The future of gideon

[Daniel Engelschalt]

[...]
> -- Code Completion --
[...]
just a few words from my side - i see 2 main problems concerning the
codecompletion:

1st is the parsing of (just written) code. currently we have 2 approaches
for this issue - roberto is reducing the code using regexps and i'm using lex,
looking for unknown expressions. IMHO the "right" way would be using a
lex-oriented solution which has to be newly created for kdevelop. i don't say this
because i did some work on that, i say it because i believe it has the most
power needed for this task. i havn't had enough time for taking a closer look
at what roberto did, but i have a strong feeling that it "could" be too time
consuming compared with lex on larger projects.
an optimal solution would be an editor which uses incremental parsing -
maybe kate can be extended with a plugin for that.

2nd is the storing of all the information we collect. working with
pre-parsed qt / kde - headers is not that slow as i expected, but completing a class
with lots of parents takes its time. again, i don't know it exactly but is the
currently used approach the fastest solution (for dealing with big projects;
maybe with hundreds of own classes ?) or should a real database be
considered ?

currently i have very limited time to work on gideon; i'm not a coding guru
and havn't read more than one book about compiler building. so please excuse
my not always correct statements :) but i encourage all to take a look at
what is there (in gideon) and if it is worth to extend / rewrite it. my opinion
always has been not to create a perfect parser but one that works for the
main tasks.

bye,
daniel

-- 
GMX - Die Kommunikationsplattform im Internet.
http://www.gmx.net