dolfin team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

Johan Jansson wrote:
> Hi,
>
> I've added dependencies for GLib and GTS (http://gts.sourceforge.net),
> GTS requires GLib. I've made the dependency conditional though, so
> it's not obligatory to have GTS (GLib could also be made
> conditional). I'm not really fond of conditional dependencies though,
> it means the code becomes less readable (requires #ifdefs) and some
> functionality will obviosuly be disabled. On the other hand, having
> too many required dependencies makes DOLFIN more tedious to
> build. What do you think?

If the use of GTS is localised, I don't mind seeing a few #ifdefs. GLib 
could be a pain as it's a big package with numerous dependencies. Does 
GLib support all platforms?

What functionality can we expect in the future that requires GTS? If 
it's something generic and neat, we might as well have a dependency on 
GTS (as long as GTS is maintained and easy to build on various platforms).

> GTS provides bounding-box trees, which is likely the only feature we
> will use. The geometric predicates also come from GTS, but since
> they're not part of the public API I had to duplicate them in DOLFIN
> (GTS has wrappers for its own mesh representation).

Unfortunate.

> The bounding-box trees are used by IntersectionDetector, which
> provides general cell-location (and overlap) functionality. This is in
> preparation for projection between non-matching meshes, 

Great! I'm looking forward to this.

Garth

but could also
> have other uses.
>
>   Johan
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev