[josm-dev] Optimizations for larger data sets

[Frederik Ramm]

Hi,

> I don't think this would require much more memory.  Say we'd add a tile index
> for the nodes, this would require just a bit more than one pointer per node,
> that's just 8M for a million nodes.  Similarly, if we cache ways as polylines,
> we wouldn't need to duplicate all the indirections, we could just store them
> as arrays of doubles, so for a million nodes that might cost us just 18M.


Hm, I hadn't thought of actually keeping the node display cache info in 
the way dislpay objects but that might not be too bad. Still, each 
object has some overhead and we also need pointers between the display 
objects and the cache objects so that the latter may be invalidated when 
the former change...

Bye
Frederik