About gbh3
----------

gbh3 is used to extract Gambas documentation from C/C++ source files and
create .help files for a component from them. It is a single Gambas project
depending only on gb.pcre.

... and why?
------------

Components written in Gambas can already be documented in-code. This docu-
mentation is then stored inside their .info files. With a similar result
provided by gbh3 for C/C++ components, we can document all[*] Gambas in
its source files (at least my components will be) which makes it, IMHO,
easier to keep the documentation up-to-date. Also the help could be
displayed locally by the IDE or remotely once these .help files are imported
to the gambaswiki.org site.

Also, documentation can be bound to a specific source code version, so that
you get docs for the version of Gambas you are running, not only for the
development branch.

[*] I'm not sure about intrinsic functions, though. Should be feasible,
    looking at gbx_class_info.c...

Great! How do I prepare my component?
-------------------------------------

The No. 0 rule is to BE CAREFUL! There are quite some rules to remember:

1. Documentation syntax.

You may write a Gambas documentation comment like

  /**G
   * Here goes the documentation.
   **/
  BEGIN_METHOD_VOID(ClassName_MethodName)
  	/* ... */
  END_METHOD

or alternatively

  /**G Class Symbol
   * Documenting Class.Symbol
   **/

Instead of the **/ at the end, you may also write */. Cool, huh?

The second syntax can also be used to document the class per se:

  /**G Class
   * Class documentation here
   **/

The difference between both notations is that the first must immediately
precede a BEGIN_{PROPERTY,METHOD,METHOD_VOID) line that defines the symbol
to which the documentation refers. The second syntax documents the indicated
symbol -- no matter where the comment is.

This is intended to document GB_CONSTANTs or GB_PROPERTY_SELFs which don't
have a BEGIN_* line. It can neverthelss be used for any other symbol but
NEVER intermix both syntaxes on a single symbol.

The ClassName_MethodName part I call "(implementation) function name". The
particular function naming convention shown is not mandatory. The scripts
take the function name and look up the corresponding symbol in your
GB_DESC[].

2. Regular expressions.

The regular expressions used to filter these comments are somewhat strict.
You have to pass the spaces exactly as indicated in the line where your
comment text starts, i.e. <space>*<space>text. Also note that Class and
Symbol in the second syntax are case sensitive!

OTOH, I can't guarantee that the expressions will not consume junk that you
left behind where I didn't expect junk to be.

3. Strict structure.

DON'T try to confuse the parsers by not giving information it needs where
it expects it.

4. Synonyms.

The first match for a function name lookup in a GB_DESC[] is taken as the
"primary" implementation of that symbol. That means the help comment will
be attched to that symbol. All other symbols which call the same function
are considered synonyms and are *automatically* documented with the default
sentence: "This is a synonym for <primary-symbol>". DO NOT document them
separately using a syntax-2 help block!

5. Don't reuse method and property implementations (except for synonyms).

As the implementation function name is searched in the GB_DESC[]s in your
source file, you may not put the same function into different classes -- or
the parser may get confused.

Also, keep the GB_DESC[] in the same source file as the help comment.

Umm... I need an example
------------------------

gb.openssl and gb.data are partially documented that way. Digest.Hash() and
Digest._call() in gb.openssl show how to deal with synonyms. There the entry
for Hash precedes the entry for _call, thus Hash is the primary symbol and
_call a synonym.

To see a sample .help file (on stdout), you can do at the source tree root:

  $ gbx3 app/src/gbh3 -- gb.openssl