Re: Function completion for GVariant maybe types?

[Simon McVittie <simon mcvittie collabora co uk>]

On Mon, 05 Dec 2011 at 19:45:39 +0100, Markus Elfring wrote:
> >   type = g_variant_get_type (gv);
> 
> How do you think about the following instructions?
> 
> 	GVariant* contained = g_variant_get_child_value (value, 0);
> 	type = g_variant_get_type (contained);

What is the type of "value"? If it's a "maybe" type like "m(xy)", then
this will only work correctly if "contained" is non-NULL. It'll produce
undefined behaviour (i.e. probably crash) if "contained" is NULL

If you want a function that, given a GVariant of type "m(xy)",
will return "(xy)", then I think Ryan already provided one, but
here is another:

    const GVariantType *vt = g_variant_get_type (value);

    /* you might want to assert that g_variant_type_is_maybe (vt) here */

    return g_variant_type_element (vt);

> Is it possible to determine the child data type without retrieving
> the complete value?

See above.

> Would it be useful to optimise this approach?

I doubt it. The usual answer to "should we optimise $THING?" applies:
if you don't have profiling data that indicates that $THING is a bottleneck,
then no, optimising $THING isn't useful.

    S