Re: [cdt-core-dev] Open C++ Parser

[Tom Tromey <tromey@xxxxxxxxxx>]

>>>>> "Doug" == Schaefer, Doug <dschaefer@xxxxxxxxxxxx> writes:

Doug> We could set up out-of-process links but I have doubts that
Doug> would meet performance requirements, especially for the
Doug> real-time users of the parser (e.g. code complete).

Yeah, the parser is run as a separate process, but the results are
kept in a database.  My recollection (I hacked on S-N in the dim and
distant past) was that this really wasn't a performance limiter.

However, to be sure I've sent some email to a couple of S-N folks with
a bunch of questions.  Perhaps we can get a more definitive answer on
the scalability, performance, and correctness issues.  I'll report
back when I get an answer.


This brings up scalability problems.  With the earlier IBM CDT, we
found we couldn't parse even medium sized projects like gcc.  They
simply required too much memory.  How/where are you planning to store
the parse information?  This issue is partly solved by the S-N
approach -- it traded disk for core.  As I recall the on-disk
databases could grow quite large.

If you're curious to see S-N in action you can grab it from
sourceforge and try it out...


I don't know if this is interesting or not, but S-N includes a lot of
parsers, not just C and C++.  It would perhaps be possible to get
useful information for other languages "for free".

Tom