This is a proposal for an architecture change.
Currently ROBODoc consists of roughly two modules:
An analyser and a generator. In an ideal situation
these would do the following:
The analyser examines the source code and extracts
the robodoc headers parses the headers and stores
the information in a list.
The generator iterates through this list and depending
on the output mode creates the documentation in
one of the supported formats (html, latex etc).
However currently, the generator does part of
the header parsing. And it does this differently
for each output mode. This makes maintenance of
ROBODoc difficult.
The proposal is to move these parsing routines back
into the analyser.
(Thanks to Gergely Budai for pointing this out).
Logged In: YES
user_id=572001
Originator: NO
I guess that there are more design options to consider ...
http://www.cs.berkeley.edu/~smcpeak/elkhound/sources/elsa/doc/design.html
http://g4re.sourceforge.net/index.shtml
http://keystone.sourceforge.net/research.shtml
Logged In: YES
user_id=1511265
Originator: NO
Especially the source highlighing is badly implemented... At the moment, it takes place solely in the html generator. It should be done in the analyser phase. Also the part in the html generator is pretty inscrutable.