+% \cw{xreadline} is like \cw{readline} except that it leaves almost
+% all catcodes unchanged so that the return value is executable
+% tokens instead of strictly character tokens of category 11 or 12.
+% \begin{macrocode}
+\def\xreadline#1#2{%
+ \begingroup
+% \end{macrocode}
+% Render some outer control sequences innocuous.
+% \begin{macrocode}
+ \xp@\let\csname bye\endcsname\relax
+ \xp@\let\csname newif\endcsname\relax
+ \xp@\let\csname newcount\endcsname\relax
+ \xp@\let\csname newdimen\endcsname\relax
+ \xp@\let\csname newskip\endcsname\relax
+ \xp@\let\csname newmuskip\endcsname\relax
+ \xp@\let\csname newtoks\endcsname\relax
+ \xp@\let\csname newbox\endcsname\relax
+ \xp@\let\csname newinsert\endcsname\relax
+ \xp@\let\csname +\endcsname\relax
+ \actively\let\^^L\relax
+ \catcode`\^^M=9 % ignore
+ \endlinechar`\^^M% reset to normal
+ \read\m@ne to#2%
+ \toks@\xp@{#2}%
+ \edef\@tempa{\def\nx@#2{\the\toks@}}%
+ \xp@\endgroup \@tempa
+ \ifx\@empty#2\def#2{#1}\fi
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\readchar}
+%
+% \cw{readchar} reduces the user response to a single character.
+% \begin{macrocode}
+\def\readchar#1#2{%
+ \readline{#1}#2%
+% \end{macrocode}
+% If the user's response and the default response are both empty,
+% we need something after \arg{1} to keep \cw{@car} from running
+% away, so we add an empty pair of braces.
+% \begin{macrocode}
+ \edef#2{\xp@\@car#2#1{}\@nil}%
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\readChar}
+%
+% \cw{readChar} reduces the user response to a single uppercase
+% character. (This is useful to simplify testing the response
+% later with \cw{if}.)
+%
+% \begin{macrocode}
+\def\readChar#1#2{%
+ \readline{#1}#2%
+ \changecase\uppercase#2%
+% \end{macrocode}
+% Reduce \arg{2} to its first character, or the first character of
+% \arg{1}, if \arg{2} is empty. The extra braces \verb'{}' are to
+% prevent a runaway argument error from \cw{@car} if \arg{2} and
+% \arg{1} are both empty.
+% \begin{macrocode}
+ \edef#2{\xp@\@car #2#1{}\@nil}%
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\changecase}
+% The function \cw{changecase} uppercases or lowercases the
+% replacement text of its second argument, which
+% must be a macro. The first argument should be \cw{uppercase} or
+% \cw{lowercase}.
+% \begin{macrocode}
+\def\changecase#1#2{\@casetoks\xp@{#2}%
+ \edef#2{#1{\def\nx@#2{\the\@casetoks}}}#2}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@casetoks}
+% We allocate a token register just for the use of \cw{changecase}
+% because it might be used at a low level internally where we don't
+% want to interfere with other uses of the scratch token registers
+% 0\dash 9.
+% \begin{macrocode}
+\newtoks\@casetoks
+% \end{macrocode}
+% \end{macro}
+%
+% A common task in reading user input is to verify, when an answer
+% of a certain kind was requested, that the response has indeed the
+% desired form\Dash for example, if a nonnegative integer is required
+% for subsequent processing, it behooves us to verify that we have
+% a nonnegative integer in hand before doing anything that might
+% lead to inconvenient error messages. However, it's not easy to
+% decide how best to handle such verification. One possibility
+% might be to have a function
+%
+%^^V \readnonnegativeinteger\foo
+%
+% to do all the work of going out and fetching a number from the
+% user and leaving it in the macro \cw{foo}. Another possibility
+% would be to read the response using \cw{readline} and then apply
+% a separate function that can be used in combination with \cw{if},
+% for example
+%
+%^^V \readline{}\reply
+%^^V \if\validnumber\reply ... \else ... \fi
+%
+% For maximum flexibility, a slightly lower-level approach is
+% chosen here. The target syntax is
+%
+%^^V \readline{}\reply
+%^^V \checkinteger\reply\tempcount
+%
+% where \cw{tempcount} will be set to \verb'-'\cw{maxdimen} if
+% \cw{reply} does {\em not\/} contain a valid integer. (Negative
+% integers are allowed, as long as they are greater than
+% \verb'-'\cw{maxdimen}.) Then the function that calls
+% \cw{checkinteger} is free to make additional checks on the range
+% of the reply and give error messages tailored to the
+% circumstances. And the handling of an empty \cw{reply} can be
+% arbitrarily customized, something that would tend to be
+% inconvenient for the first method mentioned.
+%
+% The first and second approaches can be built on top of the third
+% if desired, e.g.\ (for the second approach)
+%
+%^^V \def\validnumber#1{TT\fi
+%^^V \checkinteger#1\tempcount%
+%^^V \ifnum\tempcount>-\maxdimen }
+%
+% The curious \verb'TT\fi'\verbdots\cw{ifnum} construction is from
+% \TeXhax{} 1989, no.~20 and no.~38 (a suggestion of D. E. Knuth in
+% reply to a query by S. von Bechtolsheim).
+%
+% \begin{macro}{\checkinteger}
+%
+% The arguments of \cw{checkinteger}'s are: \arg{2}, a count register
+% to hold the result; \arg{1}, a macro holding zero or more arbitrary
+% characters of category 11 or 12.
+%
+% \begin{macrocode}
+\def\checkinteger#1#2{\let\scansign@\@empty
+ \def\scanresult@{#2}%
+ \xp@\scanint#1x\endscan}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\scanint}
+%
+% To validate a number, the function \cw{scanint} must first scan
+% away leading \qc{\+} or \qc{\-} signs (keeping track in
+% \cw{scansign@}), then look at the first token after that: if it's
+% a digit, fine, scan that digit and any succeeding digits into the
+% given count register (\cw{scanresult@}), ending with \cw{endscan}
+% to get rid of any following garbage tokens that might just
+% possibly show up.
+%
+% Typical usage of \cw{scanint} includes initializing \cw{scansign@}
+% to empty, as in the definition of \cw{checkinteger}.
+%\begin{usage}
+%\let\scansign@\@empty
+%\def\scanresult@{\tempcount}%
+%\xp@\scanint\reply x\endscan
+%\end{usage}
+% Assumption: \cw{reply} is either empty or contains only category
+% 11 or 12 characters (which it will if you used \cw{readline}!). If
+% a separate check is done earlier to trap the case where \cw{reply}
+% is empty\Dash for example, by using a nonempty default for
+% \cw{readline}\Dash then the \verb'x' before \cw{endscan} is
+% superfluous.
+%
+% Arg \arg{1} = next character from the string being tested.
+% The test whether \arg{1} is a decimal digit is similar in spirit
+% to the test \verb'\if!#1!' to see if an argument is empty
+% (\TB, Appendix~D, p.~376).
+% \begin{macrocode}
+\def\scanint#1{%
+ \ifodd 0#11 %
+% \end{macrocode}
+% Is \arg{1} a decimal digit? If so read all digits into
+% \cw{scanresult@} with the sign prefix.
+% \begin{macrocode}
+ \def\@tempa{\afterassignment\endscan
+ \scanresult@=\scansign@#1}%
+ \else
+ \if -#1\relax
+% \end{macrocode}
+% Here we flipflop the sign; watch closely.
+% \begin{macrocode}
+ \edef\scansign@{%
+ \ifx\@empty\scansign@ -\fi}%
+ \def\@tempa{\scanint}%
+ \else
+% \end{macrocode}
+% A plus sign can just be ignored.
+% \begin{macrocode}
+ \if +#1\relax
+ \def\@tempa{\scanint}%
+ \else % not a valid number
+ \def\@tempa{%
+ \scanresult@=-\maxdimen\endscan}%
+ \fi\fi\fi
+ \@tempa
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\endscan}
+% The \cw{endscan} function just gobbles any remaining garbage. It
+% uses its own name as the argument delimiter.
+% \begin{macrocode}
+\def\endscan#1\endscan{}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\dimenfirstpart}
+% \begin{macro}{\dimentoks}
+% \cw{dimenfirstpart}, a count register, receives the digits, if
+% any, preceding the decimal point. \cw{dimentoks}, a token
+% register, receives any digits after the decimal point.
+% \begin{macrocode}
+\newcount\dimenfirstpart
+\newtoks\dimentoks
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\scandimen}
+%
+% \cw{scandimen} is similar to \cw{scanint} but has to call some
+% auxiliary functions to scan the various subcomponents of a
+% dimension (leading digits, decimal point, fractional part, and
+% units, with optional {\tt true}, in addition to the sign). The
+% minimum requirements of \tex/'s syntax for dimensions are a digit
+% or decimal point \qc{\+} the units; all the other components are
+% optional (\TB, Exercise~10.3, p.~58).
+%
+% When scanning for the digits of a fractional part, we can't throw
+% away leading zeros; therefore we don't read the fractional part
+% into a count register as we did for the digits before the decimal
+% point; instead we read the digits one by one and store them in
+% \cw{dimentoks}.
+%
+% The function that calls \cw{scandimen} should initialize
+% \cw{scansign@} to \cw{@empty}, \cw{dimenfirstpart} to \cw{z@},
+% \cw{dimentoks} to empty \verb'{}', and \cw{dimentrue@} to
+% \cw{@empty}.
+%
+% Test values: {\tt 0pt}, {\tt 1.1in}, {\tt -2cm}, {\tt .3mm}, {\tt
+% 0.4dd}, {\tt 5.cc}, {\tt.10000000009pc}, \cw{hsize}, {\tt em}.
+%
+% \begin{macrocode}
+\def\scandimen#1{%
+ \ifodd 0#11
+ \def\@tempa{\def\@tempa{\scandimenb}%
+ \afterassignment\@tempa
+ \dimenfirstpart#1}%
+ \else
+% \end{macrocode}
+%
+% The following test resolves to true if \arg{1} is either a period
+% or a comma (both recognized by \tex/ as decimal point
+% characters).
+%
+% \begin{macrocode}
+ \if \if,#1.\else#1\fi.%
+ \def\@tempa{\scandimenc}%
+ \else
+ \if -#1% then flipflop the sign
+ \edef\scansign@{%
+ \ifx\@empty\scansign@ -\fi}%
+ \def\@tempa{\scandimen}%
+ \else
+ \if +#1% then ignore it
+ \def\@tempa{\scandimen}%
+ \else % not a valid dimen
+ \def\@tempa{%
+ \scanresult@=-\maxdimen\endscan}%
+ \fi\fi\fi\fi
+ \@tempa
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\scandimenb}
+%
+% Scan for an optional decimal point.
+%
+% \begin{macrocode}
+\def\scandimenb#1{%
+ \if \if,#1.\else#1\fi.%
+ \def\@tempa{\scandimenc}%
+ \else
+% \end{macrocode}
+% If the decimal point is absent, we need to put back \arg{2} and
+% rescan it to see if it is the first letter of the units.
+% \begin{macrocode}
+ \def\@tempa{\scanunitsa#1}%
+ \fi
+ \@tempa
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\scandimenc}
+%
+% Scan for the fractional part: digits after the decimal point.
+%
+% \begin{macrocode}
+\def\scandimenc#1{%
+% \end{macrocode}
+% If \arg{1} is a digit, add it to \cw{dimentoks}.
+% \begin{macrocode}
+ \ifodd 0#11 \dimentoks\xp@{%
+ \the\dimentoks#1}%
+ \def\@tempa{\scandimenc}%
+ \else
+% \end{macrocode}
+% Otherwise rescan \arg{1}, presumably the first letter of the
+% units.
+% \begin{macrocode}
+ \def\@tempa{\scanunitsa#1}%
+ \fi
+ \@tempa
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\scanunitsa}
+%
+% \begin{macrocode}
+\def\scanunitsa#1\endscan{%
+% \end{macrocode}
+% Check for \verb'true' qualifier.
+% \begin{macrocode}
+ \def\@tempa##1true##2##3\@tempa{##2}%
+% \end{macrocode}
+%
+% The peculiar nature of \cw{lowercase} is evident here as we are
+% able to apply it to only the test part of the conditional without
+% running into brace-matching problems. (Compare the braces in this
+% example to something like \cw{message}\qc{\{}\cw{iffalse}
+% \verb"A}"\cw{else} \verb"B}"\cw{fi}.)
+%
+% \begin{macrocode}
+ \lowercase{%
+ \xp@\ifx\xp@\end
+ \@tempa#1true\end\@tempa
+ }%
+% \end{macrocode}
+% No \verb'true' was found:
+% \begin{macrocode}
+ \let\dimentrue@\@empty
+ \def\@tempa{\scanunitsb#1\endscan}%
+ \else
+ \def\dimentrue@{true}%
+ \def\@tempa##1true##2\@tempa{%
+ \def\@tempa{##1}%
+ \ifx\@tempa\@empty
+ \def\@tempa{\scanunitsb##2\endscan}%
+ \else
+ \def\@tempa{\scanunitsb xx\endscan}%
+ \fi}%
+ \@tempa#1\@tempa
+ \fi
+ \@tempa
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\scanunitsb}
+%
+% Scan for the name of the units and complete the assignment of the
+% scanned value to \cw{scanresult@}. Notice that, because of the way
+% \cw{scanunitsb} picks up \arg{1} and \arg{2} as macro arguments,
+% \verb"p t" is allowed as a variation of \verb"pt". Eliminating
+% this permissiveness doesn't seem worth the speed penalty
+% that would be incurred in \cw{scanunitsb}.
+%
+% The method for detecting a valid units string is to define the
+% scratch function \cw{@tempa} to apply \tex/'s parameter-matching
+% abilities to a special string that will yield a boolean value
+% of true if and only if the given string is a valid \tex/ unit.
+% \begin{macrocode}
+\def\scanunitsb#1#2{%
+ \def\@tempa##1#1#2##2##3\@nil{##2}%
+ \def\@tempb##1{T\@tempa
+ pcTptTcmTccTemTexTinTmmTddTspT##1F\@nil}%
+% \end{macrocode}
+% Force lowercase just in case the units were entered with
+% uppercase letters (accepted by \tex/, so we had better accept
+% uppercase also).
+% \begin{macrocode}
+ \lowercase{%
+ \if\@tempb{#1#2}%
+ }%
+ \scanresult@=\scansign@
+ \number\dimenfirstpart.\the\dimentoks
+ \dimentrue@#1#2\relax
+ \else
+ \scanresult@=-\maxdimen
+ \fi
+% \end{macrocode}
+% Call \cw{endscan} to gobble garbage tokens, if any.
+% \begin{macrocode}
+ \endscan
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\checkdimen}
+%
+% Argument \arg{2} must be a dimen register; \arg{1} is expected to
+% be a macro holding zero or more arbitrary characters of category 11
+% or 12.
+% \begin{macrocode}
+\def\checkdimen#1#2{%
+ \let\scansign@\@empty \def\scanresult@{#2}%
+ \let\dimentrue@\@empty
+ \dimenfirstpart\z@ \dimentoks{}%
+ \xp@\scandimen#1xx\endscan
+}
+% \end{macrocode}
+% \end{macro}
+%
+% Finish up.
+% \begin{macrocode}
+\restorecatcodes
+\endinput
+% \end{macrocode}
+%
+% \CheckSum{858}
+% \Finale