shared wasabi implementation

[Max Wiehle]

Hi,

Am Montag, den 19.02.2007, 23:14 +0100 schrieb Mikkel Kamstrup
Erlandsen:
> 2007/2/19, Max Wiehle <max.wiehle at gmail.com>:
>         Hi,
>         
>         Hope it's okay if i just jump right into the discussion. I
>         just read
>         about wasabi on a dashboard-hackers and i am quite interested
>         in it so i
>         tried to catch up with the mailing list archive. I worked on a
>         metadata 
>         store for beagle during last years Summer of Code and i am
>         still trying
>         to follow what happens on the desktop search / metadata side
>         of things
>         as time permits.
>         
>         Am Sonntag, den 18.02.2007, 21:15 +0100 schrieb Mikkel
>         Kamstrup 
>         Erlandsen:
>         
>         >
>         > Ok. If we are to standardize something like this, I would
>         assume that
>         > we use dbus for rpc - as far as I can tell that doesn't seem
>         to be a
>         > problem..? Fx a dbus api like: 
>         >
>         >  - AddFile (in as metadata, in s input_file)
>         >  - AddText (in as metadata, in s text)
>         >
>         > where the metadata argument contains things such as uri,
>         mime, and hit
>         > type (in some specified order (and maybe some 
>         > filtering/stemming/whatnot info)). The AddFile method sorta
>         replaces
>         > the "drop-in-special-dir" approach - the drop-in-special-dir
>         method
>         > could still be allowed for apps not talking dbus. The
>         AddText method 
>         > should encapsulate the functionality of Beagles' current
>         > IndexServiceRequest/Indexable duo.
>         
>         Maybe it would be possible to have a
>           - AddMetadata (in as metadata, in s uri)
>         as well. This could be used for all kinds of metadata that is
>         added to a 
>         object aside from indexing. fx tags, emblems, notes in
>         nautilus. Or
>         epiphany might add "downloaded_from" to a files metadata -
>         saved
>         attachements could be marked as belonging to a certain email
>         and vice 
>         versa etc.
> 
> 
> Agreed. I figure this belongs under the Wasabi metadata spec and api -
> which is yet to be discussed. Maybe I better kick that off soonish -
> the search spec relies on some bits and pieces from it anyway. 
> 
> The two methods I mention is specifically only targeting an indexer
> since I don't think it should be a requirement to both an indexer and
> metadata storage.
Okay, i did not know there was a seperate metadata spec and api. Makes
sense.

> Is there any where we can have a look at your work? 
Its in the beagle svn in
http://svn.gnome.org/viewcvs/beagle/branches/beagle-metadata-branch/
mostly in beagled/SqliteMetadata/...

It's only the store and some internal beagle changes that save metadata
to the store. But there is no external api yet.

EntityStore.cs is the central store.
SqlMetaTests/TestEntityStoreBasics.cs illustrates the usage a little
bit.
SqlMetadata.cs implements an EntityStore that tries to verify metadata
against a model supplied in SqlMetaModel and stores it marked as
non_verified anyway if it can't be verified.

If there are any questions please ask. I am quite busy right now but i
will be working on similar things from april on for my diploma thesis so
i will try and keep in touch.

 Max