Re: C++ TA parser

["Susan Elliott SIM" <SESIM@uci.edu>]

Dear All,

Yuzo Kanomata wrote:

> Hi All,
>
> I ran the *old* (vintage 2001) TA C++ parse and it works fine. I ran it 
> on the Apache httpd core.c file and got a GXL file with 23389 nodes and 
> 23389 edges.
>
> Things to report:
>
> 1. running BindingTest on it, the java interpreter ran out of memory 
> after grabbing all the nodes and about 21000 edges.

I'm unfortunately not surprised by this.

> 2. The C++ parse took a long time to run over core.c, I estimate between 
> 12-15 minutes! The majority of this is I believe related to the database 
> used to store information. I am not about to change the implementation, 
> but the sources are available, and perhaps a mySQL implementation would 
> be faster -- just a guess.

Is this the TkSee parser? More than one person has wondered about this.

> 3. The nature of the information the parser generates is very similar to 
> work done in Hypertext communities specific technologies that did 
> similar things are:
> Chimera -- UCI product now at CU Boulder, stored C function signatures 
> with position information for use with either vi or emacs and related 
> information stores in design documents and specifications. "Fly through 
> your (B-2 Spirit bomber) code with Chimera" DARPA demo comes to mind.
> Xanadu -- Ted Nelson's PARC/Auto Cad/now opensource project. Ted had a 
> local file system-centric relational idea at first, but a lot of 
> developers used parts of Xanadu to make psuedo-IDE's in command line 
> environments.
> Rivendell -- Gail Kaiser's hyperbase project at Columbia. I recall the 
> Steve Dossick (now at Microsoft working on the rule-base in Longhorn) 
> had a Quake front-end so that you could "run" through the call graphs 
> and the other hyperbase relations.

These projects are an awful lot like other work that I know of.
More examples of how different sub-groups have tried to solve the
same problem and we're sitting in the middle of this.

Susan