If we share a tile provider amongst multiple views, the tile cache can be under-sized

[GoogleCodeExporter]

(this is targeting the new tile request chain branch, but the underlying issue 
probably still exists in the current trunk)

We have designed into the OpenStreetMapView and the tile provider the 
functionality to share a tile provider amongst more than one view. This is used 
in the mini-map sample, where the main map and the mini-map share the same tile 
provider. The problem is that both maps call ensureCapacity with their own 
capacity requirements, but really those requests need to be aggregated over all 
views that use the tile provider. So if the main map needs 16 tiles, and the 
mini-map needs 4 tiles, the capacity truly needs to be set to 20.

This might be accomplished in a few ways. We could somehow try to aggregate the 
ensureCapacity requirements. We could somehow assign individual tile caches to 
each view. We could prevent users from sharing top-level tile providers (like 
the array provider) and encourage them to share the modular tile providers 
(like the download tile provider).

Original issue reported on code.google.com by kurtzm...@gmail.com on 20 Dec 2010 at 1:59

[GoogleCodeExporter]

The new mini-map will create its own tile provider that is not shared between 
the main map, however the mini-map also has a constructor that accepts a 
supplied tile provider. If you want to use a custom tile provider it is not a 
problem to just create two tile providers (ProviderArrays) with the same 
modular tile providers and use one for the main map and one for the mini-map. 
However there is nothing to prevent you from sharing a tile provider between 
both, even though it will cause the issue described above.

The only solutions I can think of to resolve this in code are hackish and just 
get in the way to a problem that can be easily avoided. I don't want to make 
users of the tile provider "register" with it, but I don't know any other way 
to detect that two views are using the tile provider.

The solution to this might just be to discourage it, and clearly show in the 
javadoc how to avoid this.

Original comment by kurtzm...@gmail.com on 12 Jan 2011 at 5:04

[GoogleCodeExporter]

This has been documented and this ticket can be closed.

Original comment by kurtzm...@gmail.com on 27 Feb 2011 at 5:16

• Changed state: Fixed