path: root/Master/texmf-dist/doc/latex/nicetext/doclog/fifinddo.doc

\ProvidesFile{fifinddo.doc}[2009/04/16 automatically generated with makedoc.sty]
\begin{packagecode}
%% Macro package `fifinddo.sty' for LaTeX2e,    %% FIDO, FIND!
%% copyright (C) 2009 Uwe L\"uck,
%%   http://www.contact-ednotes.sty.de.vu
%% -- author-maintained in the sense of LPPL below --
%% for processing tex(t) files
%% (checking, filtering, converting, substituting, expanding, ...)

\def\fileversion{0.3} \def\filedate{2009/04/15}

%% This file can be redistributed and/or modified under
%% the terms of the LaTeX Project Public License; either
%% version 1.3a of the License, or any later version.
%% The latest version of this license is in
%%
%%     http://www.latex-project.org/lppl.txt
%%
%% We did our best to help you, but there is NO WARRANTY.
%% Please report bugs, problems, and suggestions via
%%
%%   http://www.contact-ednotes.sty.de.vu
%%
%% For the full documentation, look for `fifinddo.pdf'.
%% Its source starts in `fifinddo.tex'.
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  Format and package version \unskip }

\begin{packagecode}
\NeedsTeXFormat{LaTeX2e}[1994/12/01]
% 1994/12/01: \newcommand* etc.
\ProvidesPackage{fifinddo}[\filedate\space v\fileversion\space
                           filtering TeX(t) files by TeX (UL)]
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  Category codes \unskip }


We use the ``underscore'' as ``compound identifier.''
\begin{packagecode}
\catcode`\_=11 %% underscore used in control words
\end{packagecode}
|\MakeOther| is a synonym for `\@makeother', needed for matching
special characters from the input file. It is exemplified by
|\fdPatternCodes| which is the default of |\PatternCodes|.
The latter is used in setup macros for reading patterns.
\begin{packagecode}
\@ifundefined{MakeOther}{\let\MakeOther\@makeother}{}
\newcommand*{\fdPatternCodes}{\MakeOther\&\MakeOther\$}
\newcommand*{\PatternCodes}{} \let\PatternCodes\fdPatternCodes
  %% TODO adding/removing
\end{packagecode}
It would be bad to have `\MakeOther\%' and `\MakeOther\ ' here in
that this may have unexpected, weird effects with arguments of
setup macros. Therefore neither `\dospecials' nor `\@sanitize' are
used. Curly braces remain untouched as default delimiters in setup
macros. For matching them, you must use `\MakeOther\{' and
`\MakeOther' in your `\PatternCodes', or |\Delimiters| to introduce
new ones at the same time, e.g., `\Delimiters\[\]':
\begin{packagecode}
\newcommand*{\Delimiters}[2]{%
  \MakeOther\{\MakeOther\}\catcode`#1=1\catcode`#2=2\relax}
\end{packagecode}
For replacing strings or for defining other strings of ``other''
characters by `\edef', you can use some \LaTeX\ constructs---here
are copies |\PercentChar| and |\BackslashChar| of them
(do you need more?):
\begin{packagecode}
\newcommand*{\PercentChar}{}   \let\PercentChar\@percentchar
\newcommand*{\BackslashChar}{} \let\BackslashChar\@backslashcar
\end{packagecode}

\section{\hspace{1sp}\ignorespaces  File handling \unskip }

\begin{packagecode}
\newwrite\result_file %% or write to \@mainaux!?
\end{packagecode}
|\ResultFile{<output>}| opens (and empties) a file
<output> to be written into.
\begin{packagecode}
\newcommand*{\ResultFile}[1]{%
  \def\result_file_name{#1}%
  \immediate\openout\result_file=#1}
\end{packagecode}
|\WriteResult{<balanced>}| writes a <balanced> line into
<output> (or more lines with `^^J').
\begin{packagecode}
\newcommand*{\WriteResult}[1]{%
  \immediate\write\result_file{#1}}
\end{packagecode}
|\WriteProvides| writes a `\ProvidesFile' command to the
opened <output> file. This should be used when <output>
is made as \LaTeXe\ input.
\begin{packagecode}
\newcommand*{\WriteProvides}{%
  \WriteResult{%
    \string\ProvidesFile{\result_file_name}%
      [\the\year/\two@digits\month/\two@digits\day\space
       automatically generated with fifinddo.sty]}}%
\end{packagecode}
|\ProcessFileWith{<input>}{<loop-body>}| opens a file <input>
and runs a loop on its lines the main body of which is <loop-body>.
When it starts, a new line of <input> is stored as macro
|\fdInputLine|.
\begin{packagecode}
\newcommand*{\ProcessFileWith}[2]{%
  \openin\@inputcheck=#1%
%   \ifeof\@inputcheck  %% bad `exists?' test
%     \PackageError{fifinddo}{File `#1' not here}%
%                            {Mistyped?}%
%   \else
    \global\c@fdInputLine=\z@ %% line counter reset
    \begingroup
      \MakeOther\{\MakeOther\}\@sanitize
      %% from docstrip.tex:
      % \MakeOther\^^A\MakeOther\^^K%% irrelevant, not LaTeX
      \endlinechar\m@ne
      %% <- cf. TeXbook "extended keyboards" up-/downarrow
      %%    -> "math specials", cf. "space specials"
      \MakeOther\^^I% ASCII horizontal tab -- guessed!? ^^L!?
      \loop \ifeof\@inputcheck \else
        \read\@inputcheck to \fdInputLine
        \ignorespaces #2%
      \repeat
    \endgroup
%   \fi
  \closein\@inputcheck}
\end{packagecode}
|\CloseResultFile| closes <output>.
\begin{packagecode}
\newcommand*{\CloseResultFile}{\immediate\closeout\result_file}
\end{packagecode}
Peter Wilson's 'newfile' provides more powerful file handling.
% <- TODO relevant? 2009/04/12

\pagebreak %% TODO move theory to fifinddo.tex 2009/04/12

\section{\hspace{1sp}\ignorespaces  Basic handling of substring conditionals \unskip }

\label{sec:theory}

\subsection{\hspace{1sp}\ignorespaces  ``Substring Theory'' \unskip }

\begin{flushright}\it
I wished I could study string theory,\\
but I only could study substring theory.
\end{flushright}
A \TeX\ macro, say, `\find' whose parameter text
(cf.\  \TeX book p.~203)
starts with `#1<pattern>#2&' stops \TeX\ with an error if it does
not find <pattern> and then `&'. Otherwise we have a situation
`\find<split1><pattern><split2>&', and `\find' reads <split1> as
#1 and <split2> as #2. An important point to note is that <split1>
will not contain <pattern>, but possibly <pattern> has more
occurrences in <split2>. In this sense, `\find' uses the
\emph{first} occurrence of <pattern> it finds in order to delimit #1.
Finding the \emph{last} occurrence of <pattern> therefore needs a
special idea.

In order to use `\find' for a test whether <pattern> is in
<target>, we build a ``sandbox'' |\find<sand>&|, where <sand>
contains <target> \emph{and additionally} <pattern>---as a
``dummy;'' so `&' delimits the search and `\find' finds
<pattern> either in <target> or somewhere else before `&'.

Consider the simple sandbox |\find<target><pattern>&|.
We can test #1 and #2 on being empty by `\ifx$#1$' and `\ifx$#2$'.
If #2 is empty, <pattern> is \emph{not} in <target>.
If #1 is empty at the same time, <target> is empty.
If #1 is empty and #2 is not, <pattern> \emph{starts} <target>!
This can be used to implement
Wikipedia-like lists                 %% TODO 2009/04/11
and to distinguish package code from comments in 'makedoc'.

If #2 is \emph{not} empty, <pattern> occurs in <target>---or this once
was \emph{thought}, some time in developping the present package,
as well as in the version of 'substr.sty' marked
`2005-11-29',\footnote{\hspace{1sp}'substr' does not change
  category codes
  % as 'fifinddo' does
  and uses \cs{@nil} as delimiter instead of our &&.} try
\[`\IfSubStringInString{<str1><str2><str1>}{<str1><str2>}{YES}{NO}'\]
which works \emph{verbatim} as well as considering <str1> and
<str2> \emph{placeholders}, e.g., for
\begin{center}
  `\IfSubStringInString{day after day}{day after }{YES}{NO}'\footnote{%
    Likewise \texttt{t\string\^ete-\string\`a-t\string\^te} $\dots$}\\
  `\IfSubStringInString{AMSTERDAM}{AMSTERD}{YES}{NO}'\\
  `\IfSubStringInString{TORONTO}{TORON}{YES}{NO}'\\
  `\IfSubStringInString{bonbon}{bon}{YES}{NO}'\footnote{%
    Polynesian: `aku aku', `rongorongo', `wiki wiki' $\dots$}\\
  `\IfSubStringInString{bonobo}{bono}{YES}{NO}'
  (an ape)
\end{center}
or `\IfSubStringInString{ionization}{ionizat}{YES}{NO}'.\footnote{%
  Read 'substr.sty' or try ``normal'' things to convince yourself
  that the syntax indeed is
  &\IfSubStringInString{<pattern>}{<target>}{<yes>}{<no>}.}
Same with \LaTeX's internal `\in@':
\[`\makeatletter \in@{bonbon}{bon}\ifin@ YES\else NO\fi \makeatother'\]
%% \makeatletter \in@{bonbon}{bon}\ifin@ YES\else NO\fi \makeatother
%% \IfSubStringInString{ionization}{ionizat}{YES}{NO}

In general, the previous approach \emph{fails if and exactly if}
<pattern> has a \emph{period} $p$---less than its length---in the sense of that
the $p$th token to the right or left of each token in <pattern>
is the \emph{same} token.
`AMSTERDAM' has a period 8, `day after day' 10, `bonbon' 3, `bonobo' 4.
There is a counterexample <target> of length $p$ iff
<pattern> has period $p$, namely the first substring of <pattern>
having length $p$. If the length of <pattern> exceeds a multiple
$mp$ of its period, the first $mp$ tokens of <pattern> form
a counterexample <target>.

Therefore, a sandbox must have something between <target> and
<pattern>. We choose |\find<target>~<pattern>$&| as standard.
The `$' will be used as an argument delimiter to get rid of the dummy
<pattern> in <split2>, as well as to decide whether the match was
in <target> or in the dummy part of the sandbox.
The `$' can be replaced by another tilde `~' in order to
test whether <target> \emph{ends} on a <pattern>, defining a macro
like `\findatend' whose parameter text starts with `#1<pattern>~#2&'.


\subsection{\hspace{1sp}\ignorespaces  Plan for proceeding \unskip }

When we check a file for several patterns, we seem to need
\emph{two} macros for each pattern: one that has the pattern
in its parameter text and one that stores the pattern for building
the sandbox.\footnote{If it were for the pattern only, the parsing
  macro might suffice and the macro calling it might extract the
  pattern from a ``dummy expansion.'' Somewhat too much for me now;
  on the other hand the calling macro also hands some ``current''
  informations to the parsing macro---oh, even this could be
  handled by a general ``calling'' macro \dots}
  %% TODO 2009/04/15
We use a separate \emph{``name space''} for each of
both kinds. The parsing macro and the macro building the sandbox
will have a common \emph{``identifier''} by which the user or
programmer calls them. Actually, she will usually (first) call
the sandbox box builder. The sandbox builder calls the parsing
macro. When \emph{all} occurrences of a pattern in the target are
looked for, the parser may call itself.

Actually, the parsing macro will execute certain actions
depending on what it finds in the sandbox, so we call it a
\emph{``substring conditional''}. It may read additional arguments
after the sandbox that store information gathered before.
This is especially useful for designing \emph{``expandable''}
chains (sequences) of conditionals where macros cannot store information in
macros. The macro setting up the sandbox will initialize such
extra arguments at the same time.

It may be more efficient \emph{not} to use the following setup
macros but to type the macros yourself, just using the following
as templates. The setup macros are especially useful with patterns
that contain ``special characters,'' as when you are looking for
lines that might be package comments.


\subsection{\hspace{1sp}\ignorespaces  Set up conditionals \unskip }

`substr_cond' is the ``name space'' for substring conditionals.
A colon separates it from \emph{``job identifiers''} in the actual
macro names.
\begin{packagecode}
\def\substr_cond{substr_cond:}
\end{packagecode}
|\MakeSubstringConditional{<id>}[<changes>]{<pattern>}|
starts the definition of a conditional with \emph{identifier} <id>
and pattern <pattern>. <changes> optionally add commands to
be executed after `\PatternCodes' in a local group.
It may be more safe to redefine `\PatternCodes' instead.
\begin{packagecode}
\newcommand*{\MakeSubstringConditional}{%
  \afterassignment\mk_substr_cond_san \def\cond_id}
\newcommand*{\mk_substr_cond_san}[1][]{%
  \begingroup \PatternCodes #1\mk_substr_cond}
  %% #1 more changes
\end{packagecode}
`\begingroup' |\mk_substr_cond{<pattern>}|
can be directly called by other programmer setup commands when
`\cond_id' and <pattern> have been read.
\begin{packagecode}
\def\mk_substr_cond #1{%% #1 pattern string
  \endgroup \@namedef{\substr_cond \cond_id}##1#1##2&}
\end{packagecode}
This really is not \LaTeX. We are starting defining a macro
`\substr_cond:<id>' in primitive \TeX\ with `\def' in the form
\[`\def\substr_cond:<id>#1<pattern>#2&'\]
where `\csname' etc.\ render \lq`:<id>'\rq\ part of the macro name.
The user or programmer macro produces the part of the definition
until the delimiter `&' to match the sandbox. You have to add
(maybe) #3 etc.\ and the `{<definition text>}'
just as with primitive \TeX.


\subsection{\hspace{1sp}\ignorespaces  Set up sandboxes \unskip }

There was a \emph{question:} will we rather see
\emph{string macros} or \emph{strings from macro arguments}?
The input file content always comes
as `\fdInputLine' first, so we at least \emph{must account} for
the possibility of string macros as input.

One easy way to apply several checks and substitutions to
`\fdInputLine' before the result is written to <output> is
`\let\OutputString\fdInputLine' and then let `\OutputString'
be to what each job refers as \emph{its}
input and output, finally `\WriteResult{\OutputString}'.
(`\fdInputLine' might better not be touched, it could be used
for a final test whether any change applied for some message on
screen, even with an entirely expandable chain of actions.)
This way each job, indeed each recursive substitution of a single
string must start with expanding `\OutputString'.

On the other hand, there is the idea of \emph{``expandable''
chains of substitutions}. We may, e.g., define a macro, say,
`\manysubstitutions{<macro-name>}', such that
`\WriteResult{\manysubstitutions{\fdInputLine}}'
writes to <output> the result of applying many expandable
substitutions to `\fdInputLine'.
Such a macro `\manysubstitutions' may read `\fdInputLine',
but it must not redefine any macros.
Instead, the substitution macros it calls must read results
of previous substitutions as \emph{arguments}.

Another aspect:
the order of substitutions should be easy to
change. Therefore expanding of string macros should rather be
controlled by the way a job is \emph{called}, not right here
at the \emph{definition} of the job. For this reason,
a variant of the sandbox builder expanding some macro was given up.

`setup_substr_cond' is the name space for macros that build
sandboxes and initialize arguments for conditional macros.
\begin{packagecode}
\def\setup_substr_cond{setup_substr_cond:}
\end{packagecode}
|\MakeSetupSubstringCondition{<id>}[<changes>]{<pattern>}{<more-args>}|
% <- TODO allow `%' and ` ' for breaking code lines.
---same <id>, <changes>, <pattern> as for
`\MakeSubstringConditional' (this is bad, there may be
|\MakeSubstringConditional*{<more-args>}|)---creates the
% <- TODO: store args in \Make$\dots$Conditional
corresponding sandbox, by default without tilde wrap.
<more-args> may contain `{#1}' to store the string that was tested,
also `{<id>}' for calling repetitions and `{<pattern>}' for screen
or log informations.
\begin{packagecode}
\newcommand*{\MakeSetupSubstringCondition}{%
  \afterassignment\mk_setup_substr_cond_san \def\cond_id}
\newcommand*{\mk_setup_substr_cond_san}[1][]{%
  \begingroup \PatternCodes #1\mk_setup_substr_cond}
\end{packagecode}
`\begingroup' |\mk_setup_substr_cond{<pattern>}{<more-args>}|
can be directly called by other programmer setup commands after
`\cond_id' and <pattern> have been read:
\begin{packagecode}
\def\mk_setup_substr_cond #1#2{%% #1 pattern string,
  %% #2 additional arguments, e.g., `{#1}' to keep tested string
  \endgroup
  \expandafter \edef
   \csname \setup_substr_cond \cond_id \endcsname ##1{%
%       \expandafter \noexpand
%         \csname \substr_cond \cond_id \endcsname %% 2009/04/10:
      \make_not_expanding_cs{\substr_cond \cond_id}%
\end{packagecode}
By `\edef', the name of the substring conditional is stored here
as a single token. The rest of the sandbox follows.
\begin{packagecode}
          ##1\noexpand~#1\dollar_tilde&#2}%
  \let\dollar_tilde\sandbox_dollar}
\end{packagecode}
If a tilde `~' has been used instead of `$', the default
is restored.
\begin{packagecode}
\def\sandbox_dollar{$}
\let\dollar_tilde\sandbox_dollar
\end{packagecode}
The following general tool |\make_not_expanding_cs| has been used
(many definitions in 'latex.ltx' could have used it): %% 2009/04/10
\begin{packagecode}
\def\make_not_expanding_cs#1{%
  \expandafter \noexpand \csname #1\endcsname}
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  Getting rid of the tildes \unskip }

|\let~\TildeGobbles| can be used to suppress dummy patterns
(contained in <split2>)
in `\write'ing or with `\edef'. $\dots$ will probably become obsolete
$\dots$ however, it is helpful in that you needn't care
whether there is a dummy wrap left at all. (2009/04/13)
\begin{packagecode}
\newcommand{\TildeGobbles}{}       \def\TildeGobbles#1${}
\end{packagecode}
|\RemoveDummyPattern| is used to remove the dummy pattern
\emph{immediately}, not waiting for `\write'ing
or other ``total'' expansion:   %% 2009/04/13
\begin{packagecode}
\newcommand{\RemoveDummyPattern}{} \def\RemoveDummyPattern#1~#2${#1}
\end{packagecode}
|\RemoveDummyPatternArg<macro>{<arg>}| executes
`\RemoveDummyPattern' in the next argument:
\begin{packagecode}
\newcommand*{\RemoveDummyPatternArg}[2]{%
  \expandafter #1\expandafter {\RemoveDummyPattern #2}}
\end{packagecode}
|\RemoveTilde| is used to remove the tilde that separated
the dummy pattern from <split1>.
\begin{packagecode}
% %% An alternative policy is to pass
% %% <target> (as an argument) to the parsing macro.
\newcommand{\RemoveTilde}{} \def\RemoveTilde#1~{#1}
\end{packagecode}
|\RemoveTildeArg<macro>{<arg>}| executes `\RemoveTilde'
in the next argument:
\begin{packagecode}
\newcommand*{\RemoveTildeArg}[2]{%
  \expandafter #1\expandafter {\RemoveTilde #2}}
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  Calling conditionals \unskip }

|\ProcessStringWith{<target-string>}{<id>}| builds the sandbox
to search <target-string> for the <pattern> associated with the
parser-conditional that is identified by <id>, the sandbox then
calls the parser.
\begin{packagecode}
\newcommand*{\ProcessStringWith}[2]{%
  \csname \setup_substr_cond #2\endcsname{#1}}
\end{packagecode}
|\ProcessExpandedWith{<string-macro>}{<id>}| does the same but with
a \emph{macro} (like `\fdInputLine' or `\OutputString') in which
the string to be tested is stored.
\begin{packagecode}
\newcommand*{\ProcessExpandedWith}[2]{%
  \csname \setup_substr_cond #2\expandafter \endcsname
    \expandafter{#1}}
\end{packagecode}
I would have preferred the reversed order of arguments which seems
to be more natural, but the present is more efficient.
Macros with reversed order are currently stored after `\endinput'
in section~\ref{sec:pondered}, may be they once return.

Anyway, most desired will be |\ProcessInputWith{<id>}| just
applying to `\fdInputLine':
\begin{packagecode}
\newcommand*{\ProcessInputWith}[1]{%
  \csname \setup_substr_cond #1\expandafter \endcsname
    \expandafter{\fdInputLine}}
\end{packagecode}
(Definition almost copied for efficiency.)
\begin{packagecode}
  %% TODO: error when undefined 2009/04/07
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  Copy jobs \unskip }

A job identifier <id> may also be considered a mere \emph{hook},
a \emph{placeholder} for a parsing job. What function actually is
called may depend on conditions that change while reading the
<input> file. %%% On a certain condition,
|\CopyFDconditionFromTo{<id1>}{<id2>}|
\emph{creates or redefines a sandbox builder} with identifier <id2>
that afterwards behaves like the sandbox builder <id1>.
So you can store a certain behaviour as <id1> in advance in order
once to change the behaviour of <id2> into that of <id1>.
\begin{packagecode}
\newcommand*{\CopyFDconditionFromTo}[2]{%
  \expandafter \let
    \csname \setup_substr_cond #2\expandafter \endcsname
      \csname \setup_substr_cond #1\endcsname}
\end{packagecode}
(Only the \emph{sandbox} is copied here---what about
changing conditionals?) %% TODO

An ``almost'' example is typesetting documentation from a package
file where the ``Legalese'' header might be typeset verbatim
although it is marked as ``comment.'' (The present %% 2009/04/07
example changes ``hand-made'' macros instead.)

This feature could have been placed more below as a ``programming
tool.''


\section{\hspace{1sp}\ignorespaces  Programming tools \unskip }


\subsection{\hspace{1sp}\ignorespaces  Tails of conditionals \unskip }

\label{sec:tails}
When creating complex \emph{expandable} conditionals,
this may amount to have primitive `\if' $\dots$ `\fi' conditionals
nested quite deeply, once perhaps too deep for \TeX's memory.
To avoid this, you can apply the common `\expandafter' trick
which finishes the current `\if' $\dots$ `\fi' before an inside
macro is executed (cf.\  \TeX book p.~219 on ``tail recursion'').

Internally tests whether certain strings are present at certain
places will be carried out by tests on emptiness or
onwards) on starting with `~'. E.g.,
``#1~=~<split1> empty'' indicates that either the <pattern>
starts a line or the line is empty altogether (this must be
decided by another test).

|\IfFDempty{<arg>}{<when-empty>}{<when-not-empty>}|
is used to test <arg> on emptyness (without expanding it):
\begin{packagecode}
\newcommand*{\IfFDempty}[1]{%
  \ifx$#1$\expandafter \@firstoftwo  \else
          \expandafter \@secondoftwo \fi}
\end{packagecode}
|\IfFDinputEmpty{<when-empty>}{<when-not-empty>}| is a variant of
the previous to execute <when-empty> if the loop processing <input>
finds an empty line---otherwise <when-not-empty>.
\begin{packagecode}
\newcommand*{\IfFDinputEmpty}{%
  \ifx\fdInputLine\@empty \expandafter \@firstoftwo  \else
                          \expandafter \@secondoftwo \fi}
\end{packagecode}
|\IfFDdollar{<arg>}{<when-empty>}{<when-not-empty>}|
is another variant, testing <split2> for being `$',
main indicator of there is a match anywhere in <target>
(as opposed to starting or ending match):
\begin{packagecode}
\newcommand*{\IfFDdollar}[1]{%
  \ifx$#1\expandafter \@firstoftwo  \else
         \expandafter \@secondoftwo \fi}
\end{packagecode}
It is exemplified and explained in section~\ref{sec:replchain}.
(The whole policy requires that `~' remains active in any
 testing macros here!)

However, you might always just type the replacement text
(in one line) instead of
such an `\If'\,$\dots$ (for efficiency \dots)

If expandability is not desired, you can just chain macros that
rework (so re-define) `\OutputString' or so.

2009/04/11: tending towards combining $\dots$
Keeping empty input and empty arguments apart is useful in that
\emph{one} test of emptiness per input line should suffice---it
may be left open whether this should be the first of all tests
\dots


\subsection{\hspace{1sp}\ignorespaces  Line counter \unskip }

A \LaTeX\ counter |fdInputLine| may be useful for screen or log
messages, moreover you can use it to control processing of the
<input> file ``from outside,'' not dependent on what the parsing
macros find. The header of the file might be typeset verbatim,
but we may be too lazy to define the ``header'' in terms of
what is in the file. We just decide that the first $\dots$ lines
are the ``header,'' even without counting just trying whether
the output is fine. It may be necessary to change that number
manually when the header changes.

You also can insert lines in <output>
which have no counterpart in <input>---if you know what you are
doing. With 'makedoc', there is a hook `\EveryComment' that can
be used to issue commands ``from outside'' at a place where
executing the command is safe or appropriate.
\begin{packagecode}
\newcounter{fdInputLine}
\end{packagecode}
You then must insert |\CountInputLines| in the second argument
of `\ProcessFileWith' (or in a macro called from there)
so that the counter is stepped. %% TODO!? 2009/04/07
\begin{packagecode}
\newcommand*{\CountInputLines}{\global\advance\c@fdInputLine\@ne}
\end{packagecode}
At present %% 2009/04/07 TODO
the counter is reset by `\ProcessFileWith', this may change.

|\IfInputLine{<relation><number>}{<true>}{<false>}|, when called
from the processing loop (second argument of `\ProcessFileWith')
issues <true> commands if `\value{fdInputLine}<relation><number>'
is true, otherwise <false>. <relation> may usually be just `='.
\begin{packagecode}
\newcommand*{\IfInputLine}[1]{%
  \ifnum\c@fdInputLine#1\relax \expandafter \@firstoftwo
  \else                        \expandafter \@secondoftwo \fi}
\end{packagecode}

\subsection{\hspace{1sp}\ignorespaces  The ``identity job'' `LEAVE' \unskip }

\label{sec:LEAVE}
The job with identifier |LEAVE|
\emph{leaves} an (expandable) chain of jobs
(as expandable replacement in section~\ref{sec:replchain})
and \emph{leaves} the processed string without changing it
and without the braces enclosing it:
\begin{packagecode}
\expandafter \let
  \csname \setup_substr_cond LEAVE\endcsname \@firstofone
\end{packagecode}
I.e., `\ProcessStringWith{<string>}{LEAVE}' expands to <string>
$\dots$ \ProcessStringWith{(Indeed!)}{LEAVE}


\section{\hspace{1sp}\ignorespaces  Setup for expandable chains of replacements \unskip }

\label{sec:replchain}%% TODO makedoc: provide less visible label/ref 2009/04/11
By the following means, you can create macros
(`\Transform' among them) such that, e.g.,
\[`\edef\OutputString{\Transform{<string>}}'\] renders `\OutputString'
the result of applying a chain (sequence) of stringwise replacements
to <string>.
You can even write a transformed input <string> to a file
without defining anything anything after `\read to'\,.\,.\,.
In this case however, you don't get any statistical message
about what happened or not. With `\edef\OutputString' you can at
least issue some `changed!' or `left!' (maybe `\message{!}' vs.\
`\message{.}').
There is an application in 'makedoc' for ``typographical upgrading''
from plain text to \TeX\ input.

|\repl_all_chain_expandable| will be the backbone of the
replacements. It is called by some parsing macro <parser>
and receives from the latter <split1>~=~#1 and <split2>~=~#2.
#3 is the result of what happened so far.
\begin{packagecode}
\def\repl_all_chain_expandable#1#2#3#4#5#6{%
  %% #1, #2 splits, #3 past,   #4 substitute,
  %% #5 repeat parser,         #6 pass to
%   \ifx~#2\expandafter\@firstoftwo\else\expandafter\@secondoftwo\fi
\end{packagecode}
The previous line would be somewhat faster, but let us exemplify
`\IfFDdollar' from section~\ref{sec:tails} instead:
\begin{packagecode}
  \IfFDdollar{#2}%
\end{packagecode}
If #2 starts with `$'---with category code 3, ``math shift''!,
it \emph{is} `$', due to not reading `$'
from input with its standard category code 3
and the sandbox construction (where `$' appears with its standard
category code). %% TODO might be explained earlier 2009/04/11
                %%      or refer to here.
And this is the case \emph{exactly} when the <pattern> from
<parser> didn't match, again due to the input category codes.
Now on \emph{no} match, the sandbox builder #6 is called
with target string #3#1 where the last tested string is attached
to previous results. The ending `~' is removed, #6 inserts a new
wrap for the new dummy pattern.
\begin{packagecode}
             {\RemoveTildeArg #6{#3#1}}%
\end{packagecode}
Otherwise $\dots$ the \emph{sandbox builder} <sandbox>
(that will be shown below) that called <parser>
initialized #5 to be that <parser> itself. (<parser> otherwise
wouldn't know who it is.) So <parser> calls itself with another sandbox
`#2&'. Note that #2 contains \lq`~<pattern>$'\rq\ due to the initial
<sandbox> building.
\begin{packagecode}
             {#5#2&{#3#1#4}{#4}#5#6}}
\end{packagecode}
#4 is the replacement string that <sandbox> passed to <parse>.
The first argument after the `&' is previous stuff plus
the recently skipped <split1>
plus #4 replacing the string <pattern> that was matched.

Finally, #5 and #6 again ``recall'' <parser> and the sandbox
builder to which to change in case of no other match.

% TODO move following up!? 2009/04/11
|\MakeExpandableAllReplacer{<id>}{<pattern>}{<replace>}{<id-next>}|
creates sandbox and parser with common identifier <id> and search
pattern <pattern>. Each occurrence of <pattern> will be replaced by
<replace>. When <pattern> is not found, the sandbox builder for
<id-next> is called. This may be another replacing macro of the
same kind. To return the result without further transformations,
call job `LEAVE' (section~\ref{sec:LEAVE}).
\begin{packagecode}
\newcommand*{\MakeExpandableAllReplacer}{%
  \afterassignment\mk_setup_xpdbl_all_repl_san
  \def\cond_id}
\end{packagecode}
$\dots$ usual intermezzo for reading patterns with non-standard
category codes, this time we read \emph{two} patterns \dots
\begin{packagecode}
\newcommand*{\mk_setup_xpdbl_all_repl_san}[1][]{%
  \begingroup \PatternCodes #1\mk_setup_xpdbl_all_repl}
\end{packagecode}
Here comes the real work.
\begin{packagecode}
\newcommand*{\mk_setup_xpdbl_all_repl}[3]{%
  %% #1 pattern, #2 substitute, #3 pass to
  \endgroup
\end{packagecode}
We take pains to call next jobs by single
command strings and store them this way, not by `\csname',
as `\ProcessStringWith' would do it. `\edef\@tempa'
is used for this purpose, but \dots
\begin{packagecode}
    \edef\@tempa{%
      \noexpand\mk_setup_substr_cond{#1}{%
        {}{#2}%
        \noexpand\noexpand
\end{packagecode}
That `\edef\@tempa' must \emph{not expand} the controll
words after they have been computed from `\csname' etc.
Moreover, expansion of the parser commands
must be avoided another time, when `\@tempa' is executed.
\begin{packagecode}
          \make_not_expanding_cs{\substr_cond\cond_id}%
        \noexpand\noexpand
          \make_not_expanding_cs{\setup_substr_cond #3}}}%
\end{packagecode}
Those internal setup commands start with `\endgroup' to switch back
to standard category codes. We must match them here by
`\begingroup'.
\begin{packagecode}
    \begingroup \@tempa
    \begingroup \mk_substr_cond{#1}{%
      \repl_all_chain_expandable{##1}{##2}}}
\end{packagecode}
The final command is the one that we explained first. %% TODO 2009/04/11

Support for dozens of replacements in one sequence
and for screen messages
must wait for another release, sorry! %% TODO 2009/04/11


\section{\hspace{1sp}\ignorespaces  Leave package mode \unskip }

We restore the underscore `_' for math subscripts.
(This might better depend on something \dots) %% TODO 2009/04/07
\begin{packagecode}
\catcode`\_=8   %% restores underscore use for subscripts

\endinput
\end{packagecode}
\TeX\ ignores the rest of the file when it is \emph{input}
``in the sense of `\input''', as opposed to just reading
the file line by line to a macro like `\fdInputLine'.


\section{\hspace{1sp}\ignorespaces  Pondered \unskip }

\label{sec:pondered}
\begin{packagecode}
  %% TODO abbreviated commands (aliases) \MkSubstrCond...
  %% TODO \@onlypreamble!?
\newcommand*{\ApplySubstringConditional}[1]{%
  %% #1 identifier; text to be searched expected next
  \csname setup_substr_cond:#1\endcsname}
\newcommand*{\ApplySubstringConditionalToExpanded}[1]{% 2009/03/31+
  \csname setup_substr_cond:#1\expandafter \endcsname \expandafter}
\newcommand*{\ApplySubstringConditionalToInputString}[1]{% 2009/03/31+
  \csname setup_substr_cond:#1\expandafter \endcsname
    \expandafter {\fdInputLine}}
  %% TODO or `\OutputString', even `\read' to `\OutputString'!?
% \newcommand*{\ApplySubstringConditionalToExpanded}[2]{%
%   %% note: without assignments, robust!
%   %% BUT the `\csname ... \expandafter \endcsname' method is faster
%   \expandafter \reversed_apply_substr_cond
%     \expandafter {#2}{#1}}
% \newcommand*{\reversed_apply_substr_cond}[2]{%
%   \ApplySubstringConditional{#2}{#1}}
  %% ODER:
% \newcommand*{\expand_attach_arg}[2]{%% 2009/03/31
%   %% #1 command with previous args, TODO cf. LaTeX3
%   \expandafter \attach_arg \expandafter {#1}{#2}}
%   %% actually #1 may contain more than one token,
%   %% only first expanded
% \newcommand*{\attach_arg}[2]{#2{#1}}
% \newcommand*{\ApplySubstringConditionalToExpanded}[2]{%
%   \expandafter \attach_arg \expandafter
%     {#2}{\ApplySubstringConditional{#1}}}
\end{packagecode}

\section{\hspace{1sp}\ignorespaces  VERSION HISTORY \unskip }


\begin{packagecode}
v0.1   2009/04/03   very first version, tested on morgan.sty
v0.2   2009/04/05   counter fdInputLine, \ProvidesFile moved from
                    \ProcessFile to \ResultFile, \CopyFD...,
                    category section first, more sectioning,
                    suppressing empty code lines before section
                    titles; discussion, \Delimiters
       2009/04/06   more discussion
       2009/04/07   more discussion, factored \WriteProvides out from
                    \ResultFile, \ProcessExpandedWith corrected
       2009/04/08   \InputString -> \fdInputline;
                    removed \ignorespaces
       2009/04/09   \WhenInputLine[2] -> \IfInputline[3],
                    \ProcessInputWith, typos,
                    \WriteProvides message `with'
       2009/04/10   \make_not_expanding_cs
                    DISCOVERED ``IF SUBSTRING'' ALGORITHM WRONG
                    (<str1><str2><str1> in <str1><str2>)
v0.3   2009/04/11   SOME THINGS GIVEN UP EARLIER WILL BE REMOVED,
                    TO BE STORED IN THE COPY AS OF 2009/04/10
                    mainly: sandbox setup (tilde/dollar)
                    REAL ADDITION: setup for expandable replacing
       2009/04/12   played with `chain' vs. `sequence';
                    plain `...', `cf.', `etc.' for `mdcorr.cfg'
       2009/04/13   \RemoveTilde...
       2009/04/15   reworked text, same mistake \in@

\end{packagecode}