summaryrefslogtreecommitdiff
path: root/macros/latex/contrib/polytable
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/polytable
Initial commit
Diffstat (limited to 'macros/latex/contrib/polytable')
-rw-r--r--macros/latex/contrib/polytable/README14
-rw-r--r--macros/latex/contrib/polytable/polytable.dtx2378
-rw-r--r--macros/latex/contrib/polytable/polytable.ins5
-rw-r--r--macros/latex/contrib/polytable/polytable.pdfbin0 -> 288334 bytes
4 files changed, 2397 insertions, 0 deletions
diff --git a/macros/latex/contrib/polytable/README b/macros/latex/contrib/polytable/README
new file mode 100644
index 0000000000..88ba930183
--- /dev/null
+++ b/macros/latex/contrib/polytable/README
@@ -0,0 +1,14 @@
+This is the polytable package. The license is LPPL.
+
+The package depends on lazylist.sty, as available from
+"CTAN:/macros/latex/contrib/lazylist.sty".
+
+To install and get documentation, run the following commands:
+
+latex polytable.ins
+latex polytable.dtx
+latex polytable.dtx
+
+Please report bugs to "polytable@andres-loeh.de".
+
+Andres Loeh, November 2005
diff --git a/macros/latex/contrib/polytable/polytable.dtx b/macros/latex/contrib/polytable/polytable.dtx
new file mode 100644
index 0000000000..fb9219492b
--- /dev/null
+++ b/macros/latex/contrib/polytable/polytable.dtx
@@ -0,0 +1,2378 @@
+% \iffalse
+%<*driver>
+\ProvidesFile{polytable.drv}
+\documentclass{ltxdoc}
+\usepackage[info]{polytable}
+\begin{document}
+ \DocInput{polytable.dtx}
+\end{document}
+%</driver>
+%
+% Copyright (C) 2003--2013 by Andres Loeh
+%
+% \fi
+%
+% \ProvidesFile{polytable.dtx}
+% [2013/07/08 v0.8.5 `polytable' package (Andres Loeh)]
+% \GetFileInfo{polytable.dtx}
+%
+% \title{The \textsf{polytable} package}
+% \author{Andres L\"oh\\ \texttt{polytable@andres-loeh.de}}
+% \date{\filedate}
+%
+% \maketitle
+%
+% \begin{abstract}
+% This package implements a variant of tabular/tabbing-like environments
+% where columns can be given a name and entries can flexibly
+% be placed between arbitrary columns. Complex alignment-based
+% layouts, for example for program code, are possible.
+% \end{abstract}
+%
+% \changes{v0.1}{2002/11/17}{Works.}
+% \changes{v0.2}{2003/04/03}{Added pboxed. Added debug option.}
+% \changes{v0.3}{2003/05/26}{Greatly enhanced pboxed.}
+% \changes{v0.4}{2003/06/11}{Optional argument for ptable and parray.}
+% \changes{v0.4.1}{2003/07/13}{Saving and restoring enhancements.}
+% \changes{v0.4.2}{2003/07/16}{Fixed bug that caused infinite reruns.}
+% \changes{v0.4.3}{2003/07/20}{Added some missing relaxes.}
+% \changes{v0.5}{2003/07/22}{Documentation improved. To be put on CTAN.}
+% \changes{v0.6}{2003/07/24}{Depends on lazylist instead of lambda now.}
+% \changes{v0.7}{2003/11/20}{Check for undefined columns. Save counters.}
+% \changes{v0.7.1}{2003/12/26}{Counters have not been restored correctly.}
+% \changes{v0.7.2}{2004/02/27}{Fix for conflict with amsmath package.}
+% \changes{v0.7.3}{2004/06/21}{Balance last line in pboxed environment.}
+% \changes{v0.8}{2004/06/24}{New environments pmboxed, ptboxed.
+% Removed original ptabular and parray.
+% Interface overhaul.}
+% \changes{v0.8.1}{2004/11/11}{Fixed some problems with the new
+% interface, added the external (disk) queue.}
+% \changes{v0.8.2}{2005/04/26}{New interface command.
+% Minor beautifications.}
+% \changes{v0.8.3}{2009/04/28}{Make sorting algorithm stable.}
+% \changes{v0.8.4}{2009/11/01}{Make ptb output more stable.}
+% \changes{v0.8.5}{2013/07/18}{Work around lazylist name clash
+% with unicode-math.}
+%
+% \section{Introduction}
+%
+% This package implements a variant of tabular-like environments.
+% We will call these environments the |poly|-environments to
+% distinguish them from the standard ones as provided by the
+% \LaTeX\ kernel or the \textsf{array} package.
+%
+% Other than in standard tables, each column has a name. For
+% instance, the commands\\
+% |\column{foo}{l}|\\
+% |\column{bar}{r}|\\
+% -- when used within a |poly|-environment -- define a column
+% with name |foo| that is left-aligned, and a column with name
+% |bar| that is right-aligned.
+%
+% Once a couple of columns have been defined, the text is specified
+% in a series of |\fromto| commands. Instead of specifying text
+% per column in order, separating columns with |&|, we give the
+% name of the column where the content should start, and the name
+% of the column before which the content should stop.
+% To typeset the text ``I'm aligned!'' in the column |foo|,
+% we could thus use the command\\
+% |\fromto{foo}{bar}{I'm aligned}|\\
+% Several |\fromto|-commands can be used to typeset a complete
+% line of the table. A new line can be started with |\\|.
+%
+% The strength of this approach is that it implicitly handles
+% cases where different lines have different alignment properties.
+% Not all column names have to occur in all lines.
+%
+% \section{A complete example}
+%
+% Figure~\ref{tab:example}
+% is an example that is designed to show the
+% capabilities of this package. In particular, it is \emph{not}
+% supposed to look beautiful.
+%
+% \begin{figure}[htb]
+% \centering
+% \begin{ptboxed}
+% \defaultcolumn{l|}\column{.}{|l|}
+% \> left
+% \=3 first of three \> second of three \> third of three
+% \=r right \\
+% \defaultcolumn{r|}\> left \=2 middle 1/2 \> middle 2/2 \=r right \\
+% \> left \=3 middle 1/3 \> middle 2/3 \> middle 3/3 \=r right \\
+% \> left
+% \=2 first of two middle columns \> second of two middle columns
+% \=r right \\
+% \end{ptboxed}
+% \caption{Example table}%
+% \label{tab:example}
+% \end{figure}
+%
+% The example table consists of four lines. All lines have
+% some text on the left and on the right, but the middle part
+% follows two different patterns: the first and the third line
+% have three middle columnss that should be aligned, the second
+% and the fourth line have two (right-aligned) middle columns
+% that should be aligned, but otherwise independent of the three
+% middle columns in the other lines.
+%
+% Vertical bars are used to clarify where one column ends and
+% the next column starts in a particular line. Note that the first
+% and the third line are completely aligned. Likewise, the
+% second and the fourth line are. However, the fact that the
+% bar after the text ``middle 1/2'' ends up between the two
+% bars delimiting the column with ``second of three'' in it
+% is just determined by the length of the text ``first of two
+% middle columns'' in the last line. This text fragment
+% is wider than the first of the three middle columns, but not
+% wider than the first two of the three middle columns.
+%
+% Let's have a look at the input for the example table:
+% \begin{verbatim}
+% \begin{ptboxed}
+% \defaultcolumn{l|}\column{.}{|l|}
+% \> left
+% \=3 first of three \> second of three \> third of three
+% \=r right \\
+% \defaultcolumn{r|}\> left \=2 middle 1/2 \> middle 2/2 \=r right \\
+% \> left \=3 middle 1/3 \> middle 2/3 \> middle 3/3 \=r right \\
+% \> left
+% \=2 first of two middle columns \> second of two middle columns
+% \=r right \\
+% \end{ptboxed}
+% \end{verbatim}
+% First, columns are declared, including the vertical lines.
+% Note that there is a final column |end| being declared
+% that is only used as the end column in the |\fromto| statements.
+% A future version of this package
+% will probably get rid of the need to define
+% such a column.
+% After the column definitions, the lines are typeset by a
+% series of |\fromto| commands, separated by |\\|.
+% Note that the first and third column do not use |m12|, |m22|.
+% Similarly, the second and fourth column do not use
+% |m13|, |m23|, and |m33|.
+%
+% So far, one could achieve the same with an ordinary |tabular|
+% environment. The table would have 6 columns. One left and right,
+% the other four for the middle: the first and third line would
+% use the first of the four columns, then place the second entry
+% in a |\multicolumn| of length 2, and then use the fourth column
+% for the third entry. Likewise, the other lines would place both
+% their entries in a |\multicolumn| of length 2. In fact, this
+% procedure is very similar to the way the |ptabular| environment
+% is implemented.
+%
+% The problem is, though, that we need the information that
+% the first of the two middle columns ends somewhere in the middle
+% of the second of the three columns, as observed above. If we
+% slightly modify the texts to be displayed in the middle columns,
+% this situation changes. Figure~\ref{tab:variation} shows
+% two variants of the example table. The input is the same,
+% only that the texts contained in some columns have slightly
+% changed. As you can see, the separator between the first and
+% second middle column in the second and fourth lines of the
+% tables now once ends up within the first, once within the
+% third of the three middle columns of the other lines.
+%
+% \begin{figure}[htb]
+% \centering
+% \begin{ptboxed}
+% \column{left}{|l|}
+% \column{right}{l|}
+% \column{m13}{l|}
+% \column{m23}{l|}
+% \column{m33}{l|}
+% \column{m12}{r|}
+% \column{m22}{r|}
+% \column{end}{l}
+% \fromto{left}{m13}{left}
+% \fromto{m13}{m23}{first of three}
+% \fromto{m23}{m33}{second of three}
+% \fromto{m33}{right}{third of three}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m12}{left}
+% \fromto{m12}{m22}{middle 1/2}
+% \fromto{m22}{right}{middle 2/2}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m13}{left}
+% \fromto{m13}{m23}{middle 1/3}
+% \fromto{m23}{m33}{middle 2/3}
+% \fromto{m33}{right}{middle 3/3}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m12}{left}
+% \fromto{m12}{m22}{first of two}
+% \fromto{m22}{right}{second of two}
+% \fromto{right}{end}{right}
+% \end{ptboxed}
+% \par\bigskip
+% \begin{ptboxed}
+% \column{left}{|l|}
+% \column{right}{l|}
+% \column{m13}{l|}
+% \column{m23}{l|}
+% \column{m33}{l|}
+% \column{m12}{r|}
+% \column{m22}{r|}
+% \column{end}{l}
+% \fromto{left}{m13}{left}
+% \fromto{m13}{m23}{first of three}
+% \fromto{m23}{m33}{second of three}
+% \fromto{m33}{right}{third of three}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m12}{left}
+% \fromto{m12}{m22}{middle 1/2}
+% \fromto{m22}{right}{middle 2/2}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m13}{left}
+% \fromto{m13}{m23}{middle 1/3}
+% \fromto{m23}{m33}{middle 2/3}
+% \fromto{m33}{right}{middle 3/3}
+% \fromto{right}{end}{right}
+% \\
+% \fromto{left}{m12}{left}
+% \fromto{m12}{m22}{the first of two middle columns}
+% \fromto{m22}{right}{2/2}
+% \fromto{right}{end}{right}
+% \end{ptboxed}
+% \caption{Variants of the example table}%
+% \label{tab:variation}
+% \end{figure}
+%
+% If one wants the general case using the |\multicolumn|
+% approach, one thus has to measure the widths of the entries
+% of the columns to compute their relative position. In essence,
+% this is what the package does for you.
+%
+% \section{Haskell code example}
+%
+% I have written this package mainly for one purpose: to be able
+% to beautifully align Haskell source code. Haskell is a functional
+% programming language where definitions are often grouped into
+% several declarations. I've seen programmers exhibit symmetric
+% structures in different lines by adding spaces in their source
+% code files in such a way that corresponding parts in different
+% definitions line up. On the other hand, as Haskell allows
+% user-defined infix operators, some programmers like their symbols
+% to be typeset as \LaTeX\ symbols, not as typewriter code.
+% But using \LaTeX\ symbols and a beautiful proportional font
+% usually destroys the carefully crafted layout and alignment.
+%
+% With lhs2\TeX, there is now a preprocessor available that preserves
+% the source code's internal alignment by mapping the output onto
+% \textsf{polytable}'s environments.
+% Figure~\ref{tab:haskell} is an example of how
+% the output of lhs2\TeX\ might look like.
+%
+% \newcommand{\HsVar}[1]{\mathop{\mathit{\strut #1}}}
+% \newcommand{\HsCon}[1]{\mathop{\mathsf{\strut #1}}}
+% \newcommand{\HsKey}[1]{\mathop{\mathbf{\strut #1}}}
+% \newcommand{\map}{\HsVar{map}}
+% \newcommand{\round}{\HsVar{round}}
+% \newcommand{\TA}{\HsVar{a}}
+% \newcommand{\TB}{\HsVar{b}}
+% \newcommand{\TF}{\HsVar{f}}
+% \newcommand{\TX}{\HsVar{x}}
+% \newcommand{\TY}{\HsVar{y}}
+% \newcommand{\TN}{\HsVar{n}}
+% \newcommand{\TM}{\HsVar{m}}
+% \newcommand{\TR}{\HsVar{r}}
+% \newcommand{\TXS}{\HsVar{xs}}
+% \newcommand{\RealFrac}{\mathop{\mathsf{RealFrac}}}
+% \newcommand{\Integral}{\mathop{\mathsf{Integral}}}
+% \newcommand{\LET}{\mathop{\mathbf{let}}}
+% \newcommand{\IN}{\mathop{\mathbf{in}}}
+% \newcommand{\IF}{\mathop{\mathbf{if}}}
+% \newcommand{\THEN}{\mathop{\mathbf{then}}}
+% \newcommand{\ELSE}{\mathop{\mathbf{else}}}
+% \newcommand{\OF}{\mathop{\mathbf{of}}}
+% \newcommand{\CASE}{\mathop{\mathbf{case}}}
+% \newcommand{\CLASS}{\mathop{\mathbf{class}}}
+% \newcommand{\WHERE}{\mathop{\mathbf{where}}}
+% \newcommand{\Eq}{\mathop{\mathsf{Eq}}}
+% \newcommand{\Ord}{\mathop{\mathsf{Ord}}}
+%
+% \setlength{\arraycolsep}{.5ex}
+%
+% \begin{figure}
+% \[
+% \begin{pmboxed}
+% \column{left}{l}
+% \column{body}{l}
+% \column{end}{l}
+% \column{rel}{c}
+% \column{rhs}{l}
+% \>[left] \CLASS\ (\Eq\TA)\Rightarrow\Ord\TA\WHERE \\
+% \>[left] \qquad\
+% \>[body] \HsVar{compare}
+% \>[rel] ::
+% \>[rhs] \TA\rightarrow\TA\rightarrow\HsCon{Ordering} \\
+% \>[left] \qquad\
+% \>[body] (<),(\leq),(\geq),(>)
+% \>[rel] ::
+% \>[rhs] \TA\rightarrow\TA\rightarrow\HsCon{Bool}
+% \\
+% \>[left] \qquad\
+% \>[body] \HsVar{max},\HsVar{min}
+% \>[rel] ::
+% \>[rhs] \TA\rightarrow\TA\rightarrow\HsCon{Bool}
+% \\
+% \\
+% \>[body] \mbox{--- Minimal complete definition:
+% $(\leq)$ or $\HsVar{compare}$}
+% \\
+% \>[body] \mbox{--- using $\HsVar{compare}$ can be more efficient
+% for complex types}
+% \\
+% \column{cgrd}{l}
+% \>[body] \HsVar{compare}\TX\TY
+% \>[cgrd] \vert\ \TX\equiv\TY
+% \>[rel] =
+% \>[rhs] \HsCon{EQ}
+% \\
+% \>[cgrd] \vert\ \TX\leq\TY
+% \>[rel] =
+% \>[rhs] \HsCon{LT}
+% \\
+% \>[cgrd] \vert\ \HsVar{otherwise}
+% \>[rel] =
+% \>[rhs] \HsCon{GT}
+% \\
+% \>[body] \TX\leq\TY
+% \>[rel] =
+% \>[rhs] \HsVar{compare}\TX\TY\not\equiv\HsCon{GT}
+% \\
+% \>[body] \TX < \TY
+% \>[rel] =
+% \>[rhs] \HsVar{compare}\TX\TY\equiv\HsCon{LT}
+% \\
+% \>[body] \TX\geq\TY
+% \>[rel] =
+% \>[rhs] \HsVar{compare}\TX\TY\not\equiv\HsCon{LT}
+% \\
+% \>[body] \TX > \TY
+% \>[rel] =
+% \>[rhs] \HsVar{compare}\TX\TY\equiv\HsCon{GT}
+% \\
+% \>[body] \HsVar{max}\TX\TY
+% \>[cgrd] \vert\ \TX\leq\TY
+% \>[rel] =
+% \>[rhs] \TY
+% \\
+% \>[cgrd] \vert\ \HsVar{otherwise}
+% \>[rel] =
+% \>[rhs] \TX
+% \\
+% \>[body] \HsVar{min}\TX\TY
+% \>[cgrd] \vert\ \TX\leq\TY
+% \>[rel] =
+% \>[rhs] \TX
+% \\
+% \>[cgrd] \vert\ \HsVar{otherwise}
+% \>[rel] =
+% \>[rhs] \TY
+% \end{pmboxed}
+% \]
+% \caption{Haskell code example}
+% \label{tab:haskell}
+% \end{figure}
+%
+% Of course, this could be useful for other programming languages
+% as well. In fact, lhs2\TeX\ can be tweaked to process several
+% experimental languages that are based on Haskell, but I can imagine
+% that this package could generally prove useful to typeset program
+% code.
+%
+% \section{Other applications}
+%
+% Although I have written this package for a specific purpose, I
+% am very much interested to hear of other potential application
+% areas. Please tell me if you found a use for this package and
+% do not hesitate to ask for additional features that could convince
+% you to use the package for something.
+%
+% \section{The \textsf{lazylist} package}
+%
+% Internally, this package makes use of Alan Jeffrey's excellent
+% \textsf{lazylist} package, which provides an implementation of
+% the lambda calculus using fully expandable control sequences.
+% Unfortunately, |lazylist.sty| is not included in most common
+% \TeX\ distributions, so you might need to fetch it from CTAN
+% separately.
+%
+% \section{Reference}
+%
+% \subsection{Setup}
+%
+% New in v0.8: We allow to implicitly define columns by just
+% using the names during table content specification, without
+% having declared them formally using |\column| (see below).
+%
+% \DescribeMacro{\nodefaultcolumn}
+% By default, though, this behaviour is turned off, because the
+% use of a misspelled column is often an error. Thus, by default
+% or after the command |\nodefaultcolumn| is used, all columns
+% must be declared.
+%
+% \DescribeMacro{\defaultcolumn}
+% Using a statement of the form
+% |\defaultcolumn{|\meta{spec}|}|, implicit columns can
+% be activated. All undefined columns that are encountered
+% will be assumed to have format string \meta{spec}.
+%
+% \subsection{The environments}
+%
+% \DescribeEnv{pboxed}
+% \DescribeEnv{ptboxed}
+% \DescribeEnv{pmboxed}
+% \DescribeEnv{ptabular}
+% \DescribeEnv{parray}
+% New in v0.8: There are now five environments that this package
+% provides: in addition to the former |ptabular|, |parray|, and
+% |pboxed|, there is now also |ptboxed| and |pmboxed|.
+% The environment |pboxed|
+% typesets the material in boxes of the calculated length, but in
+% normal paragraph mode. The advantage is that
+% there can be page breaks within the table.
+% Note that you should start a new (probably nonindented)
+% paragraph before beginning a |pboxed|. All lines that a |pboxed|
+% produces are of equal length, so it should be possible to center
+% or right-align the material.
+
+% Both |ptboxed| and |pbmoxed| are like |pboxed|, but
+% pre-wrapped in a |tabular| or |array| environment,
+% respectively, and
+% thus not page-breakable but less fragile (or should I just
+% say: easier to use) than the unwrapped |pboxed|.
+% With those, there
+% is no need for |ptabular| and |parray| anymore --
+% which were directly based on (and translated into) usual
+% |tabular| and |array| environments as provided by the |array|
+% package. The environments are still supported, to ensure
+% compatibility, but they are mapped to |ptboxed| and |pmboxed|,
+% respectively.
+%
+% The |pmboxed| and |parray| environments assume math mode, whereas
+% |ptboxed| and |ptabular| assume text mode. The |pboxed| environment
+% works in both text and math modes and adapts to whatever is
+% currently active.
+%
+% I wrote in previous versions that one additional environment, namely
+% |plongtable|, a poly-version of the |longtable| environment, was planned.
+% Page-breaking of |pboxed| works fine, and I do not see a
+% real need for a |plongtable| environment anymore. If you would like
+% it anyway, feel free to mail me and inform me about the advantages
+% it would have over |pboxed|.
+%
+% The interface is the same for all of the environments.
+%
+% \subsection{The interface}
+%
+% \begin{center}
+% \textbf{(Note: this has changed significantly in v0.8!)}
+% \end{center}
+%
+% In each of the environments, the following commands can be
+% used. All other commands should be used with care if placed
+% directly inside such an environment: the contents of a polytable
+% are processed multiple times; hence, if your commands generate
+% output, the output will end up multiple times in the final
+% document, and destroy your layout.
+%
+% \DescribeMacro{\column}
+% With |\column[|\meta{dimen}|]{|\meta{columnid}|}{|\meta{spec}|}|,
+% a new column
+% \meta{columnid} is specified. The name of the column can be
+% any sequence of alphanumerical characters. The \meta{spec} is
+% a format string for that particular column, and it can contain
+% the same constructs that can be used in format strings of normal
+% tables or arrays.
+% However, it must only contain the description for \emph{one}
+% column.
+%
+% As long as the save/restore feature (explained below) is not used,
+% |\column| definitions are local to one table.
+% One can define a column multiple times within one table:
+% a warning will be produced,
+% and the second format string will be used for the complete table.
+%
+% Columns must be defined before use when implicit columns using
+% |\defaultcolumn| are disabled.
+%
+% A minimal horizontal position \meta{dimen} (relative to the
+% beginning of the table) can be specified for a column, which
+% defaults to 0pt.
+%
+% \DescribeMacro{\=}
+% The command |\={|\meta{fromid}|}[|\meta{spec}|]| instructs the
+% package to prepare for a new entry starting at column \meta{fromid}.
+% Everything that follows the command, up to the next
+% interface-command (except |\column|) or the end of the environment
+% is interpreted as contents of the entry and can be arbitrary
+% \LaTeX\ code that has balanced grouping. The column specifier
+% \meta{spec} can be used to define a different formatting for this
+% particular entry. If the specifier starts with an exclamation mark
+% (|!|), it will be used as specifier for all entries of that column.
+% The use of multiple exclamation-mark specifiers for the same
+% column name gives precedence to the last one, but you should not
+% rely on this behaviour as it may change in the future.
+%
+% Note that, contrary to normal \LaTeX\ behaviour, the second argument
+% is the optional argument. Therefore, if an entry starts with an
+% opening bracket |([)|,
+% and the optional argument is omitted, a |\relax| should be inserted
+% between command and opening bracket, as otherwise the entry would
+% be treated as the optional argument.
+%
+% \DescribeMacro{\>}
+% The command |\>[|\meta{fromid}|][|\meta{spec}|]| is a variant of
+% |\=| where both arguments are optional. If no column name is given
+% then the current column name, postfixed by a dot (|.|), is assumed
+% as column name. At the beginning of a line, a single dot (|.|)
+% will be assumed as a name. The \meta{spec} argument has the same
+% behaviour as for |\=|. Note that if the specifier is given, the
+% column name must be given as well, but may be left empty if the
+% default choice is desired. For instance, |\>[][r]|, will move to
+% the next column, which will be typeset right-aligned.
+%
+% \DescribeMacro{\<}
+% The command |\<[|\meta{fromid}|]| ends the current entry at the
+% boundary specified by \meta{fromid}. This macro can be used instead
+% of |\>| or |\=| if an entry should be ended without immediately
+% starting a new one. Differences in behaviour occur if \meta{fromid}
+% is associated with a non-trivial column format string.
+% TODO: Improve this explanation.
+%
+% \DescribeMacro{\fromto}
+% The call |\fromto{|\meta{fromid}|}{|\meta{toid}|}{|\meta{text}|}|
+% will typeset \meta{text} in the current line, starting at column
+% \meta{fromid} and ending before column \meta{toid}, using the
+% format string specified for \meta{fromid}.
+%
+% A line of a table usually consists of multiple |\fromto| statements.
+% Each statement's starting column should be either the same as the
+% end column of the previous statement, or it will be assumed that
+% the start column is located somewhere to the right of the previous
+% end column. The user is responsible to not introduce cycles in the
+% (partial) order of columns. If such a cycle is specified, the
+% current algorithm will loop, causing a |dimension too large| error
+% ultimately. TODO: catch this error.
+%
+% \DescribeMacro{\\}
+% \DescribeMacro{\nextline}
+% The command |\\| (or, now deprecated, |\nextline|) ends one line
+% and begins the next.
+% There is no need to end the last line. One can pass an optional
+% argument, as in |\\[|\meta{dimen}|]|, that will add
+% \meta{dimen} extra space between the lines.
+%
+% \subsection{A warning}
+%
+% The contents of the table are processed multiple times because
+% the widths of the entries are measured. Global assignments that
+% modify registers and similar things can thus result in unexpected
+% behaviour.
+% New in v0.7: \LaTeX\ counters (i.e., counters defined by
+% |\newcounter|) are protected now. They will be reset after
+% each of the trial runs.
+%
+% \subsection{Saving column width information}
+%
+% Sometimes, one might want to reuse not only the same column, but
+% exactly the same alignment as in a previous table. An example
+% would be a fragment of program code, which has been broken into
+% several pieces, with documentation paragraphs added in between.
+%
+% \DescribeMacro{\savecolumns}
+% \DescribeMacro{\restorecolumns}
+% With |\savecolumns[|\meta{setid}|]|, one can save the information
+% of the current table for later reuse. The name |setid| can be an
+% arbitrary sequence of alphanumeric characters. It does \emph{not} share
+% the same namespace as the column names. The argument is optional;
+% if it is omitted, a default name is assumed. Later, one can restore
+% the information (multiple times, if needed) in other tables, by
+% issuing a |\restorecolumns[|\meta{setid}|]|.
+%
+% This feature requires to pass information backwards in the general
+% case, as column widths in later environments using one specific
+% column set might influence the layout of earlier environments.
+% Therefore, information is written into the |.aux| file, and
+% sometimes, a warning is given that a rerun is needed. Multiple
+% reruns might be required to get all the widths right.
+%
+% I have tried very hard to avoid producing rerun warnings infinitely
+% except if there are really cyclic dependencies between columns.
+% Still, if it happens or something seems to be broken, it often
+% is a good idea to remove the |.aux| file and start over. Be sure
+% to report it as a bug, though.
+%
+% Figure~\ref{tab:saverestore} is an example of the Haskell code
+% example with several comments inserted. The source of this file
+% shows how to typeset the example.
+% \begin{figure}
+% \setlength{\parindent}{0pt}%
+% \setlength{\parskip}{1ex}%
+% \hrule
+% We introduce a new type class $\Ord$ for objects that admit
+% an ordering. It is based on the $\Eq$ class:
+%
+% \(
+% \begin{pboxed}
+% \savecolumns
+% \column{left}{>{\quad}l}
+% \column{end}{l}
+% \fromto{left}{end}{\CLASS\ (\Eq\TA)\Rightarrow\Ord\TA\WHERE}
+% \end{pboxed}
+% \)
+%
+% The next three lines give the type signatures for all
+% the methods of the class.
+%
+% \(
+% \begin{pboxed}
+% \restorecolumns
+% \column{body}{l}
+% \column{rel}{c}
+% \column{rhs}{l}
+% \fromto{left}{body}{\qquad\ }
+% \fromto{body}{rel}{\HsVar{compare}}
+% \fromto{rel}{rhs}{::}
+% \fromto{rhs}{end}{\TA\rightarrow\TA\rightarrow\HsCon{Ordering}}
+% \\
+% \fromto{left}{body}{\qquad\ }
+% \fromto{body}{rel}{(<),(\leq),(\geq),(>)}
+% \fromto{rel}{rhs}{::}
+% \fromto{rhs}{end}{\TA\rightarrow\TA\rightarrow\HsCon{Bool}}
+% \\
+% \fromto{left}{body}{\qquad\ }
+% \fromto{body}{rel}{\HsVar{max},\HsVar{min}}
+% \fromto{rel}{rhs}{::}
+% \fromto{rhs}{end}{\TA\rightarrow\TA\rightarrow\HsCon{Bool}}
+% \\[1ex]
+% \fromto{body}{end}{\mbox{--- Minimal complete definition:
+% $(\leq)$ or $\HsVar{compare}$}}
+% \\
+% \fromto{body}{end}{\mbox{--- using $\HsVar{compare}$ can be more efficient
+% for complex types}}
+% \end{pboxed}
+% \)
+%
+% As the comment above says, it is sufficient to define either
+% $(\leq)$ or $\HsVar{compare}$ to get a complete instance. All
+% of the class methods have default definitions. First, we can
+% define $\HsVar{compare}$ in terms of $(\leq)$. The result type
+% of $\HsVar{compare}$ is an $\HsCon{Ordering}$, a type consisting
+% of only three values: $\HsCon{EQ}$ for ``equality'',
+% $\HsCon{LT}$ for ``less than'', and $\HsCon{GT}$ for ``greater
+% than''.
+%
+% \(
+% \begin{pboxed}
+% \restorecolumns
+% \column{cgrd}{l}
+% \fromto{body}{cgrd}{\HsVar{compare}\TX\TY}
+% \fromto{cgrd}{rel}{\vert\ \TX\equiv\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsCon{EQ}}
+% \\
+% \fromto{cgrd}{rel}{\vert\ \TX\leq\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsCon{LT}}
+% \\
+% \fromto{cgrd}{rel}{\vert\ \HsVar{otherwise}}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsCon{GT}}
+% \end{pboxed}
+% \)
+%
+% All the other comparison operators can be defined
+% in terms of $\HsCon{compare}$:
+%
+% \(
+% \begin{pboxed}
+% \restorecolumns
+% \fromto{body}{rel}{\TX\leq\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsVar{compare}\TX\TY\not\equiv\HsCon{GT}}
+% \\
+% \fromto{body}{rel}{\TX < \TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsVar{compare}\TX\TY\equiv\HsCon{LT}}
+% \\
+% \fromto{body}{rel}{\TX\geq\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsVar{compare}\TX\TY\not\equiv\HsCon{LT}}
+% \\
+% \fromto{body}{rel}{\TX > \TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\HsVar{compare}\TX\TY\equiv\HsCon{GT}}
+% \end{pboxed}
+% \)
+%
+% Finally, there are default definitions for
+% $\HsVar{max}$ and $\HsVar{min}$ in terms of $(\leq)$.
+%
+% \(
+% \begin{pboxed}
+% \restorecolumns
+% \fromto{body}{cgrd}{\HsVar{max}\TX\TY}
+% \fromto{cgrd}{rel}{\vert\ \TX\leq\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\TY}
+% \\
+% \fromto{cgrd}{rel}{\vert\ \HsVar{otherwise}}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\TX}
+% \\
+% \fromto{body}{cgrd}{\HsVar{min}\TX\TY}
+% \fromto{cgrd}{rel}{\vert\ \TX\leq\TY}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\TX}
+% \\
+% \fromto{cgrd}{rel}{\vert\ \HsVar{otherwise}}
+% \fromto{rel}{rhs}{=}
+% \fromto{rhs}{end}{\TY}
+% \end{pboxed}
+% \)
+% \hrule
+% \caption{Commented Haskell code example}
+% \label{tab:saverestore}
+% \end{figure}
+%
+% \section{The Code}
+% \begin{macrocode}
+%<*package>
+\NeedsTeXFormat{LaTeX2e}
+\ProvidesPackage{polytable}%
+ [2013/07/18 v0.8.5 `polytable' package (Andres Loeh)]
+% \end{macrocode}
+% New in v0.7.2:
+% The \textsf{amsmath} package clashes with \textsf{lazylist}:
+% both define the command |\And|. Although it would certainly
+% be better to find another name in \textsf{lazylist}, we
+% take precautions for now. (Note that this will still fail
+% if \textsf{lazylist} is already loaded -- but then it's not
+% our problem \dots
+% \begin{macrocode}
+\let\PT@original@And\And
+\let\PT@original@Not\Not
+\RequirePackage{lazylist}
+\let\PT@And\And
+\let\PT@Not\Not
+\def\PT@prelazylist
+ {\let\And\PT@And
+ \let\Not\PT@Not}
+\def\PT@postlazylist
+ {\let\And\PT@original@And
+ \let\Not\PT@original@Not}
+\PT@postlazylist
+\RequirePackage{array}
+% \end{macrocode}
+%
+% The option debug will cause (a considerable amount of)
+% debugging output to be printed. The option info is a
+% compromise, printing some status information for each table,
+% but no real debugging information. The option silent, on
+% the other hand, will prevent certain warnings from being
+% printed.
+% \begin{macrocode}
+\DeclareOption{debug} {\AtEndOfPackage\PT@debug}
+\DeclareOption{info} {\AtEndOfPackage\PT@info}
+\DeclareOption{silent}{\AtEndOfPackage\PT@silent}
+% \end{macrocode}
+%
+% First, we declare a couple of registers that we will need later.
+% \begin{macrocode}
+\newdimen\PT@colwidth
+\newcount\PT@cols
+\newcount\PT@table
+\newtoks\PT@toks
+\newif\ifPT@changed
+\newread\PT@in
+\newwrite\PT@out
+% \end{macrocode}
+% In |\PT@allcols|, we will store the list of all columns, as a
+% list as provided by the \textsf{lazylist} package. We initialise
+% it to the empty list, which is represented by |\Nil|. In v0.9,
+% we will have a second list that only contains the public columns.
+% \begin{macrocode}
+\def\PT@allcols{\Nil}
+%\def\PT@allpubliccols{\Nil}
+\let\PT@infromto\empty
+% \end{macrocode}
+% These are flags and truth values. TODO: Reduce and simplify.
+% \begin{macrocode}
+\let\PT@currentwidths\empty
+\def\PT@false{0}
+\def\PT@true{1}
+\let\PT@inrestore\PT@false
+% \end{macrocode}
+%
+% \begin{macro}{\defaultcolumn}
+% \begin{macro}{\nodefaultcolumn}
+% \begin{macro}{\PT@defaultcolumnspec}
+% The macro |\PT@defaultcolumnspec| contains, if defined, the
+% current default specifier that is assumed for undefined columns.
+% The other two commands can be used to set the specifier.
+% \begin{macrocode}
+\newcommand{\defaultcolumn}[1]{\gdef\PT@defaultcolumnspec{#1}}
+\newcommand{\nodefaultcolumn}{\global\let\PT@defaultcolumnspec\undefined}
+\DeclareOption{defaultcolumns}{\defaultcolumn{l}}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\memorytables}
+% \begin{macro}{\disktables}
+% \begin{macro}{\PT@add}
+% \begin{macro}{\PT@split}
+% \begin{macro}{\PT@prearewrite}
+% \begin{macro}{\PT@preareread}
+% \begin{macro}{\PT@finalize}
+% These macros steer where the end-column queue is stored, which
+% can be either in memory or on disk. The default is on disk,
+% because that's more reliable for large tables. There is a package
+% option |memory| to make |\memorytables| the default.
+% \begin{macrocode}
+\newcommand{\memorytables}{%
+ \let\PT@preparewrite\@gobble
+ \let\PT@add \PT@addmem
+ \let\PT@prepareread \PT@preparereadmem
+ \let\PT@split \PT@splitmem
+ \let\PT@finalize \relax
+}
+\newcommand{\disktables}{%
+ \let\PT@preparewrite\PT@preparewritefile
+ \let\PT@add \PT@addfile
+ \let\PT@prepareread \PT@preparereadfile
+ \let\PT@split \PT@splitfile
+ \let\PT@finalize \PT@finalizefile
+}
+\DeclareOption{memory}{\AtEndOfPackage\memorytables}
+\ProcessOptions
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\PT@debug}
+% \begin{macro}{\PT@info}
+% \begin{macro}{\PT@debug@}
+% \begin{macro}{\PT@typeout@}
+% \begin{macro}{\PT@silent}
+% \begin{macro}{\PT@warning}
+% Similar to the \textsf{tabularx} package, we add macros to
+% print debugging information to the log. Depending on package
+% options, we can set or unset them.
+% \begin{macrocode}
+\newcommand*{\PT@debug}
+ {\def\PT@debug@ ##1{\typeout{(polytable) ##1}}
+ \PT@info}
+\newcommand*{\PT@info}
+ {\def\PT@typeout@ ##1{\typeout{(polytable) ##1}}}
+\let\PT@debug@\@gobble
+\let\PT@typeout@\@gobble
+\def\PT@warning{\PackageWarning{polytable}}%
+\def\PT@silent
+ {\let\PT@typeout@\@gobble\let\PT@warning\@gobble}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\PT@aligndim}
+% \begin{macro}{\PT@aligncol}
+% The first is (almost) stolen from the \textsf{tabularx}-package, to nicely
+% align dimensions in the log file. TODO: fix some issues.
+% The other command is for column names.
+% \begin{macrocode}
+\def\PT@aligndim#1#2#3\@@{%
+ \ifnum#1=0
+ \if #2p%
+ \PT@aligndim@0.0pt\space\space\space\space\space\@@
+ \else
+ \PT@aligndim@#1#2#3\space\space\space\space\space\space\space\space\@@
+ \fi
+ \else
+ \PT@aligndim@#1#2#3\space\space\space\space\space\space\space\space\@@
+ \fi}
+
+\def\PT@aligndim@#1.#2#3#4#5#6#7#8#9\@@{%
+ \ifnum#1<10 \space\fi
+ \ifnum#1<100 \space\fi
+ \ifnum#1<\@m\space\fi
+ \ifnum#1<\@M\space\fi
+ #1.#2#3#4#5#6#7#8\space\space}
+
+\def\PT@aligncol#1{%
+ \PT@aligncol@#1\space\space\space\space\space\space\space\space\@@}
+
+\def\PT@aligncol@#1#2#3#4#5#6#7#8#9\@@{%
+ #1#2#3#4#5#6#7#8\space\space}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\PT@rerun}
+% This macro can be called at a position where we know that
+% we have to rerun \LaTeX\ to get the column widths right. It
+% issues a warning at the end of the document.
+% \begin{macrocode}
+\def\PT@rerun
+ {\PT@typeout@{We have to rerun LaTeX ...}%
+ \AtEndDocument
+ {\PackageWarning{polytable}%
+ {Column widths have changed. Rerun LaTeX.\@gobbletwo}}%
+ \global\let\PT@rerun\relax}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@currentcol}
+% \begin{macro}{\PT@currentcolumn}
+% Both macros are used during typesetting to store the current
+% column. The differences are subtle. TODO: remove
+% one of the two, if possible. The |\PT@currentcol| variant
+% contains the internal name, whereas the |\PT@currentcolumn|
+% variant contains the external name.
+% \end{macro}
+% \end{macro}
+%
+% \subsection{Macro definition tools}
+%
+% \begin{macro}{\PT@listopmacro}
+% \begin{macro}{\PT@consmacro}
+% \begin{macro}{\PT@appendmacro}
+% This assumes that |#2| is a list macro and |#3| is a new list element.
+% The macro |#2| should, after the call, expand to the list with the
+% new element |#1|ed. Because we don't know the number of tokens in |#3|,
+% we use a temporary macro |\PT@temp| (which is used frequently throughout
+% the package).
+% \begin{macrocode}
+\def\PT@listopmacro #1#2#3% #1 #3 to the list #2
+ {\def\PT@temp{#1{#3}}%
+ \expandafter\expandafter\expandafter
+ \def\expandafter\expandafter\expandafter
+ #2\expandafter\expandafter\expandafter
+ {\expandafter\PT@temp\expandafter{#2}}}
+
+\def\PT@consmacro{\PT@listopmacro\Cons}
+\def\PT@appendmacro{\PT@listopmacro\Cat}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% The follwing macro can be used to add something to
+% the end of a control structure.
+% \begin{macrocode}
+\def\PT@gaddendmacro #1#2% add #2 to the end of #1
+ {\PT@expanded{\gdef #1}{#1#2}}
+% \end{macrocode}
+% \begin{macro}{\PT@expanded}
+% This macro expands its second argument once before passing
+% it to the first argument. It is like |\expandafter|, but
+% works on grouped arguments instead of tokens.
+% \begin{macrocode}
+\def\PT@expanded #1#2%
+ {\expandafter\Twiddle\expandafter\Identity\expandafter{#2}{#1}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@enamedef}
+% This is much like |\@namedef|, but it expands |#2| once.
+% \begin{macrocode}
+\def\PT@enamedef #1% sets name #1 to the expansion of #2
+ {\PT@expanded{\@namedef{#1}}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@addoptargtomacro}
+% \begin{macro}{\PT@addargtomacro}
+% \begin{macrocode}
+\def\PT@addoptargtomacro
+ {\PT@add@argtomacro\PT@makeoptarg}
+\def\PT@addargtomacro
+ {\PT@add@argtomacro\PT@makearg}
+
+\def\PT@add@argtomacro#1#2#3%
+ {\PT@expanded{\PT@expanded{\gdef\PT@temp}}{\csname #3\endcsname}%
+ #1%
+ \PT@expanded{\PT@gaddendmacro{#2}}{\PT@temp}}
+
+\def\PT@makeoptarg%
+ {\PT@expanded{\def\PT@temp}{\expandafter[\PT@temp]}}
+\def\PT@makearg%
+ {\PT@expanded{\def\PT@temp}{\expandafter{\PT@temp}}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@gobbleoptional}
+% Gobbles one optional argument. Ignores spaces.
+% \begin{macrocode}
+\newcommand*{\PT@gobbleoptional}[1][]{\ignorespaces}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\PT@addmem}
+% \begin{macro}{\PT@splitmem}
+% \begin{macro}{\PT@addfile}
+% \begin{macro}{\PT@splitfile}
+% \begin{macro}{\PT@preparereadmem}
+% \begin{macro}{\PT@preparereadfile}
+% \begin{macro}{\PT@preparewrite}
+% \begin{macro}{\PT@finalizefile}
+% \begin{macro}{\PT@queuefilename}
+% The following macros handle a simple queue of names.
+% With |\PT@addmem|, a name is added to the end of the queue. Using
+% |\PT@splitmem|, the first element of the queue is bound to another
+% command. As a replacement, we also define |\PT@addfile| and
+% |\PT@splitfile|, that implement the queue in a file.
+% \begin{macrocode}
+\def\PT@addmem#1#2{\PT@gaddendmacro #2{\PT@elt{#1}}}
+\def\PT@splitmem#1#2{#1\PT@nil{#2}{#1}}
+
+\def\PT@elt#1#2\PT@nil#3#4{\gdef #3{#1}\gdef #4{#2}}
+
+\def\PT@queuefilename{\jobname.ptb}
+
+% the \empty at the end consumes the newline space
+\def\PT@addfile#1#2{%
+ \immediate\write #2{\string\def\string\PTtemp{#1}\string\empty}}
+
+\def\PT@splitfile#1#2{%
+ \ifeof #1%
+ \let #2=\empty
+ \else
+ \read #1 to#2%
+ %\show #2%
+ #2% hack, because it essentially ignores #2
+ \PT@expanded{\def #2}{\PTtemp}%
+ %\show #2%
+ \fi}
+
+%\def\strip#1{\def #1{\expandafter\@strip #1\@dummy}}
+%\def\@strip#1\@dummy{#1}
+
+\def\PT@preparereadmem#1#2{%
+ \global\let #1=#2}
+
+\def\PT@preparewritefile#1{%
+ \immediate\openout\PT@out\PT@queuefilename\relax
+ \let #1\PT@out}
+
+\def\PT@preparereadfile#1#2{%
+ \immediate\closeout\PT@out
+ \openin\PT@in\PT@queuefilename\relax
+ \let #1\PT@in}
+
+\def\PT@finalizefile{%
+ \closein\PT@in}
+
+\disktables
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \subsection{The environment}
+%
+% The general idea is to first scan the contents of the environment
+% and store them in a token register. In a few test runs, the positions
+% of the column borders are determined. After that, the table is
+% typeset by producing boxes of appropriate size.
+%
+% \begin{macro}{\beginpolytable}
+% This macro starts the environment. It should, however, not
+% be called directly, but rather in a \LaTeX\ environment.
+% We initialize the
+% token register to the empty string and then start scanning.
+% \begin{macrocode}
+\newcommand*{\beginpolytable}%
+% \end{macrocode}
+% We save the current enclosing \LaTeX\ environment in
+% |\PT@environment|. This will be the |\end| we will be
+% looking for, and this will be the environment we manually
+% close in the end.
+% \begin{macrocode}
+ {\edef\PT@environment{\@currenvir}%
+ \begingroup
+ % new in v0.7: save counters
+ \PT@savecounters
+ \PT@toks{}% initialise token register
+ \PT@scantoend}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\endpolytable}
+% This is just defined for convenience.
+% \begin{macrocode}
+\let\endpolytable=\relax
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@scantoend}
+% Whenenver an |\end| is encountered, we check if it really
+% ends the current environment. We store the tokens we have
+% read. Once we have found the end of the environment, we
+% initialize the column queue and column reference,
+% |\PT@columnqueue| and |\PT@columnreference|. The new interface
+% commands build the queue during the first test run, containing
+% the end columns of specific entries of the table. Later, the
+% queue is copied to be the reference, and in the typesetting run,
+% the information is used to compute the correct box widths.
+% We start with an empty queue, and set the reference to |\undefined|,
+% because it is not yet needed. TODO: queue and reference must be
+% global variables at the moment; try to change that.
+% \begin{macrocode}
+\newcommand{\PT@scantoend}% LaTeX check
+\long\def\PT@scantoend #1\end #2%
+ {\PT@toks\expandafter{\the\PT@toks #1}%
+ \def\PT@temp{#2}%
+ \ifx\PT@temp\PT@environment
+ \global\let\PT@columnqueue \empty
+ \global\let\PT@columnreference \undefined
+ \PT@preparewrite\PT@columnqueue
+ \expandafter\PT@getwidths
+ \else
+ \PT@toks\expandafter{\the\PT@toks\end{#2}}%
+ \expandafter\PT@scantoend
+ \fi}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@getwidths}
+% Here, we make as many test runs as are necessary to
+% determine the correct column widths.
+% \begin{macrocode}
+\def\PT@getwidths
+% \end{macrocode}
+% We let the |\column| command initialize a column
+% in the first run.
+% \begin{macrocode}
+ {\let\column \PT@firstrun@column
+% \end{macrocode}
+% There is the possibility to save or restore columns.
+% This is new in v0.4.
+% \begin{macrocode}
+ \let\savecolumns \PT@savewidths
+ \let\restorecolumns \PT@restorewidths
+% \end{macrocode}
+% We \emph{always} define a pseudo-column |@begin@|.
+% This denotes the begin of a row. We also define a
+% pseudo-column |@end@| denoting the end of a row (as
+% of v0.8; and I'm not sure if |@begin@| is still
+% needed).
+% \begin{macrocode}
+ \column{@begin@}{@{}l@{}}%
+ \column{@end@}{}%
+ \PT@cols=0\relax%
+% \end{macrocode}
+% The two other commands that are allowed inside of the
+% environment, namely |\fromto| and |\\| are
+% initialized. The |\fromto| command may increase the
+% current widths of some columns, if necessary, whereas
+% |\\| just resets the counter that keeps track
+% of the ``current'' column, to |0|. The command
+% |\nextline| is an old name for |\\|.
+% \begin{macrocode}
+ \let\fromto \PT@fromto
+ \let\PT@processentry \PT@checkwidth
+ \let\PT@scanbegin \PT@scanbeginfree
+ \let\\= \PT@resetcolumn
+ \let\nextline \PT@resetcolumn
+ \let\>= \PT@fromopt
+ \let\== \PT@from
+ \let\<= \PT@toopt
+ \global\PT@changedfalse % nothing has changed so far
+ \PT@resetcolumn % we are at the beginning of a line
+% \end{macrocode}
+% Now we are ready for a test run.
+% \begin{macrocode}
+ \the\PT@toks
+ \@ifundefined{PT@scanning}%
+ {}{\PT@resetcolumn\relax}%
+% \end{macrocode}
+% After the first run, we print extra information.
+% We use the contents of the macro |\column| to check
+% whether we are in the first run, because it will be
+% reset below for all other runs to do nothing.
+% \begin{macrocode}
+ \ifx\column\PT@otherrun@column
+ \else
+ % we are in first run, print extra info
+ \PT@prelazylist
+ \PT@typeout@{\PT@environment: \the\PT@cols\space columns, %
+ \PT@Print\PT@allcols}%
+ \PT@postlazylist
+ \fi
+% \end{macrocode}
+% The columns are initialised after the first run.
+% Therefore, we make sure that the |\column| command
+% won't do much in the other runs. Also, saving and
+% restoring columns is no longer needed.
+% \begin{macrocode}
+ \let\PT@firstrun@column \PT@otherrun@column
+ \let\savecolumns \PT@gobbleoptional
+ \let\restorecolumns \PT@gobbleoptional
+ \let\PT@savewidths \PT@gobbleoptional
+ \let\PT@restorewidths \PT@gobbleoptional
+% \end{macrocode}
+% New in v0.7.1: restore counters after each trial run.
+% \begin{macrocode}
+ \PT@restorecounters
+% \end{macrocode}
+% If some column widths have indeed changed in the
+% test run, this will be indicated by the flag
+% |\ifPT@changed|. Depending on this flag, we will
+% either loop and rerun, or we will continue in
+% |\PT@sortcols|.
+% \begin{macrocode}
+ \ifPT@changed
+ % we need to rerun if something has changed
+ \PT@typeout@{There were changes; another trial run needed.}%
+ \expandafter\PT@getwidths
+ \else
+ % we are done and can do the sorting
+ \PT@typeout@{There were no changes; reached fixpoint.}%
+ \expandafter\PT@sortcols
+ \fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@savecounters}
+% Save all \LaTeX\ counters so that they can be restored after a
+% trial run.
+% \begin{macrocode}
+\def\PT@savecounters
+ {\begingroup
+ \def\@elt ##1%
+ {\global\csname c@##1\endcsname\the\csname c@##1\endcsname}%
+ \xdef\PT@restorecounters{\cl@@ckpt}%
+ \endgroup}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@sortcols}
+% The column borders are sorted by their horizontal
+% position on the page (width). The they get numbered
+% consecutively. After that, we are well prepared to
+% typeset the table.
+% \begin{macrocode}
+\def\PT@sortcols
+% \end{macrocode}
+% First, we sort the list. To make sure that the
+% computation is only executed once, we save the
+% sorted list by means of an |\edef|. Sorting happens
+% with \textsf{lazylist}'s |\Insertsort| which expects
+% an order and a list. As order, we provide |\PT@ltwidth|,
+% which compares the widths of the columns.
+% To prevent expansion of the list structure, given by
+% |\Cons| and |\Nil|, we fold the list with the
+% |\noexpand|ed versions of the list constructors.
+% \begin{macrocode}
+ {\PT@prelazylist
+ \edef\PT@sortedlist
+ {\Foldr{\noexpand\Cons}{\noexpand\Nil}%
+ {\Insertsort\PT@ltmax\PT@allcols}}%
+ \PT@typeout@{Sorted columns:}%
+ \PT@PrintWidth\PT@sortedlist
+ \PT@postlazylist
+% \end{macrocode}
+% Now, each column is assigned a number, starting from
+% zero.
+% \begin{macrocode}
+ \PT@cols=0\relax%
+ \PT@prelazylist
+ \PT@Execute{\Map\PT@numbercol\PT@sortedlist}%
+ \PT@postlazylist
+ \edef\PT@lastcol@{\PT@StripColumn\PT@lastcol}%
+ \PT@typeout@{Numbered successfully, %
+ last column is \PT@lastcol@}%
+% \end{macrocode}
+% Now is a good time to save table information, if needed later.
+% We will also compare our computed information with the restored
+% maximum widths.
+% \begin{macrocode}
+ \ifx\PT@currentwidths\empty
+ \else
+ \PT@typeout@{Saving table information for \PT@currentwidths .}%
+ \PT@expanded\PT@saveinformation\PT@currentwidths
+ \fi
+% \end{macrocode}
+% Finally, we can typeset the table.
+% \begin{macrocode}
+ \PT@typeset}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@typeset}
+% We redefine |\fromto| and |\\| to their
+% final meaning in the typesetting process. The
+% |\fromto| statements will be replaced by appropriate
+% calls to |\multicolumn|, whereas the |\\| will
+% again reset the counter for the current column, but
+% also call the table environment's newline macro.
+% Again, |\nextline| is allowed as an old name for |\\|.
+% \begin{macrocode}
+\def\PT@typeset
+ {\PT@typeout@{Typesetting the table ...}%
+ \let\PT@processentry \PT@placeinbox
+ \let\PT@scanbegin \PT@scanbeginwidth
+ \let\\= \PT@resetandcr
+ \let\nextline \PT@resetandcr
+ \PT@prepareread\PT@columnreference\PT@columnqueue
+% \end{macrocode}
+% The environment that will be opened soon, can, if
+% if happens to be |tabular| or |array|, redefines |\\|
+% once more, and will redefine it to |\@arraycr|. To
+% prevent this, we also set |\@arraycr| to our command.
+% \begin{macrocode}
+ \let\@arraycr \PT@resetandcr
+% \end{macrocode}
+% The array environments keep each line in a group; therefore
+% |\PT@resetcolumn|, when executed during the linebreak, will
+% not affect the current column counters. By having
+% |\PT@resetcolumn| before entering the environment, we ensure
+% that the group reset will have the correct effect anyway.
+% \begin{macrocode}
+ \PT@resetcolumn % we are at the beginning of a line
+% \end{macrocode}
+% Now we start the tabular environment with the
+% computed preamble. We redefine the |\\| to whatever
+% the environment dictates.
+% \begin{macrocode}
+ \PT@begin%
+% \end{macrocode}
+% Run, and this time, typeset, the contents.
+% \begin{macrocode}
+ \the\PT@toks
+ \PT@fill% new in 0.7.3: balance the last line
+% \end{macrocode}
+% End the array, close the group, close the environment.
+% We are done!
+% \begin{macrocode}
+ \PT@finalize% finalize the queue (possibly close file)
+ \PT@end
+ \endgroup
+ \PT@typeout@{Finished.}%
+ \expandafter\end\expandafter{\PT@environment}}%
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{New interface}
+%
+% From v0.8 on, we offer a more convenient user interface
+% for the programming of the columns.
+%
+% \begin{macro}{\PT@from}
+% \begin{macro}{\PT@fromopt}
+% \begin{macro}{\PT@toopt}
+% The macro |\PT@from| is bound to |\=| in a polytable environment,
+% and used to move to the named column specified by its argument. The
+% previous column is closed. The variant |\PT@fromopt| is bound to
+% |\>| and takes an optional argument instead of a mandatory, which defaults
+% to the current column name followed by a single dot |.|. We use an
+% empty default which is then redefined to make it easier for the
+% user to use the default definition (TODO: explain better).
+% Otherwise, it is like |\PT@from|.
+%
+% The macro |\PT@toopt| is bound to |\<|. Where |\PT@from| ends an
+% entry if one is active, and starts a new one, the |\PT@toopt|
+% variant only ends the currently active entry.
+% \begin{macrocode}
+\newcommand{\PT@from}[1]%
+ {\PT@checkendentry{#1}\PT@dofrom{#1}}
+
+\newcommand{\PT@fromopt}[1][]%
+ {\def\PT@temp{#1}%
+ \ifx\PT@temp\empty
+ % set default column name
+ \def\PT@temp{\PT@currentcolumn .}%
+ \fi
+ \PT@expanded\PT@from\PT@temp}
+
+\newcommand{\PT@toopt}[1][]%
+ {\def\PT@temp{#1}%
+ \ifx\PT@temp\empty
+ % set default column name
+ \def\PT@temp{\PT@currentcolumn .}%
+ \fi
+ \PT@expanded\PT@checkendentry\PT@temp
+ \let\PT@scanning\undefined}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\PT@dofrom}
+% \begin{macrocode}
+\newcommand*{\PT@dofrom}[1]%
+ {\edef\PT@currentcolumn{#1}%
+ \let\PT@scanning\PT@currentcolumn
+ \let\PT@currentpreamble\relax% necessary for preparescan
+ \@ifnextchar[%]
+ {\PT@expanded\PT@dospecfrom\PT@currentcolumn}%
+ {\PT@expanded\PT@dodofrom \PT@currentcolumn}}
+
+\newcommand*{\PT@dospecfrom}{}% LaTeX check
+\def\PT@dospecfrom #1[#2]%
+ {\PT@checkglobalfrom #2\PT@nil{#1}%
+ \PT@dodofrom{#1}}
+
+\newcommand*{\PT@checkglobalfrom}{}% LaTeX check
+\def\PT@checkglobalfrom
+ {\@ifnextchar!\PT@getglobalfrom\PT@ignorefrom}
+
+\newcommand*{\PT@getglobalfrom}{}% LaTeX check
+\def\PT@getglobalfrom!#1\PT@nil#2%
+ {\column{#2}{#1}}
+
+\newcommand*{\PT@ignorefrom}{}% LaTeX check
+\def\PT@ignorefrom #1\PT@nil#2%
+ {\def\PT@currentpreamble{#1}}
+
+\newcommand*{\PT@dodofrom}[1]%
+ {\@ifundefined{PT@columnreference}%
+ {% trial run
+ \ifx\column\PT@otherruncolumn
+ \else
+ % first run
+ \let\PT@storeendcolumn\PT@add
+ \fi
+ \def\PT@temp{@end@}}%
+ {% final run
+ \PT@split\PT@columnreference\PT@temp
+ %\PT@typeout@{splitted: \PT@temp}
+ }%
+ \PT@expanded{\PT@expanded\PT@preparescan\PT@currentcolumn}\PT@temp
+ \PT@scanbegin}
+
+\let\PT@storeendcolumn\@gobbletwo
+% \end{macrocode}
+% Here, |\PT@scanbegin| will scan free or using the width, depending
+% on the run we are in.
+% \end{macro}
+%
+% \begin{macro}{\PT@fromto}
+% This is the implementation for the old explicit |\fromto|
+% macro. It takes the start and end columns, and the contents.
+% It can be used for all runs.
+% \begin{macrocode}
+\newcommand*{\PT@fromto}[3]%
+% \end{macrocode}
+% We allow a |\fromto| to follow a new-style command, but we reset
+% the current column to undefined, so no text may immediately
+% follow a |\fromto| command.
+% \begin{macrocode}
+ {\PT@checkendentry{#1}%
+ \let\PT@scanning\undefined
+% \end{macrocode}
+% We check a switch to prevent nested |\fromto|s.
+% \begin{macrocode}
+ \PT@infromto
+ \def\PT@infromto{%
+ \PackageError{polytable}{Nested fromto}{}}%
+% \end{macrocode}
+% Here, the real work is done:
+% \begin{macrocode}
+ \let\PT@currentpreamble\relax% necessary for preparescan
+ \PT@preparescan{#1}{#2}%
+ \PT@scanbegin #3\PT@scanend% defines \@curfield
+ \PT@processentry{#1}{#2}%
+% \end{macrocode}
+% The commands |\PT@scanbegin| and |\PT@processentry| will perform
+% different tasks depending on which run we are in.
+%
+% We ignore spaces after the |\fromto| command.
+% \begin{macrocode}
+ \let\PT@infromto\empty
+ \ignorespaces}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@checkendentry}
+% This macro checks if we are currently scanning an entry. If so
+% (this is detected by checking if |\PT@scanning| is defined),
+% we close the entry and handle it (in |\PT@endentry|)
+% before returning. The argument
+% is the name of the column from which this macro is called.
+% \begin{macrocode}
+\newcommand*{\PT@checkendentry}% takes one argument
+ {\@ifundefined{PT@scanning}%
+ {\let\PT@temp\@gobble}%
+ {\let\PT@temp\PT@endentry}%
+ \PT@temp}
+%\newcommand*{\PT@checkendentry}% takes one argument
+% {\@ifundefined{PT@post@preamble}%
+% {\let\PT@temp\PT@discardentry}%
+% {\let\PT@temp\PT@endentry}%
+% \PT@temp}
+
+%\newcommand*{\PT@discardentry}[1]%
+% {\let\PT@postpreamble=\empty\PT@scanend}
+
+\newcommand*{\PT@endentry}[1]%
+ {\PT@scanend
+ \edef\PT@temp{#1}%
+ \PT@expanded\PT@storeendcolumn\PT@temp\PT@columnqueue
+ \let\PT@storeendcolumn\@gobbletwo
+ \PT@expanded{\PT@expanded\PT@processentry\PT@currentcolumn}\PT@temp}
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{The trial runs}
+%
+% For each column, we store information in macros that are
+% based on the column name. We store a column's type (i.e., its
+% contribution to the table's preamble), its current width
+% (i.e., its the horizontal position where the column will start
+% on the page), and later, its number, which will be used for
+% the |\multicolumn| calculations.
+%
+% \begin{macro}{\PT@firstrun@column}
+% During the first trial run, we initialise all the columns. We
+% store their type, as declared in the |\column| command inside
+% the environment, and we set their initial width to 0pt.
+% Furthermore, we add the column to the list of all available
+% columns, increase the column counter, and tell \TeX\ to ignore
+% spaces that might follow the |\column| command.
+% New in v0.4.1: We make a case distinction on an empty type
+% field to prevent warnings for columns that have been defined
+% via |\PT@setmaxwidth| -- see there for additional comments.
+% New in v0.4.2: We allow redefinition of width if explicitly
+% specified, i.e., not equal to 0pt.
+% \begin{macrocode}
+\newcommand\PT@firstrun@column[3][0pt]%
+ {\@ifundefined{PT@col@#2.type}%
+ {\PT@typeout@{Defining column \PT@aligncol{#2} at #1.}%
+ \@namedef{PT@col@#2.type}{#3}%
+ \@namedef{PT@col@#2.width}{#1}% initialize the width of the column
+ % add the new column to the (sortable) list of all columns
+ \PT@consmacro\PT@allcols{PT@col@#2}%
+ \advance\PT@cols by 1\relax}%
+ {\expandafter\ifx\csname PT@col@#2.type\endcsname\empty
+ \relax % will be defined in a later table of the same set
+ \else
+ \begingroup
+ \def\PT@temp{PT@col@#2}%
+ \ifx\PT@temp\PT@endcol
+ \relax % end column is always redefined
+ \else
+ \PT@warning{Redefining column #2}%
+ \fi
+ \endgroup
+ \fi
+ \@namedef{PT@col@#2.type}{#3}%
+ \expandafter\ifdim#1>0pt\relax
+ \PT@typeout@{Redefining column #2 at #1.}%
+ \@namedef{PT@col@#2.width}{#1}%
+ \fi
+ }%
+% \end{macrocode}
+% For the case that we are saving and there is not yet information
+% from the |.aux| file, we define the |.max| and |.trusted| fields
+% if they are undefined. If information becomes available later, it
+% will overwrite these definitions.
+% \begin{macrocode}
+ \@ifundefined{PT@col@#2.max}%
+ {\@namedef{PT@col@#2.max}{#1}%
+ \expandafter\let\csname PT@col@#2.trusted\endcsname\PT@true}{}%
+ \ignorespaces}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@otherrun@column}
+% In all but the first trial run, we do not need any additional
+% information about the columns any more, so we just gobble the
+% two arguments, but still ignore spaces.
+% \begin{macrocode}
+\newcommand\PT@otherrun@column[3][]%
+ {\ignorespaces}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@checkcoldefined}
+% This macro verifies that a certain column is defined and
+% produces an error message if it is not. New in v0.8: this
+% macro implicitly defines the column if we have a default
+% column specifier.
+% \begin{macrocode}
+\def\PT@checkcoldefined #1%
+ {\@ifundefined{PT@col@#1.type}%
+ {\@ifundefined{PT@defaultcolumnspec}%
+ {\PackageError{polytable}{Undefined column #1}{}}
+ {\PT@debug@{Implicitly defining column #1}%
+ \PT@expanded{\column{#1}}{\PT@defaultcolumnspec}}}{}%
+% \end{macrocode}
+% We also have to define columns with empty specifiers. This
+% situation can occur if save/restoring columns that are defined
+% by default specifiers.
+% \begin{macrocode}
+ \expandafter\ifx\csname PT@col@#1.type\endcsname\empty\relax
+ \@ifundefined{PT@defaultcolumnspec}{}%
+ {\PT@debug@{Implicitly defining column #1}%
+ \PT@expanded{\column{#1}}{\PT@defaultcolumnspec}}%
+ \fi}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@checkwidth}
+% Most of the work during the trial runs is done here. We increase
+% the widths of certain columns, if necessary. Note that there
+% are two conditions that have to hold if |\fromto{A}{B}| is encountered:
+% \begin{itemize}
+% \item the width of |A| has to be at least the width of the current
+% (i.e., previous) column.
+% \item the width of |B| has to be at least the width of |A|, plus the
+% width of the entry.
+% \end{itemize}
+% \begin{macrocode}
+\def\PT@checkwidth #1#2%
+ {\PT@checkcoldefined{#2}% first column should have been checked before
+% \end{macrocode}
+% Here, we check the first condition.
+% \begin{macrocode}
+ \def\PT@temp{PT@col@#1}%
+ \ifx\PT@currentcol\PT@temp
+ \PT@debug@{No need to skip columns.}%
+ \else
+ \PT@colwidth=\expandafter\@nameuse\expandafter
+ {\PT@currentcol.width}\relax
+ \ifdim\PT@colwidth>\csname PT@col@#1.width\endcsname\relax
+ % we need to change the width
+ \PT@debug@{s \PT@aligncol{#1}: %
+ old=\expandafter\expandafter\expandafter
+ \PT@aligndim\csname PT@col@#1.width\endcsname\@@%
+ new=\expandafter\PT@aligndim\the\PT@colwidth\@@}%
+ \PT@changedtrue
+ \PT@enamedef{PT@col@#1.width}{\the\PT@colwidth}%
+ \fi
+% \end{macrocode}
+% The same for the untrusted |.max| values.
+% \begin{macrocode}
+ \PT@colwidth=\expandafter\@nameuse\expandafter
+ {\PT@currentcol.max}\relax
+ \ifdim\PT@colwidth>\csname PT@col@#1.max\endcsname\relax
+ % we need to change the width
+ \PT@debug@{S \PT@aligncol{#1}: %
+ old=\expandafter\expandafter\expandafter
+ \PT@aligndim\csname PT@col@#1.max\endcsname\@@%
+ new=\expandafter\PT@aligndim\the\PT@colwidth\@@}%
+ \PT@changedtrue
+ \PT@checkrerun
+ \PT@enamedef{PT@col@#1.max}{\the\PT@colwidth}%
+ \fi
+ \ifnum\csname PT@col@#1.trusted\endcsname=\PT@false\relax
+ \ifdim\PT@colwidth=\csname PT@col@#1.max\endcsname\relax
+ \PT@debug@{#1=\the\PT@colwidth\space is now trusted}%
+ \expandafter\let\csname PT@col@#1.trusted\endcsname\PT@true%
+ \fi
+ \fi
+ \fi
+% \end{macrocode}
+% We assume that the current field is typeset in |\@curfield|; we
+% can thus measure the width of the box and then test the second condition.
+% \begin{macrocode}
+ \PT@expanded{\def\PT@temp}{\the\wd\@curfield}%
+ \global\PT@colwidth=\@nameuse{PT@col@#1.width}%
+ \global\advance\PT@colwidth by \PT@temp\relax%
+ \ifdim\PT@colwidth>\csname PT@col@#2.width\endcsname\relax
+ % we need to change the width
+ \PT@debug@{#2 (width \PT@temp) starts after #1 (at \csname PT@col@#1.width\endcsname)}%
+ \PT@debug@{c \PT@aligncol{#2}: %
+ old=\expandafter\expandafter\expandafter
+ \PT@aligndim\csname PT@col@#2.width\endcsname\@@%
+ new=\expandafter\PT@aligndim\the\PT@colwidth\@@}%
+ \PT@changedtrue
+ \PT@enamedef{PT@col@#2.width}{\the\PT@colwidth}%
+ \fi
+% \end{macrocode}
+% And again, we have to do the same for the untrusted maximums.
+% \begin{macrocode}
+ \global\PT@colwidth=\@nameuse{PT@col@#1.max}%
+ \global\advance\PT@colwidth by \PT@temp\relax%
+ \ifdim\PT@colwidth>\csname PT@col@#2.max\endcsname\relax
+ % we need to change the width
+ \PT@debug@{C \PT@aligncol{#2}: %
+ old=\expandafter\expandafter\expandafter
+ \PT@aligndim\csname PT@col@#2.max\endcsname\@@%
+ new=\expandafter\PT@aligndim\the\PT@colwidth\@@}%
+ \PT@changedtrue
+ \PT@checkrerun
+ \PT@enamedef{PT@col@#2.max}{\the\PT@colwidth}%
+ \fi
+ \ifnum\csname PT@col@#2.trusted\endcsname=\PT@false\relax
+ \ifdim\PT@colwidth=\csname PT@col@#2.max\endcsname\relax
+ \PT@debug@{#2=\the\PT@colwidth\space is now trusted}%
+ \expandafter\let\csname PT@col@#2.trusted\endcsname\PT@true%
+ \fi
+ \fi
+% \end{macrocode}
+% Finally, we update the current column to |#2|.
+% \begin{macrocode}
+ \def\PT@currentcol{PT@col@#2}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@checkrerun}
+% If we have changed something with the trusted widths, we have
+% to check whether we are in a situation where we are using previously
+% defined columns. If so, we have to rerun \LaTeX.
+% \begin{macrocode}
+\def\PT@checkrerun
+ {\ifnum\PT@inrestore=\PT@true\relax
+ \PT@rerun
+ \fi}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@resetcolumn}
+% If the end of a line is encountered, we stop scanning the current
+% entry, and reset the current column.
+% \begin{macrocode}
+\newcommand*{\PT@resetcolumn}[1][]%
+ {\PT@checkendentry{@end@}%
+ \let\PT@currentcolumn\empty%
+ \let\PT@scanning\undefined
+ \let\PT@currentcol\PT@nullcol
+ % TODO: remove these lines if they don't work
+ %\let\PT@pre@preamble\empty
+ %\PT@scanbeginfree
+ }
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@nullcol}
+% The name of the |@begin@| column as a macro, to be able to compare
+% to it with |\ifx|; dito for the |@end@| column.
+% \begin{macrocode}
+\def\PT@nullcol{PT@col@@begin@}
+\def\PT@endcol{PT@col@@end@}
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{Sorting and numbering the columns}
+%
+% Not much needs to be done here, all the work is done by the
+% macros supplied by the \textsf{lazylist} package.
+% We just provide a few additional commands to facilitate
+% their use.
+%
+% \begin{macro}{\PT@Execute}
+% \begin{macro}{\PT@Sequence}
+% With |\PT@Execute|, a list of commands (with sideeffects) can
+% be executed in sequence. Usually, first a command will be
+% mapped over a list, and then the resulting list will be
+% executed.
+% \begin{macrocode}
+\def\PT@Execute{\Foldr\PT@Sequence\empty}
+\def\PT@Sequence #1#2{#1#2}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \begin{macro}{\PT@ShowColumn}
+% This is a debugging macro, that is used to output the list
+% of columns in a pretty way. The columns internally get
+% prefixes to their names, to prevent name conflicts with
+% normal commands. In the debug output, we gobble this prefix
+% again.
+% \begin{macrocode}
+\def\PT@ShowColumn #1#2%
+ {\PT@ShowColumn@{#1}#2\PT@ShowColumn@}
+\def\PT@ShowColumn@ #1PT@col@#2\PT@ShowColumn@
+ {#1{#2} }
+\def\PT@ShowColumnWidth #1%
+ {\PT@typeout@{%
+ \PT@ShowColumn\PT@aligncol{#1}:
+ \expandafter\expandafter\expandafter
+ \PT@aligndim\csname #1.max\endcsname\@@}}
+\def\PT@StripColumn #1%
+ {\expandafter\PT@StripColumn@#1\PT@StripColumn@}
+\def\PT@StripColumn@ PT@col@#1\PT@StripColumn@
+ {#1}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@Print}
+% Prints a list of columns, using |\PT@ShowColumn|.
+% \begin{macrocode}
+\def\PT@Print#1{\PT@Execute{\Map{\PT@ShowColumn\Identity}#1}}
+\def\PT@PrintWidth#1{\PT@Execute{\Map\PT@ShowColumnWidth#1}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@TeXif}
+% This is an improved version of \textsf{lazylist}'s |\TeXif|.
+% It does have an additional |\relax| to terminate the condition.
+% The |\relax| is gobbled again to keep it fully expandable.
+% \begin{macrocode}
+\def\PT@TeXif #1%
+ {\expandafter\@gobble#1\relax
+ \PT@gobblefalse
+ \else\relax
+ \gobbletrue
+ \fi}
+\def\PT@gobblefalse\else\relax\gobbletrue\fi #1#2%
+ {\fi #1}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@ltmax}
+% The order by which the columns are sorted is given by
+% the order on their (untrusted) widths.
+% \begin{macrocode}
+\def\PT@ltmax #1#2%
+ {\Not{\PT@TeXif{\ifdim\csname #1.max\endcsname>\csname #2.max\endcsname}}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@numbercol}
+% This assigns the next consecutive number to a column.
+% We also reassign |PT@lastcol| to remember the final column.
+% \begin{macrocode}
+\def\PT@numbercol #1%
+ {%\PT@typeout@{numbering #1 as \the\PT@cols}%
+ \PT@enamedef{#1.num}{\the\PT@cols}%
+ \def\PT@lastcol{#1}%
+ \advance\PT@cols by 1\relax}
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{Typesetting the table}
+%
+% Remember that there are three important macros that
+% occur in the body of the polytable: |\column|, |\fromto|,
+% and |\\|. The |\column| macro is only really used
+% in the very first trial run, so there is nothing new
+% we have to do here, but the other two have to be
+% redefined.
+%
+% \begin{macro}{\PT@resetandcr}
+% This is what |\\| does in the typesetting phase.
+% It resets the current column, but it also calls the surrounding
+% environment's newline macro |\PT@cr| \dots
+% If we are \emph{not} in the last column, we insert
+% an implicit |fromto|. This is needed for the boxed environment to
+% make each column equally wide. Otherwise, if the boxed environment
+% is typeset in a centered way, things will go wrong.
+% \begin{macrocode}
+\newcommand{\PT@resetandcr}%
+ {\PT@expanded\PT@checkendentry\PT@lastcol@%
+ \ifx\PT@currentcol\PT@lastcol
+ \else
+ \ifx\PT@currentcol\PT@nullcol
+ \edef\PT@currentcol{\Head{\Tail\PT@sortedlist}}%
+ \fi
+ \edef\PT@currentcol@{\PT@StripColumn\PT@currentcol}%
+ \PT@typeout@{adding implicit fromto at eol from \PT@currentcol@
+ \space to \PT@lastcol@}%
+ \PT@expanded{\PT@expanded\fromto\PT@currentcol@}\PT@lastcol@
+ \fi
+ \PT@typeout@{Next line ...}%
+ \let\PT@scanning\undefined% needed for resetcolumn
+ \PT@resetcolumn\PT@cr}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@fill}
+% This variant of |\PT@resetandcr| is used at the end of
+% the environment, to insert a blank box for the |pboxed|
+% environment to balance the widths of all lines. It does
+% not start a new line, and does nothing if the current
+% column is |@begin@|. TODO: extract commonalities with
+% |\PT@resetandcr| into a separate macro.
+% \begin{macrocode}
+\newcommand{\PT@fill}%
+ {\PT@expanded\PT@checkendentry\PT@lastcol@%
+ \ifx\PT@currentcol\PT@lastcol
+ \else
+ \ifx\PT@currentcol\PT@nullcol
+ \else
+ \edef\PT@currentcol@{\PT@StripColumn\PT@currentcol}%
+ \PT@typeout@{adding implicit fromto from \PT@currentcol@
+ \space to \PT@lastcol@}%
+ \PT@expanded{\PT@expanded\fromto\PT@currentcol@}\PT@lastcol@
+ \fi\fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@placeinbox}
+% This macro is the final-run replacement for |\PT@checkwidth|. We use
+% the precomputed width information to typeset the contents of the
+% table in aligned boxes. The arguments are the same as for
+% |\PT@checkwidth|, i.e., the start and the end columns, and the assumption
+% that the entry is contained in the box |\@curfield|.
+% \begin{macrocode}
+\def\PT@placeinbox#1#2%
+% \end{macrocode}
+% We start by computing the amount of whitespace that
+% must be inserted before the entry begins. We then insert that amount
+% of space.
+% \begin{macrocode}
+ {\PT@colwidth=\@nameuse{PT@col@#1.max}%
+ \advance\PT@colwidth by -\expandafter\csname\PT@currentcol.max\endcsname
+ \leavevmode
+ \edef\PT@temp{\PT@StripColumn\PT@currentcol}%
+ \PT@typeout@{adding space of width %
+ \expandafter\PT@aligndim\the\PT@colwidth\@@
+ (\expandafter\PT@aligncol\expandafter{\PT@temp} %
+ -> \PT@aligncol{#1})}%
+ \hb@xt@\PT@colwidth{%
+ {\@mkpream{@{}l@{}}\@addtopreamble\@empty}%
+ \let\CT@row@color\relax% colortbl compatibility
+ \let\@sharp\empty%
+ %\show\@preamble
+ \@preamble}%
+% \end{macrocode}
+% The important part is to use the pre-typeset box |\@curfield|.
+% This produces real output!
+% \begin{macrocode}
+ \PT@typeout@{adding box \space\space of width %
+ \expandafter\PT@aligndim\the\wd\@curfield\@@
+ (\PT@aligncol{#1} -> \PT@aligncol{#2})}%
+ \box\@curfield
+% \end{macrocode}
+% Finally, we have to reset the current column and
+% ignore spaces.
+% \begin{macrocode}
+ \def\PT@currentcol{PT@col@#2}%
+ \ignorespaces}%
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@preparescan}
+% The macro |\PT@preparescan| sets the two macros |\PT@scanbegin|
+% and |\PT@scanend| in such a way that they scan the input between
+% those two macros and place it in a box. The width of the box is
+% determined from the given column names. The name |@end@| can be
+% used as a column name is a free scan (a scan without knowing the
+% real end column) is desired. To allow redefinition of the preamble,
+% we assume that |\PT@currentpreamble| is defined to |\relax| if
+% we want it set normally dugin |\PT@preparescan|.
+% \begin{macrocode}
+\def\PT@preparescan#1#2%
+% First, we check that both columns are defined. This will
+% actually define the columns if implicit column definitions are
+% enabled.
+% \begin{macrocode}
+ {\PT@checkcoldefined{#1}%
+ \PT@checkcoldefined{#2}%
+ \PT@colwidth=\@nameuse{PT@col@#2.max}%
+ \advance\PT@colwidth by -\@nameuse{PT@col@#1.max}\relax%
+ \ifmmode
+ \PT@debug@{*math mode*}%
+ \let\d@llarbegin=$%$
+ \let\d@llarend=$%$
+ \let\col@sep=\arraycolsep
+ \else
+ \PT@debug@{*text mode*}%
+ \let\d@llarbegin=\begingroup
+ \let\d@llarend=\endgroup
+ \let\col@sep=\tabcolsep
+ \fi
+ \ifx\PT@currentpreamble\relax
+ \PT@expanded{\PT@expanded{\def\PT@currentpreamble}}%
+ {\csname PT@col@#1.type\endcsname}%
+ \fi
+% \end{macrocode}
+% Now, we make a preamble using the macro |\@mkpream| from the
+% \textsf{array} package. This macro takes a format string as
+% argument, and defines |\@preamble| as a result, where |\@sharp|
+% occurs in the positions of the column contents.
+% We perform the operation in a group to prevent certain redefinitions
+% from escaping. The |\@preamble| is set globally anyway.
+% \begin{macrocode}
+ {\PT@expanded\@mkpream\PT@currentpreamble%
+ \@addtopreamble\@empty}%
+ \let\CT@row@color\relax% colortbl compatibility
+% \end{macrocode}
+% We split the preamble at the position of the |\@sharp|, using
+% some tricks to make sure that there really is precisely one
+% occurrence of |\@sharp| in the resulting preamble code, and
+% producing an error otherwise. The splitting defines
+% |\PT@pre@preamble| and |\PT@post@preamble|. With those and
+% the computed |\PT@colwidth|, the scan is successfully prepared.
+% \begin{macrocode}
+ \expandafter\PT@splitpreamble\@preamble\@sharp\PT@nil}
+% \end{macrocode}
+%
+% We now define the splitting of the preamble.
+% \begin{macrocode}
+\def\PT@splitpreamble #1\@sharp #2\PT@nil{%
+ \let\@sharp=\relax% needed for the following assignment
+ \def\PT@terp{#2}%
+ \ifx\PT@terp\empty%
+ \PackageError{polytable}{Illegal preamble (no columns)}{}%
+ \fi
+ \PT@splitsplitpreamble{#1}#2\PT@nil}
+
+\def\PT@splitsplitpreamble #1#2\@sharp #3\PT@nil{%
+ \def\PT@temp{#3}%
+ \ifx\PT@temp\empty%
+ \else
+ \PackageError{polytable}{Illegal preamble (multiple columns)}{}%
+ \fi
+ \def\PT@pre@preamble{#1}%
+ \def\PT@post@preamble{#2}}%
+% \end{macrocode}
+%
+% Finally, we can define the scan environment, which depends on
+% all the other macros being defined correctly. The macro
+% |\PT@scanbegin| is not defined directly, but will be set
+% to |\PT@scanbeginfree| during the trial runs and to
+% |\PT@scanbeginwidth| during the final run.
+% \begin{macrocode}
+\def\PT@scanbeginwidth
+ {\PT@scanbegin@{\hbox to \PT@colwidth}}
+
+\def\PT@scanbeginfree
+ {\PT@scanbegin@{\hbox}}
+
+\def\PT@scanbegin@#1%
+ {\setbox\@curfield #1%
+ \bgroup
+ \PT@pre@preamble\strut\ignorespaces}
+
+\def\PT@scanend
+ {\PT@post@preamble
+ \egroup}
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{Saving and restoring column widths}
+%
+% Column width information can be saved under a name and thus
+% be reused in other tables. The idea is that the command
+% |\savecolumns| can be issued inside a polytable to save the
+% current column information, and |\restorecolumns| can be used
+% to make that information accessible in a later table. All tables
+% using the same information should have the same column widths,
+% which means that some information might need to be passed back.
+% Therefore, we need to write to an auxiliary file.
+%
+% Both |\savecolumns| and |\restorecolumns| are mapped to the internal
+% commands |\PT@savewidths| and |\PT@restorewidths|. Both take an
+% optional argument specifying a name for the column width
+% information. Thereby, multiple sets of such information can be
+% used simultaneously.
+%
+% One important thing to consider is that the widths read from the
+% auxiliary file must not be trusted. The user may have edited the
+% source file before the rerun, and therefore, the values read might
+% actually be too large (or too small, but this is less dangerous).
+%
+% The way we solve this problem is to distinguish two width values
+% per column: the trusted width, only using information from the
+% current run, and the untrusted width, incorportating information
+% from the |.aux| file. An untrusted width can become (conditionally)
+% trusted if it is reached in the computation with respect to an
+% earlier column. (Conditionally, because its trustworthiness still
+% depends on the earlier columns being trustworthy.) In the end, we
+% can check whether all untrusted widths are conditionally trusted.
+%
+% We write the final, the maximum widths, into the auxiliary file.
+% We perform the
+% write operation when we are sure that a specific set is
+% no longer used. This is the case when we save a new set under the
+% same name, or at the end of the document. The command
+% |\PT@verifywidths| takes care of this procedure.
+% This command will also check if a rerun is necessary, and issue
+% an appropriate warning if that should be the case.
+%
+% \begin{macro}{\PT@setmaxwidth}
+% First, we need a macro to help us interpreting the contents of
+% the |.aux| file. New v0.4.1: We need to define the restored
+% columns with the |\column| command, because otherwise we will
+% have problems in the case that later occurences of tables in
+% the document that belong to the same set, but define additional
+% columns. (Rerun warnings appear ad infinitum.)
+% In v0.4.2: columns with width 0.0 are now always trusted.
+% \begin{macrocode}
+\newcommand*{\PT@setmaxwidth}[3][\PT@false]% #2 column name, #3 maximum width
+ {\@namedef{PT@col@#2.max}{#3}%
+ \ifdim#3=0pt\relax
+ \expandafter\let\csname PT@col@#2.trusted\endcsname=\PT@true%
+ \else
+ \expandafter\let\csname PT@col@#2.trusted\endcsname=#1%
+ \fi
+ \column{#2}{}}%
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@loadtable}
+% Now, we can load table information that has been read from the
+% |.aux| file. Note that a |\csname| construct expands to |\relax|
+% if undefined.
+% \begin{macrocode}
+\def\PT@loadtable#1% #1 table id number
+ {%\expandafter\show\csname PT@restore@\romannumeral #1\endcsname
+ %\show\column
+ \PT@typeout@
+ {Calling \expandafter\string
+ \csname PT@restore@\romannumeral #1\endcsname.}%
+ \let\maxcolumn\PT@setmaxwidth
+ %\expandafter\show\csname PT@load@\romannumeral #1\endcsname
+ \csname PT@restore@\romannumeral #1\endcsname}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@loadtablebyname}
+% Often, we want to access table information by a column width set
+% name.
+% \begin{macrocode}
+\def\PT@loadtablebyname#1% #1 set name
+ {\PT@typeout@{Loading table information for column width set #1.}%
+ \PT@loadtable{\csname PT@widths@#1\endcsname}}%
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@saveinformation}
+% In each table for which the widths get reused (i.e., in all tables
+% that use either |\savecolumns| or |\restorecolumns|, we have to store
+% all important information for further use.
+% \begin{macrocode}
+\def\PT@saveinformation#1% #1 set name
+ {\PT@expanded{\def\PT@temp}{\csname PT@widths@#1\endcsname}%
+ \PT@expanded{\def\PT@temp}%
+ {\csname PT@restore@\romannumeral\PT@temp\endcsname}%
+ \expandafter\gdef\PT@temp{}% start empty
+ % this is: \PT@Execute{\Map{\PT@savecolumn{\PT@temp}}{\Reverse\PT@allcols}}
+ \expandafter\PT@Execute\expandafter{\expandafter
+ \Map\expandafter{\expandafter\PT@savecolumn
+ \expandafter{\PT@temp}}{\Reverse\PT@allcols}}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@savecolumn}
+% A single column is saved by this macro.
+% \begin{macrocode}
+\def\PT@savecolumn#1#2% #1 macro name, #2 column name
+ {\PT@typeout@{saving column #2 in \string #1 ...}%
+ \def\PT@temp{#2}%
+ \ifx\PT@temp\PT@nullcol
+ \PT@typeout@{skipping nullcol ...}%
+% This was a bug: end column cannot be skipped, because
+% it can change.
+% \else\ifx\PT@temp\PT@endcol
+% \PT@typeout@{skipping endcol ...}%
+ \else
+ \PT@typeout@{max=\csname #2.max\endcsname, %
+ width=\csname #2.width\endcsname, %
+ trusted=\csname #2.trusted\endcsname}%
+ % we need the column command in here
+ % we could do the same in \column, but then the location of
+ % \save / \restore matters ...
+ \PT@gaddendmacro{#1}{\maxcolumn}%
+ \ifnum\csname #2.trusted\endcsname=\PT@true\relax
+ \PT@gaddendmacro{#1}{[\PT@true]}%
+ \fi
+ \edef\PT@temp{\PT@StripColumn{#2}}%
+ \PT@addargtomacro{#1}{PT@temp}%
+ \PT@addargtomacro{#1}{#2.max}%
+ \PT@gaddendmacro{#1}{\column}%
+ \PT@addoptargtomacro{#1}{#2.width}%
+ \edef\PT@temp{\PT@StripColumn{#2}}%
+ \PT@addargtomacro{#1}{PT@temp}%
+ \PT@addargtomacro{#1}{#2.type}%
+ %\show#1%
+% \fi
+ \fi
+ }
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@savewidths}
+% If we really want to save column width information, then the first
+% thing we should worry about is that there might already have been
+% a set with the name in question. Therefore, we will call
+% |\PT@verifywidths| for that set. In the case that there is no set
+% of this name yet, we will schedule the set for verification at the
+% end of document.
+% \begin{macrocode}
+\newcommand*{\PT@savewidths}[1][default@]
+ {\PT@typeout@{Executing \string\savecolumns [#1].}%
+ \def\PT@currentwidths{#1}%
+ \PT@verifywidths{#1}%
+% \end{macrocode}
+% We now reserve a new unique number for this column width set
+% by increasing the |\PT@table| counter. We then associate
+% the given name (or |default@|) with the counter value and
+% restore the widths from the |.aux| file if they are present.
+% \begin{macrocode}
+ \global\advance\PT@table by 1\relax
+ \expandafter\xdef\csname PT@widths@#1\endcsname
+ {\the\PT@table}%
+ \PT@loadtable{\PT@table}%
+ \ignorespaces}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@restorewidths}
+% Restoring information is quite simple. We just load all information
+% available.
+% \begin{macrocode}
+\newcommand*{\PT@restorewidths}[1][default@]
+ {\PT@typeout@{Executing \string\restorecolumns [#1].}%
+ \def\PT@currentwidths{#1}%
+ \let\PT@inrestore\PT@true
+ \PT@loadtablebyname{#1}%
+ \ignorespaces}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@comparewidths}
+% \begin{macrocode}
+\def\PT@comparewidths#1% #1 full column name
+ {\@ifundefined{#1.max}%
+ {\PT@typeout@{computed width for #1 is fine ...}}%
+ {\ifdim\csname #1.max\endcsname>\csname #1.width\endcsname\relax
+ \PT@typeout@{Preferring saved width for \PT@StripColumn{#1}.}%
+ \PT@changedtrue
+ \PT@colwidth=\@nameuse{#1.max}\relax
+ \PT@enamedef{#1.width}{\the\PT@colwidth}%
+ \fi}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@trustedmax}
+% \begin{macrocode}
+\def\PT@trustedmax#1%
+ {\PT@TeXif{\ifnum\csname #1.trusted\endcsname=\PT@true}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\PT@equalwidths}
+% \begin{macrocode}
+\def\PT@equalwidths#1% #1 full column name
+ {\@ifundefined{#1.max}{}%
+ {\ifdim\csname #1.max\endcsname=\csname #1.width\endcsname\relax
+ \PT@typeout@{col #1 is okay ...}%
+ \else
+ \PT@rerun% a rerun is needed
+ \fi}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@verifywidths}
+% \begin{macrocode}
+\def\PT@verifywidths#1% #1 column width set name
+ {\@ifundefined{PT@widths@#1}%
+ {\PT@typeout@{Nothing to verify yet for set #1.}%
+ \PT@typeout@{Scheduling set #1 for verification at end of document.}%
+ \AtEndDocument{\PT@verifywidths{#1}}}%
+ {\PT@typeout@{Verifying column width set #1.}%
+ \PT@expanded\PT@verify@widths{\csname PT@widths@#1\endcsname}{#1}}}
+
+\def\PT@verify@widths#1#2% #1 set id number, #2 set name
+ {\@ifundefined{PT@restore@\romannumeral #1}{}%
+ {\begingroup
+ \let\column\PT@firstrun@column
+ \PT@cols=0\relax%
+ \def\PT@allcols{\Nil}%
+ \PT@loadtablebyname{#2}%
+ \PT@table=#1\relax
+ % nullcolumn is not loaded, therefore:
+ \@namedef{\PT@nullcol .width}{0pt}%
+ % checking trust
+ \PT@prelazylist
+ \All{\PT@trustedmax}{\PT@allcols}%
+ {\PT@typeout@{All maximum widths can be trusted -- writing .max!}%
+ \PT@save@table{.max}}%
+ {\PT@typeout@{Untrustworthy maximums widths -- writing .width!}%
+ \PT@rerun
+ \PT@save@table{.width}}%
+ \PT@postlazylist
+ \endgroup}%
+ \PT@typeout@{Verification for #2 successful.}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@save@table}
+% Here we prepare to write maximum column widths to the
+% |.aux| file.
+% \begin{macrocode}
+\def\PT@save@table#1%
+ {\PT@typeout@{Saving column width information.}%
+ \if@filesw
+ \PT@prelazylist
+ {\immediate\write\@auxout{%
+ \gdef\expandafter\noexpand
+ \csname PT@restore@\romannumeral\PT@table\endcsname
+ {\PT@Execute{\Map{\PT@write@column{#1}}{\Reverse\PT@allcols}}}}}%
+ \PT@postlazylist
+ \fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\PT@write@column}
+% We define the column command to write to the file.
+% \begin{macrocode}
+\def\PT@write@column #1#2%
+ {\noexpand\maxcolumn^^J%
+ {\PT@StripColumn{#2}}%
+ {\@nameuse{#2#1}}}%
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{The user environments}
+%
+% It remains to define the environments to be
+% called by the user.
+% New in v0.8: we add the environments
+% |ptboxed| and |pmboxed| for text-mode and math-mode
+% boxed environments. In turn, we remove |ptabular| and
+% |parray|, and make the point to their new counterparts.
+% \begin{macrocode}
+\def\pboxed{%
+ \let\PT@begin \empty
+ \let\PT@end \empty
+% \end{macrocode}
+% The following assignment is a hack. If |pboxed| is called
+% from within another |tabular|- or |array|-environment, then
+% this sometimes does the right thing.
+% \begin{macrocode}
+ \ifx\\\PT@arraycr
+ \let\PT@cr \PT@normalcr
+ \else
+ \let\PT@cr \\%
+ \fi
+ \expandafter\beginpolytable\ignorespaces}
+
+\let\endpboxed\endpolytable
+
+\def\ptboxed{%
+ \def\PT@begin {\tabular{@{}l@{}}}%
+ \let\PT@end \endtabular
+ \let\PT@cr \@arraycr
+ \expandafter\beginpolytable\ignorespaces}
+
+\let\endptboxed\endpolytable
+
+\def\pmboxed{%
+ \def\PT@begin {\array{@{}l@{}}}%
+ \let\PT@end \endarray
+ \let\PT@cr \@arraycr
+ \expandafter\beginpolytable\ignorespaces}
+
+\let\endpmboxed\endpolytable
+
+\let\ptabular \ptboxed
+\let\endptabular \endptboxed
+\let\parray \pmboxed
+\let\endparray \endpmboxed
+
+% \end{macrocode}
+%
+% That is all.
+% \begin{macrocode}
+%</package>
+% \end{macrocode}
+%
+% \Finale
+\endinput
diff --git a/macros/latex/contrib/polytable/polytable.ins b/macros/latex/contrib/polytable/polytable.ins
new file mode 100644
index 0000000000..70b0b83a88
--- /dev/null
+++ b/macros/latex/contrib/polytable/polytable.ins
@@ -0,0 +1,5 @@
+\def\batchfile{polytable.ins}
+\input docstrip
+\preamble
+\endpreamble
+\generate{\file{polytable.sty}{\from{polytable.dtx}{package}}}
diff --git a/macros/latex/contrib/polytable/polytable.pdf b/macros/latex/contrib/polytable/polytable.pdf
new file mode 100644
index 0000000000..2f267be0e5
--- /dev/null
+++ b/macros/latex/contrib/polytable/polytable.pdf
Binary files differ