Re: Adding full introspection information (#139486)

[Andy Wingo <wingo pobox com>]

Hi Maciej,

A couple of thoughts:

First, consider language-bindings gnome org for this discussion.

The idea of a C library for introspecting methods is a nice one, I
think, but not for the reasons you are giving. It would be nice iff you
could get the language binding maintainers on board, because it would be
a neutral ground for collaboration. We don't even collaborate on .defs
at this point, presumably because everyone has their own extensions to
the format. As an example, guile-gnome and pygtk at least have default
values, and (null-ok). Also, guile-gnome has GList*-of-<foo>, so that
you can specify the subtype of a list. Defs are a low-tech solution to
the problem, and low-tech has its merits.

Aside from that, a bindings generator for a language will likely be
written in that language (e.g. guile's is written in guile), so there's
a lower commonality. But if that information is available from a C
library, cooperation can be pushed upstream a bit, where people can hack
in a "neutral" language.

Finally, I think you'll find a bit of resistance to this idea from
bindings authors -- the fellow from the ruby bindings posted recently
about how he overloads << and >> in gstreamer bindings, guile-gnome uses
srfi-19 date objects instead of icaltimetype in the
evolution-data-server bindings, pygtk represents tree paths as tuples of
integers, etc. These things are why we use language bindings, and would
be impossible to generate.

I don't know much about KDE, but from what I've seen, their `smoke'
project might be a good model, because of its simplicity.

Regards,
-- 
Andy Wingo <wingo pobox com>
http://ambient.2y.net/wingo/