[Loadstone] Point type/class
shawn at odyssey.cm.nu
Wed Apr 11 09:30:31 BST 2007
The system you suggest would probably work better on a more powerful system.
Unfortunately with the phone we have to account for storage, memory, and
processing power. We also have to deal with the phone's database engine that
lacks several features that would be nice to have.
I think the way this will probably work is the types won't be a table in the
database but an array stored in the program. This will allow the types to be
translated in to other languages and won't require duplicating type data for
every database created. The database would then use ids to reference the
types. Idealy I'd like no more than 15 main types with no more than 15
subtypes each. This would allow the type and sutbype ids to be stored in a
single byte so that that data would only be one extra byte per point. I'm
not sure if this will happen or not thoe. I don't want a table that joins
types to points since using it to filter searches would be almost
impossible. This is because the database engine doesn't support table joins.
The way around that would of course be to replace the database engine. We
need one that's free and open source that can run on series60 phones. No
luck so far in finding that.
More information about the Loadstone