31. Pike AutoDoc markup

31.1. Syntax

31.1.1. Line orientation

The markup is line oriented. If you need to write a line which is very long, it can be broken into several lines. A trailing @ on the line indicates that it continues on the next line. The @ and the newline character will be discarded, and the lines merged. Thus:

  //! @variable thisVariableNameIsSoLong@
  //!YouJustCantBelieveIt

will appear to the parser as:

  //! @variable thisVariableNameIsSoLongYouJustCantBelieveIt

This is sometimes necessary because keywords that take parameters expect all the parameters to appear on the same line and treat the end of the line as the end of the parameter list.

The character '\r' is also discarded. The same applies to all other non-printable and control characters except for '\n' and '\t'.

In text (see the nonterminal 'text' in the grammar at the end of this file), a blank line with surrounding non-blank lines will give a paragraph break. Thus:

  //! - I love you, said Danny.
  //!
  //! - You have no right to come here after what you did to
  //! my little dog, screamed Penny in despair.

will generate the following XML:

  <p> - I love you, said Danny.</p>
  <p> - You have no right to come here after what you did to
    my little dog, screamed Penny in despair.</p>

31.1.2. Keywords

Keywords always begin with an at-sign: @. A @ is quoted by writing two of them: @@. There are four types of keywords (the keywords in []'s are examples of keywords of the different types):

  1. Meta keywords [@decl, @class, @endclass, @module and @endmodule] Must stand alone on one line, preceded only by whitespace. These are not really part of the markup. They must come before any text or other keywords in the doc block. They provide information about what is being documented etc. and do not result in any text in the documentation. Meta keywords have keyword specific parameter syntaxes.

  2. Delimiter keywords [@param, @member, @item, @note, ...] Must stand alone on one line, preceded only by whitespace. These are keywords that begin a section inside their block. They have no end marker, but rather the section ends when the next delimiter keyword on the same level is found. Can have parameters.

  3. Block/endblock keywords [@dl - @enddl, @mapping - @endmapping, ...] Must stand alone on one line, preceded only by whitespace. These open or close a block. If a block is started with @foo, it must end with a matching @endfoo. The lines inside the block can be divided into sections by using delimiter keywords. The start keyword can have parameters, but the corresponding end keyword can not.

  4. Short markup keywords [@ref{..@}, @i{..@}, ...] These are used in the text to perform cosmetic tasks, for example changing text to italic (@i), teletype(@tt) or marking a word as a reference to a pike entity (@ref). They can be nested, but a short markup keyword can not contain a keyword of types 1-3. They begin with @keyword{ and end with @}.

  5. The magic keyword @xml{ ... @} This is a special keyword that provides an escape to XML. All ordinary text inside it is passed unquoted to the output. However, the short markup keywords are still valid and may be used inside it, and thus @ must still be quoted with @@. < and > must also be quoted unless the intention really is to write XML. For example:

    //! He is so @xml{<i>italic</i> and @b{bold@}!@}!!

    will generate the following XML:

       He is so <i>italic</i> and <b>bold</b>!!!
     

NOTA BENE: Delimiter keywords (2) and block keywords (3) must stand alone (with their parameters, if any) on one line.

31.1.3. Delimiter keyword grouping

Delimiter keywords that indicate a new section inside the block can be grouped together in the interest of not writing the same docstring for multiple parameters etc. Delimiters are grouped if they appear on consecutive lines. For example, when documenting a method:

  //! @decl int dist(int x, int y)
  //!   Calculates the distance.
  //! @param x
  //! @param y
  //!   The coordinates of the vector.

Above, the two @param's x and y are grouped together and share the same docstring: "The coordinates of the vector.". It is an error to try to group together different keywords:

  //!   Error, can't group @note and @param:
  //! @param x
  //! @note
  //!   Don't use this function. At all. Ever.

31.1.4. Keyword parameters

After the leading @keyword (that may be preceded only by whitespace) on the line, the rest of the line is interpreted as a parameter list. The syntax of this parameter list can be different depending on the keyword:

  1. Special keyword parameter list syntax Here the parameters can be parsed according to Pike syntax or in some other way. Examples of keywords that use these kinds of special syntaxes are all the meta keywords, @member and @elem.

  2. Default parameter list syntax The meaning of parameters is determined by the order in which they appear, much like the arguments in a unix shell command line - hence they are not named as in HTML or XML. Parameters are separated by whitespace. If you wish to have a parameter string with whitespace in it, you must surround the string with a pair of ' or ". When the quoting character itself is in the string, a duplication is used to quote it:

    //! @question "He is a ""swapper""?!"

    However, if your parameter contains only one of ' and ", then it is smarter to choose the other one as the quouting character:

    //! @question 'He is a "swapper"?!'

    It is an error not to end a quote before the end of the line:

    //! @wrong "Aww, come on now, this worked fine in C64 basic!

    If a quoted parameter is too long to fit in one line, use the @ at the end of the line to merge it with the following:

    //! @right "Oh, joy! Now I can make my parameters just as@ //! long as I want them!"

    The parameters are not parsed so you can not have markup inside them. Pike style quoting is not used either, which means that if you write:

    //! @look "\n"

    The parameter will be a string with two characters, a backslash followed by the letter n.

31.1.5. Grammar

Here comes a BNF-ish grammar for documentation blocks. Note that before the parsing of the block, all lines ending with a @ will be merged with the next line (see (a) above)

docblock:
  metaline*, blockbody

metaline:
  start_of_line, white_space*, metakeyword, any_char_but_newline,
  end_of_line

blockbody:
  section?, (delimiter+, section)*, delimiter?

delimiter:
  start_of_line, white_space*, delimiterkeyword, parameterlist,
  end_of_line

section:
  (text|block)+

block:
  blockbegin, blockbody, blockend

blockbegin:
  start_of_line, white_space*, blockkeyword, parameterlist,
  end_of_line

blockend:
  start_of_line, white_space*, blockkeyword, white_space*, end_of_line

parameterlist:
  white_space*, (parameter, white_space+)*

parameter:
  qoutedparameter | any_char_but_white_space+

quotedparameter:
  ('"', (any_char_but_new_line_or_" | '""'), '"')
  | ('\'', (any_char_but_new_line_or_' | '\'\''), '\'')

text:
  (character|shortmarkup|xmlescape)+

xmlescape:
  '@xml{', any_char_sequence_not_containing_@}, '@}'

character:
  any_char_but_@ | '@@'

shortmarkup:
  shortmarkupkeyword, '{', text, '@}'

metakeyword, blockkeyword, delimiterkeyword, shortmarkupkeyword:
  keyword

keyword:
  '@', alpha_char+

endblockkeyword:
  '@end', alpha_char+

white_space:
  ' ' | '\t'

31.2. Pike autodoc inlining

The autodoc extractor works either in C mode or in Pike mode. The reason why the two modes are not the same is that they perform two very different tasks. The C mode only has to find comments in the file, and ignores the surrounding code totally, while the Pike mode on the other hand is supposed to be smarter and distill a lot of information from the Pike code that surrounds the comments.

Both work at the file level. That makes it easy to use for example "make" to generate documentation for the source tree. Another benefit is that the generation will not have to reparse all of the tree if only one source file is changed.

For Pike module trees, the extractor can recurse through the file tree on its own, but for C files, where the directory structure gives insufficient cues about what is what, there must be make targets set up manually. All generated XML files can then be merged together into the final Pike namespace.

31.2.1. C files

In C files, the doc comments look like:

  /*! yadda yadda
   *! yadda yadda yadda
   */

Note that only lines that start with *! count, so above are only two doc lines. Any whitespace before the leading *! is skipped, so that the lines can be indented freely.

In the C files, no parsing of the surrounding code is done. The context lies completely in the doc comments themselves, and the target of the doc comment is determined by special meta keywords that are not really part of the doc blocks, but rather modifiers that tell which Pike entity the doc is about.

  /*! @module Foo
   *!   ... doc for the Foo module ...
   *!           ...                 */

    /*! @decl int a()
     *!   ... doc for the method Foo.a() ...
     *!    ....                      */

    /*! @class Bar
     *!   ... doc for the class Foo.Bar  ...
     *!           ...                 */

      /*! @decl mapping(string:string) userprefs()
       *!   ... doc for the method Foo.Bar->userprefs() ...
       *!           ...                 */

      /*! @decl int a
       *! @decl int b
       *!   ... doc for the variables Foo.Bar->a and Foo.Bar->b ...
       *!      ...                      */

    /*! @endclass */

  /*! @endmodule */

The @module and @class too keywords are to work like segment directives in assembler source files. That is, you can have @module foo in several C files, if the module source is spread over multiple files. However, if you write doc for the module itself in several places, an error will be triggered.

31.2.2. Pike files

Doc comments look like:

  //! yadda yadda yadda
  //! yadda yadda

To be considered one doc block, the comments must be separated only by whitespace and \n, that is they have to be on adjacent lines in the code. Each doc block in the Pike code has one or more targets; the Pike entities (modules, classes, variables etc.) that the doc block is documenting. The target of a doc comment is the coherent block of declarations adjacent to (immediately before or after) it in the code, without intervening blank lines. Examples:

  //! Doc for alpha
  int alpha()
  {
    return 4711;
  }

  protected int undocumented;

  //! Error! This doc block has no destination!

  int beta;
  //! Doc for beta

  //! Doc for gamma, delta, and epsilon
  int gamma, delta;
  float epsilon;

  //! Error here!
  int zeta;
  //! ambiguous which doc to associate with zeta.

  int eta;
  //! Error here too! ambiguous which variable is documented.
  int theta;

  //! Doc for two methods. This is so UGLY! We strongly recommend
  //! using the decl keywords instead to accomplish this effect.
  int methodOne()
  {
    ...
  }
  int methodTwo()
  {
    ...
  }

  //! However, it can be useful sometimes, for really short methods:
  int very_short() { return 4711; }
  int even_shorter() { return 0; }

In Pike files, you can not use @class or @module to tell which module or class you are in. To document a class, you simply write:

  //! Doc for the class
  class CheeseMaker
  {
    //! You can even document inherits!
    inherit Earth : earth;

    //! Doc for CheeseMaker->a()
    int a()
    {
      ...
    }

    void create(string s) { ... }
  }

The parser will automatically identify a() as a member method of the class CheeseMaker, and will detect that Earth is inherited by CheeseMaker. If a class has no documentation comment, it's internals will not be examined, thus it is an error if a class contains documentation comments but is itself undocumented:

  class a()
  {
    //! @decl foo
    //!    ... doc for foo ...
  }

A special inlining case is that of functions and classes. When documenting these, the doc comment can be put between the head of the function/class, and the opening {, like this:

  class Roy
  //! Documentation for Roy
  {
    ....
  }

  int un_randomize(int x)
  //! This function takes a random number, and transforms it into
  //! a predictable number.
  {
    return x = 4711;
  }

If a doc block is the first in a file, and it has no target, then it is treated as doc for the file (or rather: the module/class that the file compiles into) itself. In any other case it is an error to have a targetless doc block. A target can also be set with the @decl meta keyword. If a doc comment begins with some @decl keywords, these @decl's act just like real declarations standing next to the doc. Thus:

  //! @decl int a(int x)
  //! @decl int b(int x)
  //! 	Here is some doc for these functions....

is autodocwise equivalent to:

  //! Here is some doc for these functions....
  int a(int x)
  {
     .....
  }
  int b(int x)
  {
     .....
  }

In case it is legal to have both an adjacent declaration and the @decl keyword at the block beginning. That is when you document "polymorph" methods. Then the adjacent declaration must be a method, and all @decl's must be methods that have the same name as the real method:

  //! @decl float cube(float x)
  //! @decl int cube(int x)
  //! 	Gives x**3.
  //! @param x
  //! 	The number to cube.
  int|float cube(int|float x)
  {
     ....
  }

The real method prototype is discarded in favour to the @decl'ed variants, who will be shown in the documentation instead.

One problem that is unsolved so far is how to handle #if .. #else .. #endif constructions. The approach so far has been to ignore preprocessor directives totally. For example, the parser does not handle:

  #ifdef MALE
    int bertil()
  #else
    int berit()
  #endif
  {
    ... body ...
  }

It a portion of the code is unextractable because it contains too much preprocessor macros and stuff, you can make the extractor skip it by using @ignore:

  //! @ignore

  HERE_ARE_SOME_STRANGE_THINGS
  #ifdef A
  A
  #endif

  //! @endignore

All @ignore-@endignore sections of the file are removed before any extraction is done, so they can cross class boundaries and the like. You can nest @ignore inside eachother. Another application for @ignore is to hide actual class boundaries from the extractor:

  //! @ignore
  class C {
  //! @endignore

    //! To the parser, this function appears to be on the top level
    int f() { ... }

  //! @ignore
  }
  //! @endignore

31.3. Pike autodoc tags

We have defined some different categories of markup, each with its own semantics. Seen from the parser, there are two main construct levels:

All markup can also be divided into categories based on their look and semantics; there are three categories here (with examples):

31.3.1. Meta level tags

These tags all serve the purpose of denoting what pike entities your comments should be tied to. This is particularly needed in (and in part only applies to) C files, where the autodoc extractor does not try to interpret and parse class and method definitions.

Keyword:   @module
Legal for: C files only (neither needed nor legal for Pike files)
Arguments: (the last segment of) the module name (ie no "." allowed)

Example: To document the module Parser.XML module, you need two consecutive module tags:

  /*! @module Parser */
  /*! @module XML */

A @module keyword sets the scope of documentation comments following it. @module tags nest, as shown in the example above, and must be ended with a matching number of @endmodule tags, as in:

  /*! @endmodule XML */
  /*! @endmodule Parser */

A @module keyword may also have a text child documenting the module itself:

  /*! @module Parser
   *!
   *! The common roof under which parsers of various kinds, such as
   *! @[Parser.XML], @[Parser.Pike] and @[Parser.C] are housed.
   */

There are two special @module targets available; the predef:: module and the lfun:: module. The predef:: module, although more of a scope than a module, contains the definitions for all methods in Pike's predef:: scope, as in:

/*! @module predef::
 *!   @decl int equal(mixed a, mixed b)
 *!   	This function checks if the values @[a] and @[b] are equal.
 *! @endmodule
 */

The lfun:: module scope does not legally exist in Pike, but it houses the docs for all lfuns (the virtual methods for fulfilling the Pike object API). An example:

/*! @module lfun::
 *!   @decl int(0..1) `!=(mixed arg1, mixed arg2, mixed ... extras)
 *!     The inequality operator.
 *! @endmodule
 */

This also means that referencing (via @ref{...@} or @[...]) the lfun documentation strings can be done using @[lfun::`!=] and the like, as can predefs via @[predef::sort()] et cetera.

Keyword:   @endmodule
Legal for: C files only (neither needed nor legal for Pike files)
Arguments: (the last segment of) the module name (optional)
>

When the optional argument to @endmodule is given (for code clarity), the extractor will verify that the module scope you close was indeed the one you opened, as in the @module example above. The following would trigger an error:

  /*! @module Parser
   *! @module XML */

  /*! ... some autodoc comments ... */

  /*! @endmodule Parser
   *! @endmodule XML */

while the same example, ending in

  /*! @endmodule
   *! @endmodule */

would be considered legal.

Keyword:   @class
Legal for: C files only (neither needed nor legal for Pike files)
Arguments: (the last segment of) the class name (ie no "." allowed)

Example: to document the Process.create_process class, you would use:

  /*! @module Process
   *! @class create_process */

And end the scope similar to @module:

  /*! @endclass create_process
   *! @endmodule Process */

Like @module tags, @class tags may be nested any number of levels (when documenting subclasses to subclasses to subclasses to ...).

Keyword:   @endclass
Legal for: C files only (neither needed nor legal for Pike files)
Arguments: (the last segment of) the class name (optional)

When the optional argument to @endclass is given (for code clarity), the extractor will verify that the class scope you close was indeed the one you opened, just as with @endmodule above.

Keyword:   @decl
Legal for: All source code (C and Pike files)
Arguments: (the last segment of) the identifier name (ie "." illegal)

The @decl keyword is used to target a specific (...more to come! :-)




                 +--------------------------------------+
                 | Pike autodoc markup - the XML format |
                 +--------------------------------------+

======================================================================
a) Introduction
----------------------------------------------------------------------

When a piece of documentation is viewed in human-readable format, it
has gone through the following states:

  1. Doc written in comments in source code (C or Pike).

  2. A lot of smaller XML files, one for each source code file.

  3. A big chunk of XML, describing the whole hierarchy.

  4. A repository of smaller and more manageable XML files.

  5. A HTML page rendered from one such file.
     (Or a PDF file, or whatever).

The transition from state 1 to state 2 is the extraction of
documentation from source files. There are several (well, at
least two) markup formats, and there are occasions where it is
handy to generate documentation automatically &c. This document
describes how a file in state 2 should be structured in order to
be handled correctly by subsequent passes and presented in a
consistent manner.

======================================================================
b) Overall structure
----------------------------------------------------------------------

Each source file adds some number of entities to the whole hierarchy.
It can contain a class or a module. It can contain an empty module,
that has its methods and members defined in some other source file,
and so on. Suppose we have a file containing documentation for the
class Class in the module Module. The XML skeleton of the file
would then be:

  <module name="">
      <module name="Module">
          <class name="Class">
              ... perhaps some info on inherits, members &c ...
              <doc>
                  ... the documentation of the class Module.Class ...
              </doc>
          </class>
      </module>
  </module>

The <module name=""> refers to the top module. That element, and its
child <module name="Module">, exist only to put the <class name="Class">
in its correct position in the hierarchy. So we can divide the elements
in the XML file into two groups: skeletal elements and content elements.

Each actual module/class/whatever in the Pike hierarchy maps to at most
one content element, however it can map to any number of skeletal elements.
For example, the top module is mapped to a skeletal element in each XML
file extracted from a single source file. To get from state 2 to state 3
in the list above, all XML files are merged into one big. All the elements
that a module or class map to are merged into one, and if one of those
elements contains documentation (=is a content element), then that
documentation becomes a child of the merger of the elements.

======================================================================
c) Grouping
----------------------------------------------------------------------

Classes and modules always appear as <module> and <class> elements.
Methods, variables, constants &c, however, can be grouped in the
source code:

  //! Two variables:
  int a;
  int b;

Even a single variable is considered as a group with one member.
Continuing the example in the previous section, suppose that Module.Class
has two member variables, a and b, that are documented as a group:

  <module name="">
    <module name="Module">
      <class name="Class">
          ... perhaps some info on inherits, members &c ...

        <docgroup homogen-type="variable">
          <variable name="a"><type><int/></type></variable>
          <variable name="b"><type><int/></type></variable>
          <doc>
            ... documentation for Module.Class.a and Module.Class.b ...
          </doc>
        </docgroup>

        <doc>
           ... the documentation of the class Module.Class ...
        </doc>
      </class>
    </module>
  </module>

If all the children of a <docgroup> are of the same type, e.g. all are
<method> elements, then the <docgroup> has the attribute homogen-type
(="method" in the example). If all the children have identical name="..."
attributes, then the <docgroup> gets a homogen-name="..." attribute aswell.

The <docgroup> has a <doc> child containing the docmentation for the other
children of the <docgroup>. An entity that cannot be grouped (class, module,
enum), has a <doc> child of its own instead.

======================================================================
d) Pike entities
----------------------------------------------------------------------

Pike entities - classes, modules, methods, variables, constants, &c, have some
things in common, and many parts of the xml format are the same for all of
these entities. All entities are represented with an XML element, namely one
of:

  <class>
  <constant>
  <enum>
  <inherit>
  <method>
  <modifier>
  <module>
  <typedef>
  <variable>

The names speak for themselves, except: <modifier> which is used for modifier
ranges:

  //! Some variables:
  protected final {
    int x, y;

    string n;
  }

A Pike entity may also have the following properties:

  Name - Given as a name="..." attribute:
    <variable name="i"> ... </variable>

  Modifiers - Given as a child element <modifiers>:
    <variable name="i">
      <modifiers>
        <optional/><protected/><private/>
      </modifiers>
      ...
    </variable>
  If there are no modifiers before the declaration of the entity, the
  <modifiers> element can be omitted.

  Source position - Given as a child element <source-position>:
    <variable name="i">
      <source-position file="/home/rolf/hejhopp.pike" first-line="12"/>
      <modifiers>
        <optional/><protected/><private/>
      </modifiers>
      ...
    </variable>
  The source position is the place in the code tree where the entity is
  declared or defined. For a method, the attribute last-line="..." can be
  added to <source-position> to give the range of lines that the method
  body spans in the source code.

And then there are some things that are specific to each of the types of
entities:

<class>
   All inherits of the class are given as child elements <inherit>. If there
   is doc for the inherits, the <inherit> is repeated inside the appropriate
   <docgroup>:

     class Bosse {
       inherit "arne.pike" : Arne;
       inherit Benny;

       //! Documented inherit
       inherit Sven;
     }

     <class name="Bosse">
       <inherit name="Arne"><source-position ... />
                            <classname>"arne.pike"</classname></inherit>
       <inherit><source-position ... />
                <classname>Benny</classname></inherit>
       <inherit><source-position ... />
                <classname>Sven</classname></inherit>
       <docgroup homogen-type="inherit">
         <doc>
           <text><p>Documented inherit</p></text>
         </doc>
         <inherit><source-position ... />
                  <classname>Sven</classname></inherit>
       </docgroup>
       ...
     </class>

<constant>
   Only has a name. The element is empty (or has a <source-position> child.)

<enum>
   Works as a container. Has a <doc> child element with the documentation of
   the enum itself, and <docgroup> elements with a <constant> for each enum
   constant. So:

     enum E
     //! enum E
     {
       //! Three constants:
       a, b, c,

       //! One more:
       d
     }

   becomes:

     <enum name="E">
         <doc><text><p>enum E</p></text></doc>
         <docgroup homogen-type="constant">
             <doc><text><p>Three constants:</p></text></doc>
             <constant name="a"/>
             <constant name="b"/>
             <constant name="c"/>
         </docgroup>
         <docgroup homogen-name="d" homogen-type="constant">
             <doc><text><p>One more:</p></text></doc>
             <constant name="d"/>
         </docgroup>
     </enum>

   Both the <enum> element and the <constant> elements could have
   <source-position> children, of course.

<inherit>
   The name="..." attribute gives the name after the colon, if any. The name
   of the inherited class is given in a <classname> child. If a file name is
   used, the class name is the file name surrounded by quotes (see <class>).

<method>
   The arguments are given inside an <arguments> child. Each argument is
   given as an <argument name="..."> element. Each <argument> has a <type>
   child, with the type of the argument. The return type of the method is
   given inside a <returntype> container:

     int a(int x, int y);

     <method name="a">
       <arguments>
         <argument name="x"><type><int/></type></argument>
         <argument name="y"><type><int/></type></argument>
       </arguments>
       <returntype><int/></returntype>
     </method>

<modifier>
   Works as a container ... ???

<module>
   Works just like <class>.

<typedef>
   The type is given in a <type> child:

     typedef float Boat;

     <typedef name="Boat"><type><float/></type></typedef>

<variable>
   The type of the variable is given in a <type> child:

     int x;

     <variable name="x"><type><int/></type></variable>

======================================================================
e) Pike types
----------------------------------------------------------------------

Above we have seen the types int and float represented as <int/> and <float/>.
Some of the types are complex, some are simple. The simpler types are just on
the form <foo/>:

  <float/>
  <mixed/>
  <program/>
  <string/>
  <void/>

The same goes for mapping, array, function, object, multiset, &c that have
no narrowing type qualification: <mapping/>, <array/>, <function/> ...

The complex types are represented as follows:

array
   If the type of the elements of the array is specified it is given in a
   <valuetype> child element:

     array(int)

     <array><valuetype><int/></valuetype></array>

function
   The types of the arguments and the return type are given (the order
   of the <argtype> elements is significant, of course):

     function(int, string: mixed)

     <function>
       <argtype><int/></argtype>
       <argtype><string/></argtype>
       <returntype><mixed/></returntype>
     </function>

int
   An int type can have a min and/or max value. The values can be numbers or
   identifiers:

     int(0..MAX)

     <int><min>0</min><max>MAX</max></int>

mapping
   The types of the indices and values are given:

     mapping(int:int)

     <mapping>
       <indextype><int/></indextype>
       <valuetype><int/></valuetype>

multiset
   The type of the indices is given:

     multiset(string)

     <multiset>
       <indextype><string/></indextype>
     </multiset>

object
   If the program/class is specified, it is given as the text child of
   the <object> element:

     object(Foo.Bar.Ippa)

     <object>Foo.Bar.Ippa</object>

Then there are two special type constructions. A disjunct type is written
with the <or> element:

  string|int

  <or><string/><int/></or>

An argument to a method can be of the varargs type:

  function(string, mixed ... : void)

  <function>
    <argtype><string/></argtype>
    <argtype><varargs><mixed/></varargs></argtype>
    <returntype><void/></returntype>
  </function>

======================================================================
f) Other XML tags
----------------------------------------------------------------------

p
   Paragraph.

i
   Italic.

b
   Bold.

tt
   Terminal Type.

pre
   Preformatted text.

code
   Program code.

image
   An image object. Contains the original file path to the image. Has the
   optional attributes width, height and file, where file is the path to
   the normalized-filename file.

======================================================================
g) XML generated from the doc markup
----------------------------------------------------------------------

The documentation for an entity is put in a <doc> element. The <doc> element
is either a child of the element representing the entity (in the case of
<class>, <module>, <enum>, or <modifiers>) or a child of the <docgroup> that
contains the element representing the entity.

The doc markup has two main types of keywords. Those that create a container
and those that create a new subsection within a container, implicitly closing
the previous subsection. Consider e.g.:

  //! @mapping
  //!   @member int "ip"
  //!     The IP# of the host.
  //!   @member string "address"
  //!     The name of the host.
  //!   @member float "latitude"
  //!   @member float "longitude"
  //!     The coordinates of its physical location.
  //! @endmapping

Here @mapping and @endmapping create a container, and each @member start a
new subsection. The two latter @member are grouped together and thus they
form ONE new subsection together. Each subsection is a <group>, and the
<group> has one or more <member> children, and a <text> child that contains
the text that describes the <member>s:

  <mapping>
      <group>
          <member><type><int/></type><index>"ip"</index></member>
          <text>
            <p>The IP# of the host.</p>
          </text>
      </group>
      <group>
          <member><type><string/></type><index>"address"</index></member>
          <text>
              <p>The name of the host.</p>
          </text>
      </group>
      <group>
          <member><type><float/></type><index>"latitude"</index></member>
          <member><type><float/></type><index>"longitude"</index></member>
          <text>
              <p>The coordinates of its physical location.</p>
          </text>
      </group>
  </mapping>

Inside a <text> element, there can not only be text, but also a nested level
of, say @mapping - @endmapping. In that case, the <mapping> element is put in
the document order place as a sibling of the <p> that contain the text:

  //! @mapping
  //!   @member mapping "nested-mapping"
  //!     A mapping inside the mapping:
  //!     @mapping
  //!       @member string "zip-code"
  //!         The zip code.
  //!     @endmapping
  //!     And some more text ...
  //! @endmapping

  becomes:

  <mapping>
    <group>
      <member><type><mapping/></type><index>"nested-mapping"</index></member>
        <text>
          <p>A mapping inside the mapping:</p>
          <mapping>
            <group>
              <member><type><string/></type><index>"zip-code"</index></member>
              <text>
                <p>The zip code.</p>
              </text>
            </group>
          </mapping>
          <p>And some more text ...</p>
        </text>
      </group>
  </mapping>

Inside the <p> elements, there may also be some more "layout-ish" tags like
<b>, <code>, <tt>, <i>, needed to make the text more readable. Those tags are
expressed as @i{ ... @} in the doc markup. However there are no <br>. A
paragraph break is done by ending the <p> and beginning a new. A </p><p> is
inserted for each sequence of blank lines in the doc markup:

  //! First paragraph.
  //!
  //! Second paragraph.
  //!
  //!

  becomes:

  <p>First paragraph.</p><p>Second paragraph.</p>

Note that the text is trimmed from leading and ending whitespaces, and there
are never any empty <p> elements.

In the example above the keyword `@mapping' translated into <mapping>, whereas
the keyword `@member string "zip-code"' translated into:
  <member><type><string/></type><index>"zip-code"</index></member>

The translation of keyword->XML is done differently for each keyword. How it
is done can be seen in lib/modules/Tools.pmod/AutoDoc.pmod/DocParser.pmod. Most
keywords just interpret the arguments as a space-separated list, and put their
values in attributes to the element. In some cases (such as @member) though,
some more intricate parsing must be done, and the arguments may be complex
(like Pike types) and are represented as child elements of the element.

======================================================================
h) Top level sections of different Pike entities.
----------------------------------------------------------------------

In every doc comment there is an implicit "top container", and subsections can
be opened in it. E.g.:

  //! A method.
  //! @param x
  //!   The horizontal coordinate.
  //! @param y
  //!   The vertical coordinate.
  //! @returns
  //!   Nothing :)
  void foo(int x, int y)

becomes:

  <docgroup homogen-name="foo" homogen-type="method">
      <doc>
          <text><p>A method.</p></text>
          <group>
              <param name="x"/>
              <text><p>The horizontal coordinate.</p></text>
          </group>
          <group>
              <param name="y"/>
              <text><p>The vertical coordinate.</p></text>
          </group>
          <group>
              <returns/>
              <text><p>Nothing :)</p></text>
          </group>
      </doc>
      <method name="foo">
         ......
      </method>
  </docgroup>

Which "top container" subsections are allowed depends on what type of entity is
documented:

ALL      -  <bugs/>
            <deprecated> ... </deprecated>
            <example/>
            <note/>
            <seealso/>

<method> -  <param name="..."/>
            <returns/>
            <throws/>