@shorttitle(Overview)
@title(Pasdoc Sources Overview)

This is the documentation of the pasdoc sources, intended for pasdoc developers.
For user's documentation see [http://pasdoc.sourceforge.net/].

Contents:
@tableOfContents

General overview of the data flow in pasdoc:

@section(1 SecParsing Parsing)

@link(TTokenizer) reads the source file, and converts it to a series
of @link(TToken)s.

@link(TScanner) uses an underlying @link(TTokenizer) and
also returns a series of @link(TToken)s,
but in addition it understands and interprets $define,
$ifdef and similar compiler directives. While @link(TTokenizer)
simply returns all tokens, @link(TScanner) returns only those tokens
that are not "$ifdefed out". E.g. if WIN32 is not defined then
the @link(TScanner) returns only tokens "@code(const LineEnding = #10;)"
for the following code:
@longcode# const LineEnding = {$ifdef WIN32} #13#10 {$else} #10 {$endif}; #

Finally @link(TParser) uses an underlying @link(TScanner) and interprets
the series of tokens, as e.g. "here I see a declaration of variable Foo,
of type Integer". The Parser stores everything it reads in a @link(TPasUnit)
instance.

If you ever wrote a program that interprets a text language,
you will see that there is nothing special here: We have a lexer
(@link(TScanner), a simplified lexer in @link(TTokenizer))
and a parser (@link(TParser)).

It is important to note that pasdoc's parser is somewhat unusual,
compared to "normal" parsers that are used e.g. in Pascal compilers.

@orderedList(
  @item(
    Pasdoc's parser does not read the implementation section of a unit file
    (although this may change some day, see
    [https://github.com/pasdoc/pasdoc/wiki/WantedFeaturesParsingImplementation]).)

  @item(
    Pasdoc's parser is "cheating": It does not really understand everything it
    reads. E.g. the parameter section of a procedure declaration is parsed
    "blindly", by simply reading tokens up to a matching closing
    parenthesis. Such cheating obviously simplifies the parser
    implementation, but it also makes pasdoc's parser "dumber",
    see [https://github.com/pasdoc/pasdoc/wiki/ToDoParser].)

  @item(
    Pasdoc's parser collects the comments before each declaration,
    since these comments must be converted and placed in the final documentation
    (while "normal" parsers usually treat comments as a meaningless
    white-space).)
)

@section(1 SecStoring Storing)

The unit @link(PasDoc_Items) provides a comfortable class hierarchy
to store a parsed Pascal source tree. @link(TPasUnit) is a "root class"
(container-wise), it contains references to all other items within a unit,
every item is some instance of @link(TPasItem).

@section(1 SecGenerators Generators)

The last link in the chain are the generators. A generator uses the
stored @link(TPasItem)
tree and generates the final documentation. The base abstract class
for a generator is @link(TDocGenerator), this provides some
general mechanisms used by all generators. From @link(TDocGenerator)
descend more specialized generator classes, like
@link(TGenericHTMLDocGenerator), @link(THTMLDocGenerator),
@link(TTexDocGenerator) and others.

@section(1 SecNotes Notes)

Note that the parser and the generators do not communicate with each
other directly. The parser stores things in the @link(TPasItem) tree.
Generators read and process the @link(TPasItem) tree.

So the parser cannot do any stupid thing like messing with
some HTML-specific or LaTeX-specific issues of generating documentation.
And the generator cannot deal with parsing Pascal source code.

Actually, this makes the implementation of the generator independent
enough to be used in other cases, e.g. to generate
an "introduction" file for the final documentation, like the one
you are reading right now.

