Re: [Multi-valued V2 (grilo) 11/12] core: Merge GrlDataMulti into GrlData

["Juan A." Suárez Romero <jasuarez igalia com>]

On Thu, 2011-03-03 at 09:15 +0000, Iago Toral wrote:
> > We can discuss if we can get rid of some of this functions, but that
> > would change the API, so I prefer to keep it as it is, at least for 
> > some
> > time, after we think which functions can be removed, and what 
> > functions
> > should be renamed.
> 
>  No, it won't change the API.
>  What I am saying is that we should have grl_data_remove *only* but
> its 
>  implementation should be that of grl_data_remove_property and we
> should 
>  not have grl_data_remove_property, of course.


Currently we have:

void grl_data_remove (GrlData *data, GrlKeyID key);

void grl_data_remove_related_keys (GrlData *data, GrlKeyID key, guint
index);

The first one is invoking second one with index == 0.

If we want to keep the first one only, we would need to add the index
parameter, and hence changing the api.

> >
> >> > But users can still remove a specific element if they really
> want:
> >> > get
> >> > the property with the element, and then remove the key from that
> >> > property.
> >>
> >>  Shouldn't we remove that API from GrlProperty then?
> >
> > The API to remove the key in GrlProperty is there to implement the
> > grl_data_remove(), which acts on single-values (i.e., the first
> value 
> > of
> > multi-valued keys).
> 
>  If it is there for internal use as you suggest here, then that's
> fine 
>  with me, but then it should not be part of the public API, since as
> I 
>  said, its use by clients could break the semantics of the related
> keys, 
>  and I don't really see a good reason why they would like to use this 
>  instead of grl_data_remove (updated to deal with related keys)
> 
> > Of course, I can still get rid of it, as doing
> >
> > grl_property_remove(prop, key)
> >
> > is almost the same as
> >
> > grl_property_set (prop, key, NULL)
> 
>  But this is not exactly the same, as grl_property_set would actually 
>  leave the key with NULL value.

Fair enough. I'll remove the function.


	J.A.