\documentclass[11pt]{article}
\newcommand{\runningtitle}{SWI-Prolog's latex2html}
\usepackage{pl}
\usepackage{html}
\onefile
\makeindex

\htmloutput{.}
\htmlmainfile{ltx2htm}

\title{Converting LaTeX into HTML}
\author{Jan Wielemaker\\
	HCS, University of Amsterdam\\
	Kruislaan 419\\
	Amsterdam\\
	E-mail: \email{wielemak@science.uva.nl}}

\begin{document}

\maketitle

\begin{abstract}
This document describes a Prolog package for translating LaTeX into HTML
documents. It is not a reimplementation of \program{latex2html} in
Prolog. Instead of aiming at the production of a HTML document that
looks as closely as possible to the original printed version with as
little hassle as possible, it aims the simultaneous maintenance of a
paper and WEB version of a document.

The main virtue of this package is a simple, yet powerful, interface
for extending the translator.
\end{abstract}

\tableofcontents

\section{Introduction}

Paper and a hypertext browser are different media. For various
documents, it is desirable to have both a paper version and a version
that can be consulted through the WEB as a hypertext document. For the
latter, the document should be represented using HTML. HTML however is
not the ideal language for maintaining a document:

\begin{itemlist}
    \item [Not extendable]
It is not possible to define macros in HTML, which makes the maintenance
of documents in HTML itself difficult. Highly structured documents, such
as manuals and reports benefit from a description in the terminology of
the domain, which are later translated into domain-independent
formatting commands.  For example, when writing a Prolog manual, it
is better to start the description of a predicate using a formatting
command \const{predicate}, rather then directly using textual layout
primitives such as list, item, font, etc.

    \item [Difficult-to-read syntax]
This can always be debated, but lets say I don't like it.

    \item [Not sufficiently rich to specify a quality paper document]
Notably spacing, alignment, mathematics and tables are less powerful
than one would wish for generating a quality paper document.
\end{itemlist}



\section{Maintaining a structured document}

This toolkit aims first of all at structured documents: manuals,
technical report are examples. Such documents tend to describe many
instances of domain-specific data. This toolkit assumes the author
defines a set of LaTeX macros to map the domain-specific frames into the
general text terminology LaTeX uses. In parallel, the user defines
translation of these frames into HTML commands.  This translation is
not necessarily the same.  For example, when defining a syntax in BNF,
it may be desirable to make links from the usage of BNF terms to their
definition.  For the paper version, an index of the terms defined is
probably the desired cross-referencing aid.


\section{Overview}			\label{sec:overview}

The translator consists of two parts.  The first part is a C-program
that translates a LaTeX file or string into a Prolog list of TeX
tokens.  This package also contains the larger part of a pretty-printer
for HTML.  In between is a Prolog program that translates the LaTeX
token list into an HTML token list. For this process, Prolog's
unification and non-determinism allow for easy definition of the
translator.  The translator itself is separated into a number of
components, some of which may be extended by the user.

\begin{description}
    \item[Toplevel driver]
This component maintains files, options, etc.  It is responsible for
initialising and controlling the overall process. see latex2html/1.

    \item[Command and environment translator]
Maps individual commands and environments into either plain HTML tokens
or HTML macros (see below).  This level is represented by the predicates
cmd/[2-4] and env/2, which may be extended or redefined by the user.

    \item[HTML macro expander]
Generating plain HTML in one step is cumbersome.  The macro expander
allows for parameterised expansion of terms into sequences of HTML
commands.  Of course, macros may be expressed in macros.  The set
of macros is represented by the \predref{#}{2} predicate.  Macros
are called using \functor{#}{1} terms.

    \item[Reference resolver]
References to local (i.e.\ inside the document) labels are translated
into the appropriate HTML commands.  Warnings are printed for undefined
references.

    \item[Pretty printer]
The pretty-printer provides the final output.  It deletes, maps and
translates blank space in the output to arrive at a reasonable looking
HTML document.
\end{description}


\section{Installing the translator}

To install the translator, edit first install
\href{http://www.swi-prolog.org}{SWI-Prolog}. Next edit the
\file{Makefile} and finally run \program{make} to install the system.
After successful installation, the program may be run as:

\begin{verbatim}
$ latex2html file
\end{verbatim}


\section{Running the translator}

\subsection{Running LaTeX}

If the documents contains \program{BibTeX} references, it should first
be processed by \program{LaTeX} and \program{BibTeX} to create the
\fileext{bbl} file holding the references. Section numbers, footnotes,%
	\footnote{Here is an example using {\sc smallcaps}.}.
cross-references, etc.\ are generated by the translator. Running
\program{LaTeX} before using the translator is also useful for finding
and fixing \program{LaTeX} syntax errors. The translator checks the
syntax, but gives other difficult to understand error messages as
\program{LaTeX}.


\subsection{Translating to HTML}

Normally, the command

\begin{verbatim}
% latex2html file
\end{verbatim}

should translate \file{file.tex} into \file{file/file.html}.  Separate
\fileext{html} files are generated for each section, the table of
contents, all (foot)notes and the index, unless the command
\texcmd{onefile} appears in the preamble of the document, in which
case the entire document is created in \file{file/file.html}.

The translator handles \texcmd{include} and \texcmd{input} commands
to process LaTeX included files.  If a \texcmd{usepackage} command is
encountered, the system will look for a corresponding Prolog file to
define additional translation rules.  The content of packages or
\fileext{sty} files in not examined.  See \secref{extending}.


\subsection{Running the translator from the Prolog toplevel}

Notably when testing and extending the translator, it may be wise to
run the translator from the Prolog toplevel.  The easiest way is to
use:

\begin{verbatim}
% latex2html -pl
\end{verbatim}

which will start Prolog, set the appropriate library directory, load
\pllib{latex2html} and finally enters the Prolog toplevel. After this,
it is possible to load additional packages and/or initialise the
debugger for debugging the translator. To run the translator, use
latex2html/1:

\begin{description}
    \predicate{latex2html}{1}{+InputFile}
\arg{InputFile} is the name of the toplevel LaTeX file, either with
or without the \fileext{tex} extension.	  The translator will create
a directory using the basename of the \arg{InputFile} for the output.
The toplevel HTML output file is constructed from the same basename
with the extension \fileext{html}.

For example.  Suppose your toplevel document is called \file{gnat.tex}.
This is translated into \file{gnat/gnat.html} using the following
commands:

\begin{verbatim}
% latex gnat
% bibtex gnat
% latex2html -pl
Welcome to LaTeX2HTML version 0.5

1 ?- latex2thml(gnat).
2 ?- halt.
\end{verbatim}
\end{description}


\subsection{Warnings from the translator}

Warnings from the translator come in three flavours. First of all, there
are warnings from the tokeniser, either because it does not know about
the argument syntax for a command or environment, or because it detects
an error in the LaTeX document. For any command or environment it does
not know, it assumes the command or environment has no arguments, and
prints a warning:

\begin{code}
[WARNING: Unknown command: \gnat]
[WARNING: undefined environment: gnu]
\end{code}

See \secref{extending} for defining these commands.

Next, there are warnings issued by the translator.  If a command could
nor be translated, it will print a message of the form:

\begin{code}
Failed to translate \gnat in mode "document"
Term: "gnat"
\end{code}

If this happens, the definition of cmd/[2-4] should be debugged, or
extended to deal with the command.  The command will be ignored.
If an environment could not be processed the env/2 command should
be debugged or extended.  The warning is:

\begin{code}
translate_2(env(gnu, [], [\par]), document, ...) failed
\end{code}

If an HTML macro could not be expanded, the definition of
\predref{#}{2} should be examined. The system will warn:

\begin{code}
Post-processing macro #gnus/1 could not be expanded.
\end{code}

Finally, if local labels could not be resolved, messages of the format
below will be printed:

\begin{code}
No label for reference "sec:summary"
\end{code}


\section{Extending the translator}		\label{sec:extending}

This section defines how the translator may be programmed.  It defines the
hook predicates available to an extension module, as well as the generic
translation predicates that may be called from user-defined hooks.  First,
we explain how an extension is declared.

Normally, a document requiring additional structuring LaTeX macros will
start like this:

\begin{verbatim}
\documentclass[11pt]{article}
\usepackage{mymacros}

<preamble commands>

\begin{document}

<document contents>

\end{document}
\end{verbatim}

The file \file{mymacros.sty} will include additional packages required,
dedicated style settings, structuring macros, etc. The latex2thml
translator will not examine the contents of this file. Instead, it will
search for \file{mymacros.pl} to find a Prolog module defining the
additional translation rules. It will first use the \const{tex} path
alias, and then try the \const{library} path alias. If successful, this
file will be loaded into the \const{user} module using the Prolog
predicate ensure_loaded/1.  If the file cannot be located, it is
silently ignored.

The typical structure of the extension file is:

\begin{verbatim}
<file header>
:- module(<Module Name>, []).
:- use_module(library(latex2html)).

:- latex2html_module.
:- tex_load_commands(pl).

<clauses for the extension hooks>

\end{verbatim}

The directive latex2html_module/0 registers the module as a module
containing additional rules for latex2html. These rules will be tried
{\em before} any other rules. Alternatives need not be cut on any of the
hook predicates, as this is done immediately by the translators main
control loop. If you want to redefine a rule only under specific
circumstances, simply make sure the clause only succeeds in those cases
that redefinition is desired. It is generally not (yet) possible to
`wrap' the existing translation rule.

It is not required to define clauses for all extension hooks.

\subsection{Specifying the LaTeX syntax for the tokeniser} \label{sec:cmdspec}

The C-defined LaTeX tokeniser needs to know about the argument syntax
for each of the LaTeX commands used in the document.  The syntax of
commands is defined in \fileext{cmd} files.  The initial definition
is in the file \file{latex.cmd}.

Blank space is ignored, and the \% sign starts a comment field that
lasts till the end of the current line, just as with LaTeX itself.

A command declared using a line of the form:

\begin{tabular}{lrll}
<command>	\isa \const{\}<name> { <arg> } [\const{=}<cmdfunc>] & \\
<arg>		\isa \verb${$ <type> \verb$}$ & \\
		\ora \const{[}  <type> \const{]}  & \\
<type>		\isa \const{-} & Do not tokenise \\
		\ora \const{+} & Tokenise argument \\
		\ora \const{d} & Argument is a dimension \\
<cmdfunc>	\isa \const{verb} & Act as \texcmd{verb} \\
		\ora \const{begin} & Act as \texcmd{begin} \\
		\ora \const{end} & Act as \texcmd{end} \\
		\ora \const{item} & Separate items in a list \\
		\ora \const{prolog} & Call prolog_function/1 immediately \\
\end{tabular}

An environment declaration is of the form:

\begin{tabular}{lrll}
<environment>	\isa \verb${$ <name> \verb$}$ [\const{=}<envfunc>] & \\
<envfunc>	\isa \const{verbatim} & Act as a verbatim environment \\
		\ora \const{list} & My body consists of items \\
		\ora \const{float} & I am a floating environment
\end{tabular}

Command specification files are normally loaded from the Prolog extension
files using the predicate tex_load_commands/1.


\subsection{The HTML token list}		\label{sec:htmltokens}

The output of the translator is a list of HTML tokens. This section
describes the tokens that are recognised by the pretty printer.

\begin{description}
    \termitem{html}{Command}
Literally outputs the given command. No spaces are inserted in the
output, neither before nor after the command.
    \termitem{ref}{Label}
Expands to the number associated with \arg{Label}.  Thus, if \arg{Label}
appeared in a section, this is the section number.  If \arg{Label}
appeared in a figure or table, this is the number of the figure or
table.
    \termitem{label}{Label, Text, Ref}
Outputs \exam{<A NAME="\arg{Label}">\arg{Text}</A>}.
    \termitem{lref}{Label, Text}
Outputs \exam{<A HREF="<File>\#\arg{Label}">\arg{Text}</A>}.  <File>
is the file to which the section containing \arg{Label} has been (or
will be) written.  See label/3.
    \termitem{lref}{fileof(Label), Text}
Outputs \exam{<A HREF="<File>">\arg{Text}</A>}.  <File>
is the file to which the section containing \arg{Label} has been (or
will be) written.  See label/3.
    \termitem{iflref}{Arg}
As \term{lref}{Arg}, but emits nothing if the label is not defined. For
example, the page-header uses this construct to refer to to table of
contents, the summary and the index. If one of these does not exist, the
corresponding link is not created.
    \termitem{cite}{Key}
Translated from the LaTeX \texcmd{cite} command. It will lookup
\arg{Key} in the cite/2 predicate, and emit the label for the key,
rather then the key itself.  The output format depends on the BibTeX
style used.
    \termitem{yearcite}{Key}
As \term{cite}{Key}, but only emits the year of the label.
    \termitem{tableofcontents}{document}
Emits the entire table of contents of the document in this place.
    \termitem{tableofcontents}{section(\arg{Tag})}
Emits the table of contents of all subsections of the indicated section.
Normally used to create the footer for sections that have subsections.
    \termitem{tell}{FileBaseName}
Divert further output to the named file.  \arg{FileBaseName} is only
the basename of the output file, i.e. without trailing \fileext{html}
or leading directory.  This is the only way to switch output files.
The output is scanned for functor{tell}{1} terms and \functor{label}{3}
terms for building the label/3 predicate, which in turn is used for
outputting the proper reference names for labels.
    \termitem{verb}{Text}
Output \arg{Text} without any interpretation of spaces and newlines.
The HTML special characters (\verb$<>&$) are automatically mapped to
the required escape sequences.
    \termitem{verbatim}{Text}
Same as \functor{verb}{1}.
    \termitem{nospace}{Atom}
Output atom, translating the HTML special characters.  Do not insert
spaces before or after the word.
    \termitem{<Atom>}{}
Atoms or SWI-Prolog strings are treated as a {\em word}. HTML special
characters are translated. If two words are emitted adjecently, a space
or newline (for line-wrap) is inserted between them. No spaces are
inserted between any of the other forms and a word.
\end{description}


\subsection{The macro expander}			\label{sec:htmlmacros}

The macro expander is implemented by macro_expand/2 and is called by
the toplevel driver after any successfully translated command or
environment.

\begin{description}
    \predicate{macro_expand}{2}{+In, -Out}
The input is a (nested) list or a single HTML token or a macro
reference (terms of \functor{#}{1}, \functor{+}{1} or \functor{+}{2}).
The \arg{Out} is a (nested) list of HTML tokens as described in
\secref{htmltokens}.
\end{description}

The translator translates terms using the following syntax:

\begin{description}
    \termitem{+}{Tokens}
\arg{Tokens} is a list of raw LaTeX tokens, and the translator is
called recursively as \exam{translate(\arg{Tokens}, normal, \arg{Out})}.
    \termitem{+}{Mode, Tokens}
As \functor{+}{1}, translating in mode \arg{Mode} rather then
\const{normal}.
    \termitem{#}{MacroCall}
Calls \predref{#}{2} to expand the macro.
\end{description}

The following macros have been predefined:

\begin{description}
    \termitem{tell}{File}
If \texcmd{onefile} is specified in the header, do nothing, else emit
\term{tell}{File} to divert subsequent output to the named file.
    \termitem{thetitle}{}
Contains the document title defined by the LaTeX \texcmd{title}
command.
    \termitem{theauthor}{}
Contains the document author defined by the LaTeX \texcmd{author}
command.
    \termitem{title}{Title}
Defines the HTML document title using the \HTML{TITLE} tag.
    \termitem{defitem}{Label}
Definition-list item.  Translates to \exam{<P><DT>\arg{Label}<DD>}.
    \termitem{var}{Var}
Indicates a variable.  Translates to \exam{<VAR>\arg{Var}</VAR>}.
    \termitem{code}{Text}
Indicates code in running text. Translates to \exam{<CODE>\arg{Text}</CODE>}.
    \termitem{pre}{Text}
Preformatted multiline text. Translates to \exam{<PRED>\arg{Text}</PRE>}.
    \termitem{xmp}{Text}
Preformatted multiline example. Translates to \exam{<XMP>\arg{Text}</XMP>}.
    \termitem{strong}{Text}
Emit \arg{Text} as extra emphasized text (normally printed {\bf bold}).
Translates to \exam{<STRONG>\arg{Text}</STRONG>}.
    \termitem{em}{Text}
Emit \arg{Text} as emphasized text (normally printed {\em italic}).
Translates to \exam{<EM>\arg{Text}</EM>}.
    \termitem{b}{Text}
Emit \arg{Text} as bold text. Translates to \exam{<B>\arg{Text}</B>}.
    \termitem{i}{Text}
Emit \arg{Text} as italic text. Translates to \exam{<I>\arg{Text}</I>}.
    \termitem{center}{Text}
Emit \arg{Text} centered. Translates to \exam{<CENTER>\arg{Text}</CENTER>}.
    \termitem{quote}{Text}
Emit \arg{Text} as a quote. Translates to
\exam{<BLOCKQUOTE>\arg{Text}</BLOCKQUOTE>}.
    \termitem{embrace}{Text}
Emit \arg{Text} between "()" braces. Translates to \exam{(\arg{Text})}.
    \termitem{embrace}{"AZ", Text}
Emit \arg{Text} between "AZ" braces. Translates to \exam{A\arg{Text}Z}.
    \termitem{h}{Level, Text}
Emit a header.  Translates to \exam{<H\arg{N}>\arg{Text}</H\arg{N}>},
where \arg{N} is \arg{Level}.
    \termitem{row}{Columns}
Emits a row of a table.  Translates to \exam{<TR>\arg{C1}<TD>...</TR>}.
\arg{Columns} is a list containing the HTML code for each column in the
row.
    \termitem{label}{Label, Text, Tag}
Emits \term{label}{Label, Text, Tag}, but canonicalises the \arg{Label} and
\arg{Text} fields first.  \arg{Tag} is the tag of the current section or
float.
    \termitem{label}{Label, Text}
As \term{label}{Label, Text, Tag}, but emits a label not related to a
section or float.
    \termitem{lref}{Label, Text}
Emits \term{lref}{Label, Text}, but canonicalises the \arg{Label} and
\arg{Text} fields first.
    \termitem{iflref}{Label, Text}
Emits \term{iflref}{Label, Text}, but canonicalises the \arg{Label} and
\arg{Text} fields first.
    \termitem{lback}{Label, Text}
Emits \term{lback}{Label, Text}, but canonicalises the \arg{Label} and
\arg{Text} fields first.
    \termitem{lforw}{Label, Text}
Emits \term{lforw}{Label, Text}, but canonicalises the \arg{Label} and
\arg{Text} fields first.
    \termitem{url}{URL, Text}
Emits an external URL: \exam{<A HREF="\arg{URL}">\arg{Text}</A>}.
    \termitem{cite}{Keys}
Handle a \texcmd{cite} command.  \arg{Keys} is translated into a
sequence \mbox{\term{cite}{Key1}, ',', ' ', \ldots} and the whole
is enclosed in \exam{<CITE> \ldots </CITE>}.
    \termitem{opencite}{Keys}
Currently as \term{cite}{Keys}
    \termitem{yearcite}{Keys}
As \term{cite}{Keys}, but emits \term{yearcite}{Key}.
    \termitem{header}{Tag}
Produce a header for the file containing section \arg{Tag}.
This macro may be redefined to modify the section header.
    \termitem{header}
Produce a header for a file not defining a section.
    \termitem{footer}{Arg}
Produce a footer for the file containing section \arg{Tag}.
This macro may be redefined to modify the section header.  Normally,
this macro creates a `local' table of contents for all subsections of
this section (or nothing if there are no subsections).
\end{description}


\subsection{User extension hook predicates}		\label{sec:extpreds}

As discussed in \secref{overview}, the translator may be extended at
various levels.  Normally, is should suffice to extend the definitions
of cmd/[2-4], env/2, list_command/4 and \predref{#}{2}. They are
defines below.

\begin{description}
    \predicate{cmd}{4}{+Command, +ModeIn, -ModeOut, -HTML}
Translate the given \arg{Command}.  A command is represented by
a term of the format \term{<CommandName>}{<CommandArg>, \ldots}.
Each <CommandArg> is a term of the form \exam{\{<Arg>\}} if it is
an obligatory argument.  If it is an option argument and not
specified, <CommandArg> is the empty list (\const{[]}).  Otherwise
it is a list holding a single element representing the argument.

If the command defines an optional modifier (as the
\texcmd{section} command to suppress the section number), this
is translated into the first argument to the \arg{Command}, which is
either bound to the atom \const{*} if the modifier was specified, or
\const{-} of not.

If the command argument is specified as \const{-} or \const{d} (see
\secref{cmdspec}), the tokeniser returns the verbatim text of the
argument as a SWI-Prolog string. Otherwise, if the specifier is
\const{+}, the argument is a list of LaTeX tokens.  For example,
the \texcmd{url} command is defined as

\begin{code}
\url[+]{-}
\end{code}

and the command
\verb$\url[SWI-Prolog]{ftp://swi.psy.uva.nl/pub/SWI-Prolog}$
will be handed to cmd/4 as:

\begin{code}
url([[SWI-Prolog]], {'ftp://swi.psy.uva.nl/pub/SWI-Prolog'})
\end{code}

\arg{ModeIn} specifies the mode of the translator.  Only a few values
are defined and have defined behaviour:

\begin{description}
    \termitem{math}{}
We are processing a LaTeX math environment.

    \termitem{group}{Font}
Used to translate \exam{\{\ldots\}}, possibly holding font switches, as
well as other grouped environments, such as \texcmd{mbox} and
the cells of a table (\texenv{tabular}).

    \termitem{\arg{ListName}}{}
When processing a list (see list_command/4), the mode is the name of the
list environment.  This allows for the item commands to consider the
list environment in which they are active.
\end{description}

\arg{ModeOut} defines the mode for further processing.  Currently,
cmd/4 only implements the font-switches to manipulate the argument of
the \functor{group}{1} mode term.

    \predicate{cmd}{3}{+Command, +Mode, -HTML}
Version of cmd/4 that does not need to change modes, but realises mode
dependent commands. Example are the \texmode{math} commands. This
translator will only be called {\bf after} cmd/4 fails.

    \predicate{cmd}{2}{+Command, -HTML}
Version of cmd/4 that does not change, nor verify the current mode.
This translator will only be called {\bf after} cmd/3 fails.

    \predicate{env}{2}{+Environment, -HTML}
Expand an entire environment, i.e.\ from \texcmd{begin} upto and
including the matching \texcmd{end} command.  \arg{Environment}
is a term of the form \term{<EnvName>}{[<CommandArg>, \ldots],
<Tokens>}.  <CommandArg> describes the environment arguments (as
the column specifier for a \texenv{tabular} environment), while
<Tokens> is a list of LaTeX tokens, describing the contents of the
environment.  Note that \texenv{document} is an environment as
well, containing the entire document.  Redefining the document
environment provides the opportunity for pre- and post-processing
of the entire document.

Environments that are flagged with the \const{list} type (see
\secref{cmdspec}) are handled by a default translator. See
list_command/4.

    \predicate{list_command}{4}{+ListName, +Mode, -OpenHTML, -CloseHTML}
If env/2 traps an environment marked as \const{list}, it will break the
environment into items by scanning for commands flagged as \const{item},
output the \arg{OpenHTML} code, process the items and add the
\arg{CloseHTML} code. Below is the default definition of this predicate.
Note that, as always, the HTML output is either	a simple HTML token,
an HTML macro, or a (nested) list of these.

\begin{code}
list_command(description,     _, html('<DL>'), html('</DL>')).
list_command(itemize,         _, html('<UL>'), html('</UL>')).
list_command(enumerate,       _, html('<OL>'), html('</OL>')).
list_command(thebibliography, _, html('<DL>'), html('</DL>')).
\end{code}

    \predicate{#}{2}{+Macro, -Expansion}
This predicate defines the macro-expander as invoked through
the macro_expand/2 command.  \arg{Macro} is an arbitrary Prolog
term, possibly containing variables to act as macro arguments,
\arg{Expansion} is the expansion.  The expansion is an arbitrary
HTML token list, possibly containing references to other macros.
Recursive processing of the same macro is possible, but there
is no guarding against infinite loops.  See \secref{htmlmacros}
for the default macro-definitions available.

The usage of these macros is highly encouraged, both to keep the
translator simple and to provide sharing of definitions.
\end{description}


\subsection{Translation support predicates}

This section describes the public predicates of \pllib{latex2html}.
These predicates are available to the extension hooks to realise the
desired translation.

\begin{description}
    \predicate{welcome}{0}{}
Simply prints the package banner message.
    \predicate{translate}{4}{+In, +ModeIn, -ModeOut, -Out}
Call the general translator for a list of LaTeX tokens.  \arg{ModeIn}
is the starting mode, \arg{ModeOut} is the exit mode.  \arg{Out} is
unified with a (nested) list of HTML tokens.  Note that the hook
predicates can return \term{+}{TeXTokens}.  If this construct is
detected in the output of cmd/[2-4] or env/2, the translator is
called automatically.
    \predicate{translate}{3}{+In, +ModeIn, -Out}
As translate/4, but does not return the final mode.
    \predicate{latex2html_module}{0}{}
Register the calling module as an extension module module for the
translator.  All hooks will be called in the reverse order of these
registrations, ending with the built-in translation rules.
    \predicate{tex_input_directory}{1}{+Dir}
Register \arg{Dir} as a directory for finding LaTeX files.  The initial
directory is \file{.}.  If the variable \env{TEXINPUTS} is defined, the
initial directory list is read from this variable.  The definition of
the input files is maintained using in the user:file_search_path/2
alias \const{tex}.  Files are located using absolute_file_name/3.
    \predicate{step_counter}{2}{+Name, -Value}
Step the indicated counter.  Available counters are \const{figure} and
\const{table}.  The returned value is the new value of the counter.  The
first call returns 1.
    \predicate{ps2gif}{3}{+In, +Out, +Options}
Translate a \idx{PostScript} figure into a \idx{GIF} figure.  Incomplete.
Please see the source for the current definition.
    \predicate{translate_command}{4}{+In, +ModeIn, -ModeOut, -HTML}
Translate an individual command in the form as passed to cmd/[2-4].  May
be used to express the translation of a command in terms of the translation
of another.
    \predicate{translate_environment}{4}{+In, +ModeIn, -ModeOut, -HTML}
As translate_command/4, but for the translation of environments.
    \predicate{translate_reference}{4}{+TextLabel, +RefPrefix, +Label, -HTML}
Used for the translation of \texcmd{secref}, etc. The output will be
\arg{TextLabel}~\arg{Tag}, where \arg{Tag} is the tag-number of the
section or float of the target label. The label searched for is
\exam{\arg{RefPrefix}:\arg{Label}}.
    \predicate{translate_footnote}{2}{+Text, -HTML}
Create a footnote.  It will output a reference to the node and remember
the \arg{Text} to emit the final footnote.
    \predicate{current_setting}{1}{+Name(-Value, \ldots)}
Request a setting for the translator.  Currently defined:
    \begin{description}
        \termitem{keep_figures}{}
	For figure translation.  If this test succeeds, the old translation
        should be kept.
        \termitem{html_output_dir}{Dir}
        Unifies \arg{Dir} with the directory to which the output files are
        written.
    \end{description}
    \predicate{do_float}{2}{+FloatType(+Number), +Goal}
Run \arg{Goal}, such that labels will refer to the number of the
float environment (figure, table).  So, when processing the contents
of a figure environment, step the figure counter, and process the
body:

\begin{verbatim}
	...,
	step_counter(figure, Fig),
        do_float(figure(Fig), ( translate(...),
				...
			      )),
	...
\end{verbatim}
    \predicate{tex_load_commands}{1}{CmdBaseName}
Load a file holding LaTeX command declarations for the tokeniser.  See
\secref{cmdspec}.  The file normally has extension \fileext{cmd} and is
searched for in the same directories as extension Prolog files (\const{tex}
and \const{library}).
    \predicate{add_to_index}{2}{+Index, +Label}
Add \arg{Index} to the index.  \arg{Label} is the place from which the
item is referenced.  If \arg{Label} is a term \term{+}{Label}, it is
considered the {\em principal} reference, and the word in the index
itself will be rendered as a hyper-link to the principal location.
    \predicate{clean_tt}{2}{+Raw, -CleanAtom}
When writing manuals, it is often necessary to use TeX special commands
for specifying the TeX special characters.  For example, to get a nice,
not mathematical tilde, without using \texcmd{verb}, so it can be used
as a command argument, one should use \verb${\tt\string~}$.  This predicate
eliminates all these special constructs to arrive at an atom describing
the plain character sequence.
\end{description}


\subsection{Dynamic predicates for maintaining cross-referencing}

The predicates below are dynamic predicates maintained by the translator
to deal with cross references. They are currently internal to the
translator, but described anyhow as they are referenced in this
document.

\begin{description}
    \predicate{label}{3}{Label, File, Tag}
Built after the initial translation by scanning the output token list
for terms of the form \functor{tell}{1} (for keeping track of the
current file), \functor{label}{3} \functor{lback}{2} and
\functor{lforw}{2}.  Used by the final output stage.
    \predicate{cite}{2}{Key, Cite}
Built from processing \texcmd{bibitem} commands from the \fileext{bbl}
file.  Used by \texcmd{cite} to insert the proper reference to the
bibliography.
\end{description}


\section{Problems and limitations}

The current translator is not very complete.  It is mainly maintained
to translate the SWI-Prolog documentation into clean HTML.

\printindex

\end{document}