Re: using gxref on a single module

On 02/04/2014 08:38 PM, Abdallah, Samer wrote:
> Hello all,
> I'm trying to use gxref to understand the structure of a single module
> (actually, the GULP4 feature structure translator I mentioned in an
> earlier post about feature structure unification) and the information it 
> is displaying doesn't match what I would expect: 
> 
> Two predicates that
> are defined and called internally withing the module are showing up
> as "Not Called" in the "File Info" window of the gxref GUI. Is there some
> reason why that would be the case? They are called normally, not via
> a meta-predicate.

Dunno.

> In addition, there are two predicates which *are* only
> called via meta-predicates (one of which is bagof/3) which also show
> up as "Not Called". Since the system does a meta-predicate analysis
> on load now, isn't the information necessary to detect such 'static'
> calls via meta-predicates available now?

It should ...

> Finally, it would seem to be useful to get a dependency graph between
> predicates in a module - is there any tool which can do this?
> In the meantime, I am drawing the graph with pen and paper...

gxref us based on library(prolog_xref), which performs analysis based
on the source file.  This is always a bit limited as one can do pretty
wild things in SWI-Prolog while loading code.  The alternative is
library(prolog_codewalk), which analyses the _loaded_ code, and this
is not subject to hard to track code rewriting.  The documentation
[1] contains an example that generates a call-graph.

Next, there is the pack graphml, which writes GraphML files.  Then
you have tools like Gephi to draw wonderful graphs :-)  If you go that
way, please send the picture :-)

	Cheers --- Jan

[1] http://www.swi-prolog.org/pldoc/doc/swi/library/prolog_codewalk.pl