A Brief Introduction to Regular GELLMU

William F. Hammond

Department of Mathematics & Statistics
The University at Albany
Albany, NY 12222, USA

September 2000
Last revision: January 10, 2010
(still needs revision as well as a better proof reading)


Note: Since this document was written, there have been some enhancements to the system it describes. For more information see the GELLMU Manual.

Table of Contents

1  Another Mark-Up Language?... *
2  How did GELLMU arise?... *
3  Translation of GELLMU Markup... *
4  Advanced GELLMU vs. Basic GELLMU... *
5  Some Fundamentals on Regular GELLMU... *
5.1  The Markup Commands... *
5.2  Multiple Argument/Option Syntax... *
5.3  Mathematics... *
5.4  Examples... *
6  Notes... *

1.  Another Mark-Up Language?

The purpose of regular GELLMU1 is to provide a markup language with provision for mathematics under a system design that is structured enough to admit rigorous, modularly-staged, automatic processing into many other formats while at the same time being a comfortable input format for those accustomed to using LaTeX. There is not a set list of translation target formats. Examples of possible translation targets include:

GELLMU is not LaTeX although its style of markup resembles that of LaTeX. An example of how there is provision for “enrichment of information” lies in the markup used for a quoted phrase in the source for this document.2 Whereas the usual LaTeX way of writing the phrase in the last sentence would be

``enrichment of information'',

and that markup would be meaningful in regular GELLMU, the markup

\quophrase{enrichment of information}

can be detected more easily by a processor looking for quophrase contents, while the separate double-quoting indicators may or may not be adequate for such a processor seeking to glean similar information. Furthermore, the effort by a processor to glean such information might be spoiled by the occurrence of one or more unbalanced double-quoting indicators. Possibly even more useful ways of marking the same phrase that still would yield the same printed appearance, depending on the overall context, might be:

\jargon{enrichment of information}


\topic{enrichment of information}.

An important characteristic of this markup is that the processing of a document always begins with a faithful translation to an instance under the umbrella of “Standard Generalized Markup Language (SGML)”. It is widely known in the text processing community (though not widely observed in the mathematical community) that sufficiently structured markup languages under SGML can be robustly converted to many other formats. GELLMU has been designed to provide a bridge from LaTeX-like markup to the world of SGML.

The translation from source markup to SGML markup is performed by my program called the GELLMU Syntactic Translator. Once that translation is performed the author's document exists in the world of SGML, and in that world many programs, almost all of which are very highly configurable, are available for automatic processing.

Beyond the syntactic translation from source markup to SGML the first release of materials from this project will contain a didactic article SGML document type, and some of the statements herein are, for illustration, non-explicit allusions to the article document type, which is just one example of an SGML document type in the advanced GELLMU category.

Realistically today, SGML is a document standard for in-house work, and its “subset” XML3 has become the only shareable form of SGML. Major SGML document types such as DocBook and TEI have canonical translations to XML. Still one does not abandon SGML in an authoring environment, because some of the features of full SGML provide substantial convenience for authors.

What added convenience there is in using LaTeX-like markup as a writing interface for XML is greatly aided by taking advantage of a few of the features in full SGML that are not available in XML. On the other hand, basic GELLMU is available for the direct preparation of an XML document as the direct output of the syntactic translator.

2.  How did GELLMU arise?

Aside from markup languages under SGML another example of a markup language that is structured enough to admit robust conversion is Texinfo, which is the markup language of The GNU Documentation System. The “Info” side of Texinfo is the Info hypertext self-documentation system within the popular fully programmable editing system GNU Emacs. Documents authored in Texinfo can be automatically converted into both Info (for viewing in an GNU Emacs buffer) and TeX (for printing).

The Info system is a hypertext system that pre-dates the World Wide Web (WWW) and the latter's default presentation markup language known as HTML, which falls under the SGML regime. One should, therefore, not be surprised to learn that with the dawn of HTML Lionel Cons of CERN (the birthplace of the World Wide Web) furnished the world with a PERL program for converting a Texinfo document into HTML with (internal) hypertext references.

Although Texinfo is a TeX-like markup language, it makes wide use of the character @ as command-introducer where in TeX one would (normally) use \. It seems that many authors find this superficial difference uncomfortable. Ulrik Vieth of The TeX Users Group (TUG) wrote an Emacs Lisp program that served to convert into Texinfo a LaTeX file written in January 1998 as the draft report of the TUG Working Group on a directory structure for TeX systems.

This example, which was an ad hoc conversion, played an important role in my arrival at this design for a bridge from LaTeX to SGML.

3.  Translation of GELLMU Markup

GELLMU is essentially, but for appearance, a markup language under SGML. This means that after a document is translated by the syntactic translator into the corresponding markup language falling formally under SGML, any general-purpose SGML processing tool can be used.

The project's prototype production system is built around Regular GELLMU, which uses features for enhanced authoring convenience in the syntactic translator that require some assumptions about the SGML document type under which the author is writing. Basic GELLMU, on the other hand, may be used as a simple LaTeX-like markup interface with a large number of document types, including XML document types.

Two SGML processing tools that are easily available form the core of the prototype production system.

  1. James Clark's nsgmls, an SGML parser that is part of his SP library package.

  2. David Megginson's PERL program sgmlspl that is an interface for his Perl library SGMLSPM. It may be found at CPAN or at his office. sgmlspl is a programmer-friendly interface for writing simple or complex “event-driven” translators (to other formats) that operate on nsgmls output.

In the prototype production system a document under the didactic article SGML document type made with the GELLMU Syntactic Translator from LaTeX-like source markup is parsed by nsgmls and then passed through a simple sgmlspl translation to yield an instance under the corresponding XML document type.

In the prototype production system the XML version of an article may be translated either to HTML or to LaTeX. Other translations are possible. The two existing translations are done with sgmlspl.

But a number of other freely available translation tools may be considered for the XML version of a GELLMU article including:

  1. James Clark's jade, an engine for DSSSL with a wide variety of possible translation targets.

  2. James Clark's xt, an XSLT engine also with a wide variety of possible translation targets.

  3. David Carlisle's xmltex, a package for TeX, the Program, that enables its user to write what amounts to a “style sheet” in TeX, rather than, say, XSL, for formatting (with its own parse) an instance of an XML document type in TeX (actually DVI since one does not get a chance to “see” a TeX source image).

4.  Advanced GELLMU vs. Basic GELLMU

This document is about the regular GELLMU. It is important first to know a bit about basic mode and advanced mode. What is called regular in this document is an instance of advanced mode. The term regular is being used here to reference the didactic article document type, which is an SGML document type that admits the more elaborate and succinct handling of advanced mode while not being in any way incompatible with basic mode.

Neither the basic nor the advanced mode involves in any way adoption of the language of LaTeX. (But many command names under the didactic document type, mimic LaTeX command names.)

To use either basic or advanced mode one must be familiar with the SGML document type for which one is writing. Ordinary HTML is an example where the use of basic mode is indicated. Only a very few of the features in advanced markup not part of basic markup4 make any sense for use in the direct preparation of HTML with LaTeX-like input.

Basic GELLMU uses the LaTeX special characters ‘\’, ‘{’, and ‘}’ along with LaTeX-like argument/option syntax, where braces immediately following a command name indicate command arguments and square brackets, i.e., ‘[’ and ‘]’, indicate command options. A command corresponds to an SGML element, and in basic mode a command may have at most one argument, the content of which corresponds to SGML element content, and at most one option, the content of which corresponds to a list of SGML attribute specifications. Thus for example, in basic mode for HTML one may use the markup

\a[href="http://www.w3.org/"]{The World Wide Web Consortium}

to form the HTML anchor:

<a href="http://www.w3.org/">The World Wide Web Consortium</a> .

(The formation of anchors with the didactic document type in advanced mode is slightly more complicated because the characters ‘=’ and ‘/’, which may acquire special (and “overloaded”) semantic significance in mathematical contexts, are held for delayed evaluation as empty elements and because the syntactic translator, which does not recognize command names, regards this usage in advanced mode as multiple argument/option syntax, which is not part of basic mode.)

An example of the distinction between basic and advanced GELLMU is that in advanced mode it is possible and easy to arrange to have a blank line, as in LaTeX, represent the beginning of a paragraph. In basic mode for classical HTML one must5 use “\p” to begin a paragraph, and for the XML version of HTML one must also provide markup for the end of every paragraph, which may be done in several ways.

For some of the details on using the basic markup with HTML see Using the GELLMU Syntactic Translator to Write HTML.

5.  Some Fundamentals on Regular GELLMU

5.1.  The Markup Commands

There are several kinds of commands:

5.2.  Multiple Argument/Option Syntax

An essential point in the present design is that the whole system is built from components, each of which has its own function9. Consistent with this design the syntactic translator operates with knowledge of syntax but little or no knowledge of language.

Multiple argument/option syntax has been built into advanced mode as part of the overall idea of providing, where sensible, LaTeX-like features in a precise user markup interface for writing in document types under SGML and XML.

What are the rules for converting the multiple argument/option syntax in source markup into SGML? Direct conversion by the syntactic translator of this type of usage into XML is not available because such conversion requires some language knowledge and the program does not operate with knowledge of language at that level. One obtains an XML version of a document in the prototype production system by using a translator with minimal knowledge of the command vocabulary to create the XML version from an SGML version that is the immediate output of the syntactic translator.

In multiple argument/option syntax, which is much like that of LaTeX, arguments and options follow command names. Arguments are delimited by braces, i.e., ‘{’ and ‘}’ and options by square brackets, i.e., ‘[’ and ‘]’. There must be no white space between the arguments and options nor between the command name and the first member of an argument/option sequence.

Each command with a multiple argument/option sequence is translated to an open tag whose name is the name of the command. Each argument is translated to an ag0 element and each option to an op0 element lying in GELLMU's reserved name space. There are two exceptional cases.

  1. The first argument or option is an option inside which the very first character is a colon, i.e., ‘:’. This is the method provided in advanced mode for the direct entry of an SGML attribute sequence. The entire contents of the option string, apart from the leading ‘:’, which is discarded, are understood to be a sequence of SGML attributes for the SGML element whose name is the name of the command. There is no syntax check of the attribute contents by the syntactic translator. Such an attribute option is not treated as an op0 element. In particular, an attribute option is correctly followed immediately by a semi-colon, i.e., the character ‘;’, if and only if the corresponding SGML element is a defined-empty element under the SGML document type. Since SGML attributes correspond to very little of classical LaTeX, attribute options may be entirely ignored.

  2. The first argument is the only argument and there are no options apart from a possible attribute option. This case, which is extremely common, is exceptional relative to argument/option handling since the sole argument simply becomes element content without an ag0 wrapper.

5.3.  Mathematics

The Greek letter Gamma is marked, up as in LaTeX, with


The didactic article document type provides mathematical markup that is similar to that of LaTeX. For example, one may use the markup

\[ 2^{p-1} \equiv 1 \pmod{p^2} \eos\]

to speak of the congruence

 2^{p-1} EQUIV 1 (mod p^{2})   .

In this example the markup “\eos” is a formal end-of-sentence10.

In a LaTeX math environment, input text “abc” is set the same as the the input text “a b c” following the presumption that each glyph is a separate symbol. At one point I had planned to provide for a distinction between these two forms of input for article. However, I am concerned about author confusion; so there will be provision for having “abc” be a math symbol, but an instance will be invoked as “\abc”.

In math mode all symbols are assumed to be single-characters as in LaTeX. A \mathsym command (part of language design that, for the moment, has no extant implementation) in the document preamble may be used to specify that a string is to be regarded as a single mathematical entity whenever it occurs in math mode. This contrasts with LaTeX in that there is no formal declaration of math symbols.

mathsym will be a variant of newcommand, which is a meta-command that is handled entirely by the syntactic translator with no trace in the output. However, the design of mathsym, which will also be a meta-command but which will leave marks in the output, is not yet decided. The design issues surround how to provide the author who is its user with fexible open means of planting semantic information in the output without undue verbosity. The syntax might be simply

\mathsym{symbol-name} {presentation-defintion} [semantic-information-without-markup]

Omitting the option could make author usage equivalent to that of newcommand, but still trace information planted in the output of the syntactic translator could be sufficient that in the future a derived MathML object or a derived XML formatted object produced by translation and viewed in Mozilla might enable the reader to ascertain the symbol-name and then launch a search for other occurrences. The option would serve as a mechanism for the author to pass semantic enrichment information to downstream formatters.

5.4.  Examples

The following table refers to the didatic article document type. It demonstrates how source is translated by the GELLMU Syntactic Translator to the SGML version of article and from there is translated using sgmlspl to the XML version of article.

Source \emph{this}
SGML <emph>this</emph>
XML <emph>this</emph>
Source \tex;
SGML <tex/>
XML <tex/>
Source $ \frac{2}{3} $
SGML <tmath> <frac><ag0>2</ag0><ag0>3</ag0> </tmath>
XML <tmath> <frac><numr>2</numr><denm>3</denm></frac></tmath>
Source \anch[href="nil"]{Null}\anch:
SGML <anch><op0>href<eqs/>"nil"</op0><ag0>Null</ag0></anch>
XML <anch><anchref>href="nil"</anchref><anchv>Null</anchv></anch>

One may find many other examples of GELLMU markup in the project archive.

6.  Notes

Although the meta-command newcommand, which is handled internally by the syntactic translator, is now available, the variant meta-command mathsym is not yet available.

This document is still at draft stage.

It is itself a GELLMU document. The following versions are available:


  1. * GELLMU stands for “Generalized Extensible LaTeX-like MarkUp”
  2. * The base for relative URI references given here may be found at the end of the printed form of this document.
  3. * Subcategory might be a better term here than subset.
  4. * With several minor exceptions, one related to the direct writing of SGML attributes (which cannot contain markup and which do not have many parallels in LaTeX) and another related to the way of escaping the character ‘\’, everything about basic mode also applies to advanced mode.
  5. * There is a way, with the setting of several variables for the syntactic translator, to have blank lines begin new paragraphs in basic input for HTML
  6. * For the first alpha release “letter” means something matching the regular expression “[A-Za-z]”. This will not be a limitation of the syntactic translator after the alpha releases. I believe that GNU Emacs, from version 20, is fully capable of meeting the needs of GELLMU for supporting international character sets. At present, it is possible to use Emacs “word constituent characters” in names defined with newcommand, and there is no reason why the definition strings cannot involve standard SGML entity notation. It is also possible to define symbolic SGML entities in the SGML internal declaration subset, which is the content of an option (at most one) that follows the unique argument to the meta-command “\documenttype”. Additionally, the SGML version of the didactic article document type provides commands for the document preamble that may be used to construct an internal declaration subset for the XML version of the document. (Of course, someone using this markup to edit directly for a document type under XML would need to use the “\documenttype” option provided for the SGML internal declaration subset.)
  7. * It is possible to merge the inline math and tmath zones at any level of processing beyond the syntactic translator. (These are indeed the same in LaTeX.) The syntactic translator resists the temptation here to go beyond syntax, and with the didactic article document type the formatting to LaTeX inserts the LaTeX markup “\,” for a small horizontal space before and after math, but not before and after tmath.
  8. * The syntactic translator simply outputs the SGML defined-empty element brk0, which belongs to its reserved name space. The dual use of brk0 involves some SGML chicanery that is resolved during translation to the XML version of the article document type, where tabular is converted to table and non-tabular use of brk0 is converted to brk.
  9. * In the prototype production system based on the didactic article document type the output from each stage is available for examination and, where necessary, intervention. However, such use of intervention is intended only for temporary expedient use while a GELLMU system is being designed or enhanced. As with LaTeX, enhancement is an ongoing process.
  10. * Regular GELLMU recognizes ‘.’ followed by two blank spaces or by a newline as ends-of-sentence markup when these occur outside of mathematical contexts.