simple search api (was Re: mimetype standardisation by testsets)

[Fabrice Colin]

On 11/24/06, Mikkel Kamstrup Erlandsen <mikkel.kamstrup at gmail.com> wrote:
> > The Wiki is clear enough. It's just that it may be useful to provide
> > the consumer
> > application with a list of supported groups... unless we dictate which
> > groups should
> > be supported by all engines.
>
> Well, my idea was to come up with a spec at some point in the mid-term
> future (after the simple api is complete). On top of that the spec should of
> course be extensible so that backends can have custom groups - which implies
> the need for a method to do runtime introspection. I just don't think the
> belongs in a simple search api.
>
That's fair enough. So should we dictate values for the "group" switch ?

> > Hmmm it depends on. As the dbus specification says, you can have "array of
> > array of array of ... struct of struct of struct of ...", which is
> > probably flexible enough
> > to pass all the data held by a Query object.
>
> Yeah, we can hold the data of a query object, but you can't assign any
> methods to it (unless you export it over dbus (which I don't think is a good
> idea)). Methods to construct a query programmatically would have to reside
> in a toolkit-dependent lib I think.
>
That's right. Nobody is suggesting full-blown objects. Data is good enough
if the different fields are documented.

Fabrice