This is the AGSParser class... and some autogsdoc examples.
The AGSParser class is designed to produce a property-list
which can be handled by AGSOutput... one class is not much
use without the other.
Copyright: (C) 2001 Free Software Foundation, Inc.
Returns the methods, functions and C data types in
their header declaration order, by organizing them
into arrays as described below. Methods are grouped by
class, category or protocol references. For example,
valid keys could be ClassName,
ClassName(CategoryName) and
(ProtocolName). Functions and C data types
are grouped by header file names. For example,
AGParser.h would a valid key. TODO: Collect
functions and C data types. Only methods are
currently included in the returned dictionary.
Return the list of known output files depending on
this source/header. If there are any classes,
categories, or protocols, there will be an
output file for them whose name is based on the name
of the header. If there are any constants,
variables, typedefs or functions, there will
either be a shared output file for them (defined by
a template name set in the user defaults system), or they
will go in the same file as classes etc.
In spite of its trivial name, this is one of the key
methods - it parses and skips past comments, but it
also recognizes special comments (with an additional
asterisk after the start of the block comment) and
extracts their contents, accumulating them into
the 'comment' instance variable. When the data
provided by a comment is appended to the data
stored in the 'comment' instance variable, a line
break (<br />)is automatically forced to
separate it from the proceding info. In
addition, the first extracted documentation is
checked for the prsence of file header markup,
which is extracted into the 'info' dictionary.
There are various sections we can extract from the
document - at most one of each.
We handle struct, union, and enum declarations by
skipping the stuff enclosed in curly braces. If
there was an identifier after the keyword we use it
as the struct name, otherwise we use '...' to denote a
nameless type.
If this is parsing a header file (isSource ==
NO) then we reset the list of known
source files associated with the header before
proceeding. We initially assume that the
location of a source file is the same as the
header, but if there is no file at that location,
we expect the source to be in the documentatation
directory or the current directory instead.
Attempt to parse an identifier/keyword (with
optional whitespace in front of it). Perform
mappings using the wordMap dictionary. If a
mapping produces an empty string, we treat it as if
we had read whitespace and try again. If we read end of
data, or anything which is invalid inside an
identifier, we return nil. If we
read a GS_GENERIC... macro, we return its first
argument.
Parse a macro definition... we are expected to have
read #define already It's common to have macros
which don't need commenting... like the ones used to
protect a header against multiple inclusion for
instance. For this reason, we ignore any macro
which is not preceded by a documentation comment.
Parse a preprocessor statement, handling preprocessor
conditionals in a rudimentary way. We keep
track of the level of conditional nesting, and we
also track the use of #ifdef and #ifndef with some
well-known constants to tell us which standards
are currently supported.
Skip past any whitespace characters (as defined by the
supplied set) including comments. Calls
parseComment if neccesary, ensuring that any
documentation in comments is appended to our
'comment' ivar.
Set the name of the file in which classes
are to be documented as being declared. The default
value of this is the last part of the path of the
source file being parsed.
This method is used to enable (or disable)
documentation of all instance variables. If
it is turned off, only those instance variables that are
explicitly declared 'public' or 'protected' will
be documented.
Turn on or off parsing of preprocessor conditional
compilation info indicating the standards
complied with. When this is turned on, we assume
that all standards are complied with by default.
You should only turn this on while parsing
the GNUstep source code.
Store the current standards information derived from
preprocessor conditionals in the supplied
dictionary... this will be used by the AGSOutput
class to put standards markup in the gsdoc output.
Sets up a dictionary used for mapping
identifiers/keywords to other words.
This is used to help cope with cases where C
preprocessor definitions are confusing the
parsing process.
Read in the file to be parsed and store it in a
temporary unicode buffer. Perform basic
transformations on the buffer to simplify
the parsing process later - including stripping out of
escaped end-of-line sequences. Create mapping
information to convert positions in the new
character buffer to line numbers in the original
data (for logging purposes).
Skip until we encounter a semicolon or closing brace.
Strictly speaking, we don't skip all statements
that way, since we only skip part of an if...else
statement.
Special method to skip a statement and up to the
end of the last line it was on, discarding any comments
so they don't get used by the next construct that
actually needs documenting.