Configuration files of Tralics: syntax and usage

Table of Contents
1. Introduction
2. What Tralics prints on the terminal
     2.1. First example
     2.2. Second example
3. Explanations
     3.1. Launching Tralics
     3.2. Finding the Potential Type
     3.3. Finding the Type
     3.4 Using the Type
     3.5 Bootstrap code
4. The titlepage info
     4.1 Syntax
     4.2 Semantics
     4.3. Interpreting the CES item
     4.4. Interpreting the CES flags
     4.5. Comments
     4.6 Special action
5. Classes and packages


     5.1. User configuration file

1. Introduction

This document explains how to modify the output of Tralics:

In this document, we explain the syntax of the configuration file using this configuration file. We consider two TeX files this one, and that one, and show the XML result (this XML) and (that XML) and the log file produced by Tralics (first) and (second). We shall alse use the following tpa.tcf files.

2. What Tralics prints on the terminal

First example, using original syntax.

grimm@medee$ tralics tpa -configfile ../config_tpa -verbose
This is tralics 1.9t, a LaTeX to XML translator
Copyright INRIA/MIAOU/APICS 2002-2004, JosÚ Grimm
Starting xml processing for tpa.
<!-- S1 -->
Trying config file from user specs:../config_tpa
Read configuration file ../config_tpa.
Configuration file has type \documentclass
Seen \documentclass tpa
<!-- S2 -->
Potential type is tpa
Defined type: default
Defined type: tpa
Defined type: unused
Defined type: eurotex
Using type tpa
dtd is tpa from tpa.dtd (standard mode)
Ok with the config file, dealing with the TeX file...
<!-- S3 -->
Error signaled at line 35:
No value given for command \cmdp.
<!-- S4 -->
Seen 0 bibliographic entries
Used 1237 commands
math stats : formulas 0, non trivial kernels 0, cells 0,
   special 0 trivial 0, \mbox 0 large 0 small 0.
List stats: short 0 inc 2 alloc 529
Buffer realloc 0 string 1455 size 17863; merge 0
Macros created 28 deleted 4
Save stack +12 -12
Attribute list search 191(119) found 156 in 1081 elements (1037 after boot)
number of ref 0, of used labels 0, of defined labels 0, of ext. ref. 0
Modules with 0, without 0, sections with 0, without 0
There was no image.
Output written on tpa.xml (1436 bytes).
There was one error.
(For more information, see transcript file tpa.log)

First example, using new syntax.

grimm@medee$ tralics tpa -verbose 
This is tralics 2.8, a LaTeX to XML translator
Copyright INRIA/MIAOU/APICS 2002-2006, Jos\'e Grimm
Licensed under the CeCILL Free Software Licensing Agreement
Starting xml processing for tpa.
<!-- S1 -->
Configuration file identification: standard $ Revision: 2.24 $
Read configuration file /user/grimm/home/cvs/tralics/.tralics_rc.
Configuration file has type \documentclass
Seen \documentclass tpa
<!-- S2 -->
Potential type is tpa
Defined type: std
Configuration file identification: tpa.tcf $ Revision: 1.3 $
Read tcf file for type: ../confdir/tpa.tcf
Using type tpa
dtd is tpa from tpa.dtd (standard mode)
Ok with the config file, dealing with the TeX file...
<!-- S3 -->
Error signaled at line 35 of file tpa.tex:
No value given for command \cmdp.
<!-- S4 -->
Bib stats: seen 0 entries
Seen 0 bibliographic entries
Used 1810 commands
Math stats: formulas 0, non trivial kernels 0,
   special 0, trivial 0, \mbox 0, large 0, small 0.
Buffer realloc 0, string 1642, size 17963, merge 0
Macros created 86, deleted 4.
Save stack +24 -24.
Attribute list search 1700(1549) found 1019 in 1495 elements (1450 after boot).
Number of ref 0, of used labels 0, of defined labels 0, of ext. ref. 0.
Modules with 0, without 0, sections with 0, without 0
There was no image.
Output written on tpa.xml (1378 bytes).
There was one error.
(For more information, see transcript file tpa.log)

Second example, non-verbose mode

grimm@medee$ tralics bo
This is tralics 2.8, a LaTeX to XML translator
Copyright INRIA/MIAOU/APICS 2002-2006, Jos\'e Grimm
Licensed under the CeCILL Free Software Licensing Agreement
Starting xml processing for bo.
Configuration file identification: standard $ Revision: 2.24 $
Read configuration file /user/grimm/home/cvs/tralics/.tralics_rc.
Configuration file identification: cedram.tcf $ Revision: 1.1 $
Read tcf file for type: cedram.tcf
Translation terminated after title page
Bib stats: seen 0 entries
Seen 3 bibliographic entries
Used 1866 commands
Math stats: formulas 16, non trivial kernels 10,
   special 0, trivial 0, \mbox 0, large 0, small 8.
Buffer realloc 7, string 1881, size 25733, merge 2
Macros created 114, deleted 10.
Save stack +154 -154.
Attribute list search 1847(1549) found 1010 in 1620 elements (1450 after boot).
Number of ref 0, of used labels 0, of defined labels 0, of ext. ref. 0.
Modules with 0, without 0, sections with 0, without 0
There was no image.
Output written on bo.xml (5209 bytes).
No error found.
(For more information, see transcript file bo.log)

3. Explanations

The lines above show the command used to call Tralics, and its response. We have added some markers in order to make the explanations easy to follow.

The first example was run in 2004, using an old version of the software. We have rerun the example, with the new version of Tralics. As you can see, the resulting XML is differerent (1378 bytes instead of 1536). You can compare the old version with the new version. The differences are the following: first, spaces before a \par command are removed, this means that the XML contains no space before a closing </p> tag; second, <foo></foo> is now printed as <foo/>; third, the language atttribute is now added to the document element only if the configuration file contains att_language="language" (the right hand side of the assignment defines the attribute name).

3.1. Launching Tralics

For a list of all arguments of the Tralics command, see the page Options of the program. The command line used for the first example is:
   tralics tpa -configfile ../config_tpa -verbose
It tells the program to translate a file named tpa.tex, in verbose mode, using a given configuration file. The verbose option is equivalent to a \tracingall command in the TeX source (its effect is to print a lot of things in the transcript file); moreover some additional lines are printed on the screen (especially concerning configuration).

Tralics analyses first the arguments. All options start with a single or double dash. Instead of -configfile foo, you can say -configfile=foo (whitout spaces). For historical reasons, you can also say -config_file=foo, and, since version 2.8, the short form -config=foo is allowed. There is exactly one argument to the tralics command that does not start with a dash, it is the name of the TeX input file (extension .tex is optional).

The transcript file (here tpa.log) is opened after all arguments have been successfully parsed. Almost everything printed on the terminal is also printed to the transcript file. In the example, the version number is printed, the Copyright notice, then the line: `Starting xml processing for tpa'. The TeX input file is read next, once and for all.

3.2. Finding the Potential Type

Step S1 shows Tralics examining the configuration file, in order to find the Document Type. This information will be used to select a part of the configuration file, or the whole of a tcf file, using the rules that follow.

  1. If you say tralics -noconfig, then no configuration file is read at all.
  2. If you say tralics -configfile=foo, then Tralics will print Trying config file from user specs, and try to use this file.
  3. If you say tralics -configfile=foo.tcf, then Tralics will print the same as above; it will also search the file in the `confdir' directory.
  4. If the source file contains % tralics configuration file 'foobar', then Tralics will print Trying config file from source file, and try to use this file. In case of failure, and if the name 'foobar' contains no dot, the suffix .tcf is added, and the next rule is applied.
  5. If the source file contains % tralics configuration file 'foobar.tcf', then Tralics will print the same as above; it will also search the file in the `confdir' directory.
  6. The default configuration file is named .tralics_rc (or tralics_rc on Windows). The current directory is looked at first, then the tralicsdir, finally the home directory.
  7. If you say tralics -dir TOTO, or tralics -dir=TOTO, then TOTO/src/.tralics_rc is the second try.
  8. The home directory, or its src subdirectory, is searched next. (Depending on the operating system, this can fail, because there is no standard way of defining the home directory of the user).
  9. If you set the shell variable TRALICSDIR to somedir, or RAWEBDIR to somedir, then somedir/src/.tralics_rc is the last try. If neither variable is set, then some default location will be used.

In the configuration file, you can use # or % as comment characters. This file contains declarations inside groups (for instance BeginAlias ... End on lines 11--14 or BeginType ... End on lines 16--25). The end of a group is defined by a line that starts with the 3 letters `End', followed by anything. Some commands are allowed outside groups.

In the example, you can see that the user defined configuration file has been read and an important information extracted from line 8, namely 'Configuration file has type \documentclass' and Tralics has Seen \documentclass tpa.

The documentclass is defined as the argument of the first \documentclass command seen in the TeX source. This may be different from the real document class because of one of the following

If you say tralics -type FOO, then the type will be FOO. Otherwise, if the configuration file contains, at toplevel, the declaration Type =fOO, then the type will be fOO. If the configuration file contains Type =\documentclass, then the type will be the documentclass, as explained above. If you say nothing, then the type is the documentclass. This will be called the potential type, that needs to be converted into a real type.

3.3. Finding the Type

When Tralics enters step S2, it says: Potential type is tpa. It tries to match this type with all the types that are defined. In this case they are four types defined, default (l.16), tpa (l.48), unused (l.94) and eurotex (l.134), and one of them is to be selected.

In case of failure, aliases are considered. There is line 12 that says that `report' is the same as `article', but `article' is undefined, so that this line discarded. In the same fashion, `inriaslides' and `foiltex' are aliases to `slides', which is undefined. Near the end of the file, lines 154 to 156 say that `foo' and `foo2004' are alias to tpa, which is undefined. Thus we have only two useful lines, one that says that `foo' is an alias for `TPA' and `unused', while `foo2003' is an alias for `unused'. Only the first valid alias for `foo' is considered.

Hence, for this configuration file, aliasing adds two new types: `foo' for `TPA', `foo2003' for `unused'.

In the case of the first example, new syntax, we use the standard configuration file. This defines a single type, std, plus some aliases. We show here the content of it (without the comments).

Type = \documentclass

BeginType std
  DocType = std classes.dtd 
End

BeginAlias
  torture torture1 torture2 
  std report book article minimal
End

Since version 2.8, tcf files are considered. This means that, if the potential type is tpa, and the file tpa.tcf exists (or its alternate location ../confdir/tpa.tcf), it will be used. This is what happens in our example and Tralics says: Read tcf file for type: ../confdir/tpa.tcf. If the potential type is, as 'torture2' above, aliased to `torture', and if `torture' is not a type defined in the configuration file, then torture.tcf and ../confdir/torture.tcf are considered; if the file exists, it will be read, and `torture' becomes the effective type.

If aliasing fails, a second try is made, ignoring final digits in the name. Thus `fooNNNN' is equivalent to `tpa', whatever NNNN (if NNNN is a number, different from 2003). Since the standard distribution contains a file ra.tcf, if you say \documentclass{ra2003} (and this is required for the RAWEB, year 2003), the type will be `ra'. See example of the raweb.

In case of failure, the first type of the configuration file is used (this is `std' for the default configuration file).

In case of failure (if the configuration file does specify no type) then `book', `report' or `article' classes are recognized as a special case. In the same fashion, if no \documentclass command is seen in the TeX source, then plainTeX input is assumed. No error is signaled, but no parameterization code is executed.

In our example, we get the message Using type tpa. Everything that is out of the `tpa' scope is ignored from the configuration file. If a tcf file, is given, its entire content is considered.

3.4 Using the Type

What remains in the configuration file, after removing what is irrelevant to the current type, consists in two blocks: BeginTitlePage ... End (lines 50 -- 86) BeginCommands ... End (lines 87 -- 89) and some isolated commands (in this example there is only line 49, but for other types lines 17 -- 20 or lines 29 -- 43 are considered).

From now one, all line numbers refer to the tpa.tcf file; the first block is at lines 311--347, the second at lines 348--350, the isolated line at 308.

The important command is the `DocType' line. Tralics says: dtd is tpa from tpa.dtd (standard mode). It might say something like: dtd is raweb from raweb.dtd (mode RAWEB2003), case where you are creating Inria's Raweb for year 2003.

Now Tralics prints: Ok with the config file, dealing with the TeX file...

3.5 Bootstrap code

In the transcript file, at the start of Step S3, you can see that the source file has 38 lines. After that, the transcript file contains, for every source line its content, and for every command executed some information (this is because we are in verbose mode).

Some lines are not shown, they are of the form \let\endgraf\par. Then comes a bunch of lines of the form Defining \InsertTitlepage as \TitlePageCmd 0. Almost every line from the TitlePage block is associated to such a line. In fact, every line that start with a backslash (a LaTeX command) defines a command, and if you say alias \foo, then \foo is an alias to the command defined on the previous line. In the case of <UR> -, no command is created, but such lines are nevertheless useful.

Each line is followed by a second line, of the form usual <abstract ab='AB1'/> (flags +par +env). We shall see in a moment how the flags have to be interpreted.

You can see this in the transcript file:
[1] \cmdb{\cmdBval}
++ End of virtual file.
This is a consequence of the B or C flag, as explained later.

The transcript file contains [1] \def\cmdAval{CMDA}, plus some other lines, not shown here. This is the first line of the TeX source.

4. The titlepage info

In the configuration file, there are some lines that define the titlepage info. In this way, you can add, to Tralics, some commands like \author, \title, etc, that can be used (exactly once, at most once, at least once, as you like) before the \maketitle command (you chose your own name, variants are allowed). These commands should appear before the main text; they are forbidden in a paragraph (you can always insert a \par before them). Only one \maketitle command is allowed in the document.

4.1 Syntax

Each line of the configuration file in the title page info section is formed of one to four tokens; these have a type, say A, E, S, C; the interpretation of the line depends on the lists of these types.

Our example starts with a CESS, followed by an AESS, and two CES. After that comes a CEES, and the first E has plus as modifier. The order of elements is important. There can be only one CESS element, and it has to be the first. You can put an AESS only after an AESS or a CESS. You cannot put an AS element after a CESS, AESS or E.

4.2 Semantics

We shall discuss the meaning of these token list, in the order of use in the example.

CESS
Example: \cmd <elt> "att1" "att2".

This declaration has to be the first in the list. It defines a command \cmd, that can be used only once in the document, after \begin{document}. The effect is to insert the <elt> element into the XML tree. In what follows, we shall call it the TPA element. This element is formed of other elements defined by the titlepage info, the names of these elements are statically defined, their content is dynamic (i.e. the names depends on the configuration file, the content on the TeX document). The string att1 is a list of attributes added to the TPA element and the string att2 is a list of attributes added to the document element.

There is an example on lines 312 and 313; the title page element is named `titlepage', it has two attributes, with value `att1' and `att2' with value `foo1' and `foo2'. If used via the command defined on line 312, it has two additional attributes `a1' and `a2', and the document element has attributes `from_type' and `from_tpa'. There is another example line 207; the attribute name is empty, thus the attribute value will be ignored by the translator. See section Special action for the use of this special value.

AESS
Example: alias \cmda "att1p" "att2p". This declaration is valid only after a CESS declaration (or after another AESS declaration). It defines a command \cmda that can be used instead of \cmd (only one of these commands can be used). The result is the same; however att1p is used instead of att1 for the attribute list of the TPA element, and att2p is used instead of att2 for the document element (same remark as above for special attribute values in att2p).

In what follows, the \TPA command means one of the commands defined by this rule or the preceding one.

CEES

Example: \cmdb +<master> <local> "val". Note that the plus sign is required before the master element. Examples can be seen on line 316 and 317. See comments below.

This declaration has as side effect that the TPA element will contain a <master> element, formed of a number of <local> elements. Initially there is only one, initialized with "val".

The declaration has another effect, it defines a command \cmdb, that has to be used before \TPA command. It takes one argument, and creates a <local> whose content is the translation of the argument. This element is added to the end of the <master> element. Note that the default value is removed in case at least one value is given.

CES
Example: \cmdc <elt> "value". The element can have the flags p, q, e or E, and the value can have the flags +ABC. The effect is to define a command \cmdc (or an environment cmdc if the E flag has been given), that can be used only before the \TPA command. The argument of the command, (or the content of the environment) is translated, put in a <elt> element, and added to the TPA element.
If no flag is given for the element, paragraphs are forbidden in the argument. If you want to use paragraphs (either \par or \\) you must use the P flag (lower-case letter). In the same fashion, a lower case E means environment without paragraphs, an upper case E means environment with paragraphs. If the q flag is given, paragraphs are forbidden, but you can use \\, which is ignored. (in fact, the command reads an optional star, an optional argument, and the result is replaced by a space). Other flags are described in the section Interpreting the CES flags below.
There is a special trick for the case where the name of the element associated to the command is empty. Assume that the configuration file contains \cmd <> +"text". In the case where the user does not use \cmd, an error will be signaled, and text will appear in the resulting XML. If the user says \cmd{foo}, then Tralics remembers the use and issues no complain. Moreover, it reads the argument, and pushes foo\par in the input stream (the reason why \par is executed is to make sure that Tralics remains in vertical mode).
CCS
Example \cmdd \cmde "text". The effect is the same as \def\cmdd{\cmde{text}}. However, "text" is not translated, it is taken verbatim.
AC
If you say alias \foo, then \foo is an alias for the command defined on the previous line.
E
Example: <UR> -. The dash after the element is required. Another example is: <sUR fr='unitÚ de recherche' en='research unit'> -. In this second example, we have an element named `sUR', that has two attributes. The effect is to put, in the XML result, this element (with its attributes), and its content is a list of items declared in the configuration file (the list can be empty).
CE
The syntax is \cmdf ?+ <elt>. Example: \sURsop ?+ <sUR fr='dans le sud'>. This has as effect to define a command, here \sURsop, that takes no argument, whose effect is to insert, to the element <elt> (that must be defined by the previous rule) an empty element, whose name is cmdf, and that has the attributes of <elt>. In the example, we would get: <sURsop fr='dans le sud'/>.
CEE
The syntax is \cmdg ?+ <elt> <eltx> . Same as above, but the element created is <eltx> instead of one named cmdf with the attributes of <elt>. Example \sParis ?<sUR> <Rocq en='nearparis'> gives <Rocq en='nearparis'/>.
S
The character string is inserted verbatim in the XML tree.
AC
The syntax is execute \cmd or action \cmd. When Tralics sees the \InsertTitlepage command, it executes \xbox{}{\cmd}, this gives an XML element that is inserted in the current tree.

Note. Since Version 2.8, Tralics can produce UTF-8 output. Some arguments are copied verbatim from the input to the output, independently of this setting. The example shown above for case 'E' contains the string fr='unitÚ de recherche' (see configuration file, line 71). If you read the file containing this string in UTF8 mode, an error will be signaled by Tralics. If you use it in UTF8 output mode, no error is signaled, but the result is invalid. The solution is to use seven bit input and character entities, like &#xe9;. See line 332 of the configuration file, where this patch has been applied.

4.3. Interpreting the CES item

If the titlepage contains \cmd ?<elt> ?"value", this defines a command \cmd that produces an element <elt>, with a default value "value". The first question mark has to be replaced by nothing or one of p, q, e or E, the second by one of A, B, C or plus sign. If e or E is given, then an environment is defined, instead of a command.

Look at the second transcript file, line 649. First, the translator sees a command; here it is \lastname, but what the translator knows is the command code and its subcode; no name to is attached to this command code so that you will see ``{(Unknown)}''. The translator executes a piece of code that depends on the command code; this piece of code prints line 650, namely ``{\titlepage 8}'', where 8 is the subcode of the command (the transcript file says, on line 330, ``Defining \lastname as \TitlePageCmd 8'', so that this is coherent.) The title page data structure, contains for line 8, not only the name of the element to create, but also the name of the command. Thus, you can see on line 651 the name of the command, ``{\titlepage 8=\lastname}'' and on line 652 the start of the action ``{Push nom 1}''.

The effect of \cmd{foo} is roughly the same as \setbox8\xbox{elt}{foo} (replace `\cmd' by the name of the command, `elt' by the name of the element, and `\setbox8' by the effective location in the titlepage structure, this can differ slightly from the index of the command in the table). In the first version of Tralics the argument of the command was read, translated, and the stack was popped after that. As a consequence category codes were fixed once and for all. When you say \setbox8=\hbox, then TeX will execute the \hbox command, and push a special marker on the stack. When the closing brace of the command is found, TeX will fix the glue of the box, then pop its special stack, and then finish the assignment. No argument is read, category changes are allowed, all modifications are local to the group defined by the braces.

In the case of a titlepage element, another special marker is put on the stack by Tralics. In general, the opening brace in \cmd{foo} does not change the grouping level (this is different from \hbox), so that the transcript file will contain: +stack: level = 2 for titlepage argument (see transcript file, line 653). In a simple case, for instance \Atitle, the opening brace is sensed line 539, the closing brace on line 547, and this is followed by a line of the form ``+stack: level - 2 for titlepage argument''. This means that the current level will decrease from 2 to 1, which is the bottom level (what you do not see in the current version is that the level is incremented by one, when the special closing brace is seen).

Assume that \cmd@hook is defined. Then \cmd{foo} behaves like \setbox8 \xbox{elt} {\cmd@hook {foo}}. This does not work if you use `cmd' as an environment. In fact, the argument of \foo is read, and the token list {\cmd@hook {foo}} is pushed back in the input stream. An example is given in the configuration line 263. The transcript file, on line 653, shows the inserted brace, not the initial one; on line 654, you see the call of \lastname@hook. On line 657, you will see ``stack: level + 3'', this is because the hook expands to \textit{#1}, and this is the same as {\it #1}. On line 660, you see the closing brace that delimits the argument of the \it command, on line 664 the closing brace of the titlepage command, and on line 667 another closing brace (input line 7 has the form \author{XYZ}, the \author command command is defined to be \relax by the configuration file line 238, the 3 items X, Y, Z are commands. The transcript file says, line 666: ``{Pop 1: document_v nom_t}''. This means that the current element (to be popped) is a <nom> in a <document>.

Assume that \cmd@helper is defined. Then \cmd{foo} behaves like \setbox25 \xbox{elt} {\cmd@helper foo}, and \begin{cmd}foo\end{cmd} behaves like \setbox25= \begin{xmlelement} {elt}\cmd@helper foo\end{xmlelement}. More precisely: if the command is defined, it is inserted and executed. Braces are group delimiters. If we consider the case of `Btitle', whose execution starts line 554, we can see line 556 that the opening brace that follows is a special one (that leaves the level unchanged), and on line 557 that the level is increased by one. The closing brace is seen on line 568, and line 570 shows that the level goes down from 3 to 2 (as previously, you do not see it raising to 3). But you see the effects: on line 569 the value of \@nomathml is restored. Line 572 matches line 557: the brace level goes down from 2 to 1.

The situation is similar for environments. On lines 805, 849, \begin{Aabstract} or \begin{Babstract} is seen. On lines 806, 811 (and similarly, 850, 855) we have two lines that say that the current level increases or is unchanged. This is independent of the existence of a helper or not. On line 842 and 887 we execute \end{Aabstract} or \end{Babstract}. In general, this expands to \endAabstract. But this is a special case, so that, after restoring what is needed (line 888 for instance), and unwinding the stack (line 890), execution resumes with popping the XML stack. As is the case of a non-environment, you see the level decrease twice, but one of them (line 891) is immediately corrected.

4.4. Interpreting the CES flags

We finish the discussion of a command defined by: \cmd <elt> ?"value", where, instead of the question mark, you can use one of A, B, C or plus sign. You can also say nothing. Then, if the command is never used, <elt>value</elt> is added in proper place to the TPA element.

On the titlepage example, lines 342 --345, we define \cmdp, \cmdA, \cmdB, \cmdC in a similar fashion, but add a flag before the value. None of these commands is used in the TeX file (first example), and you can observe the following facts.

4.5. Comments

The purpose of the CEES construction is to define a list of authors. See example 1, where we have four authors, each defined by a single name. Example 2 is more complicated, but incomplete. The idea would be the following: the XML file contains a <auteur> element for each author (the configuration file works for a single author, elements are hard-coded l.213, l.221. Authors can be divided into real authors and contributors, the number of them is automatically computed (l.222) and l.223). It is easy to create the <nombre_auteurs>: define a CES command, and do not use it, add flag C; then some command is evaluated at the end, when the number of authors is known.

The source file shows a \author command, containing the name (structured via 3 commands) but not the affiliation (address, email, etc), that are defined elsewhere. Very often, this is given via footnotes, or special notes, like \thanks. Some people use shorthand notations like {A,B}@foo.bar for A@foo.bar and B@foo.bar.

4.6. Special action

When you launch Tralics with the option -tpa_status=X, this controls what happens after the titlepage has been translated. If X is all (or starts with the letter `a'), then the whole document is translated as usual. If X is title (or starts with the letter `t'), then only the titlepage is translated. The resulting XML will contain the TPA element, maybe some junk, maybe the bibliography. If X is config (or starts with the letter `c'), then the result depends on the configuration file, more precisely on the second attribute list of the titlepage command (see examples lines 207 or 312). In the case where one of the attributes of att2 has the value 'only title page', the translation stops. Note that the string above contains two spaces and two single quotes. If you really want this string to appear in the XML, replace the quote by &apos;. Any other value has the following meaning: compilation stops if an error was signaled, continues otherwise. By default, the configuration file has precedence.

If translation stops after the titlepage, the bibliography is not inserted. Exception: if 'translate also bibliography' (with the quotes), appears in the attribute list, then the equivalent of \nocite{*}, \bibliography{\jobname} is executed. This has as side effect to insert the bibliography at the end of the document.

Lines 244, 245, 246 of the configuration file contain the magic command \@reeavaluate. When you say \@reevaluate\foo\bar{gee}, then commands \foo and \bar are called with gee as argument, as in \def\@reevaluate#1#2#3{#1{#3}#2{#3}}, but the command can change the category codes. A star is allowed, and the behaviour is as in \def\@reevaluate*#1#2#3{\begin{#1}#3\end{#1}\begin{#2}#3\end{#2}}, There is an example in the transcript file line 375. We say \def \title {\@reevaluate \Atitle \Btitle}. On line 378 there is \def\abstract {\@reevaluate* {Aabstract} {Babstract}}. The star form of the command says that we consider environments. You can see the first use on line 529.

First, Tralics reads the optional star, then two arguments. After that, it reads all characters up to a closing brace, or (in the case of a star) the end of the environment. It can happen that some characters have already been read and tokenized. In this case, the token is converted into a character string (with a leading backslash, and a trailing space if the token has at least two characters, or has a single character that is a letter (this is independent of the category code). When characters are read from the buffer, the backslash character is considered specially: the character that follows it is normal. Otherwise an open brace increases the brace counter, a closing brace decreases the brace counter. In the case of \@reevaluate*, parsing ends when the end of the environment is seen; this means that Tralics counts the number of \begin and \end at brace-level zero (the command name must be followed by a non-letter), and stops when there is one more \end than \begin (the characters in this \end token are not read).

On line 530 and following, you can see see the list of all these characters; they are put in a local file that has a two lines, first is read at l.534, second is read at l.551, end of file detected at l.574 The interesting part starts on line l.558. Originaly, we had \catcode `\$=12 \@nomathml=1, its effect is to replace the math formula $L^\infy$ by $L^\infty $. Translation is given here. A better solution would be to use \@nomathml=-1 (not category code changes!). Then the translation of the math formula will be <texmath type='inline'>L^\infty </texmath>. The translation of the whole document is given here.

In the case where \@nomathml is zero (normal case) or negative (as above), parsing a math formula produces lines like l.543 (or l.564) because a math formula is a group, maybe other lines that increase of decrease the grouping level, maybe lines that show expansion of all local macros, followed by the end-of-group line, followed by the list of all tokens read (see line l.827 for a more complicated example). After that, the math formula is translated. You may see lines of the for l.829 whenever some tables are reallocated.

Consider now line l.781 in the transcript file. We have the start of some environment. A new grouping level is created (line 784), and the command associated to the environment is executed (line 785), and this calls the magic command \@reevaluate. All sources lines, from 24 to 27 are read (only the first four characters of line 27 are consulted, none of them is read). But Tralics reads one token (this should be the \end token) and a sequence of characters (in the same way as the \end command reads is argument); it does not test that this sequence is the current environment. It closes the current environment (see lines 791, 792). After that, Tralics inserts the set of lines, of the form \begin {cmd1} text \end{cmd1}% \begin{cmd2} text \end {cmd2}%. As above, a virtual file is created. It starts at line 802 and lasts until 893.

5. Classes and packages

5.1 User configuration file

If you compile file foo.tex, then Tralics will insert the line \InputIfFileExists*{foo.ult}{}{} just near the line with the documentclass declaration, if there is any. As mentioned above, there are other insertions, and in the transcript file, they are all after the line containing the \documentclass declaration. This is wrong in the case where an optional argument follows the declaration on the following line: we cannot insert text between the command and its argument.

In the current version, all insertions are before the line containing the \documentclass. The order if the following: all commands from a BeginCommand block in a configuration file (as Insert 2 above), then insertions from the titlepage (as Insert 1 above) then the InputIfFileExists line.

The semantics of \InputIfFileExists is the following. an argument is read, here, foo.ult. If the file exists, it is input; moreover the first token list is added to the current stream, otherwise the second list is inserted. Here both lists are empty, so that nothing special happens. The star after the command has a special meaning: it says that the @ character should be of category code 11 (letter) while reading the file. The old category code is saved on a special stack.

If the file uses the document class 'bar', then bar.clt will be used, and if package 'gee' is required, gee.plt will be used. The files bar.clt and gee.plt are looked at it the current directory, then in the directories specified by the confdir options to the program, then in a default directory, in the same way as tcf files, described here. No error is signalled if one of these files is missing. The @ character has always category code 11 (letter) while loading these files. Normally, bar.clt contains the same code as bar.cls (we shall explain later what is in the standard classes), and gee.plt contains the same code as gee.sty.

In order to implement commands like \LoadClassWithOptions, a table of all packages or classes, loaded not not, is maintained somewhere. A global variable contains the current position in the table, it is saved on the stack mentioned above, set to zero when a file opens, restored when a file closes. A command like \DeclareOption adds an option to the current class or package. If you say \input{baraux}, the content of the file `baraux' does not belong to the current class or package. In order to share options between standard classes, Tralics provides \InputClass{baraux}. This loads the file baraux.clt (found in the same way as aux.clt), but the content is added to the current class.

The following shows a bit what happens. The transcript file has a line for every file whose existence is tested. The file 'article.clt' identifies itself via \ProvidesClass line, it loads std.clt, that identifies itself with a \ProvidesFile. You can see how the integer cur_file_pos is restored to zero (outside class or package, or to one, the class). You can also see the options that are executed (in verbose mode, you would see the code of the options as well). The first list corresponds to the \ExecuteOptions command of the class file, the second to the \ProcessOptions

++ file article.clt does not exist 
++ file ../confdir/article.clt exists 
++ Made @ a letter
++  Opened file ../confdir/article.clt; it has 25 lines
Document class: article 2006/08/19 v1.0 article document class for Tralics
++ file std.clt does not exist 
++ file ../confdir/std.clt exists 
++ Made @ a letter
++  Opened file ../confdir/std.clt; it has 51 lines
File: std 2006/08/19 v1.0 Standard LaTeX document class, for Tralics
++ End of file ../confdir/std.clt
++ cur_file_pos restored to 1
{Options to execute->letterpaper,10pt,oneside,onecolumn,final}
{Options to execute->}
++ End of file ../confdir/article.clt
++ cur_file_pos restored to 0


Valid XHTML 1.0 Strict back to home page © INRIA 2004, 2006 Last modified $Date: 2008/08/11 10:10:08 $