Re: [Banshee-List] Some preliminary work in trunk - opinions needed

[LCID Fire <lcid-fire gmx net>]

Gabriel Burt wrote:
>> concern is that it appears the "ModelID" value being used to identify each
>> model in the cache table is dependant upon the order in which models are
>> instantiated (line 61 of BansheeCacheableModelAdapter.cs). Am I mistaken, or
>> would changing the order of the model/cache-adapter instantiation screw up
>> the cache on subsequent launches? If so, perhaps something a /little/ more
>> sturdy should serve as the ModelID, like the hashcode of the model's class
>> name. Or something.
> 
> The cache, as it is now, is not intended to live beyond a single
> Banshee session.  It is actually cleared on start up.   Each cacheable
> model is given a unique ModelID which could be different each run, but
> that's ok.  It will be simple enough to change it if we need to
> persist the cache.
What I wonder is - why aren't the tables normalized? I think you'd have
much less work in generating the "cache" - let alone the fact that you'd
have to do this once and then it would be done.

The reason why I'm asking is - I'm currently rewriting the database part
in c++ so I can access banshee data from another app. Initially I
thought about opening the database in readonly mode. Since almost every
operation is done on the cache tables I either implement my own ways to
access data or I, too, have to generate the cache tables (I actually
have the code running - but it really hurts).
For readonly access I think I'd go with the former - but hereby I
officially complain about the data structure ;)

ym2c