[mkgmap-dev] Should inter tile routing still work when splittingSplitter results again?

Tue Sep 29 15:46:36 BST 2009

L> A solution:
L> I created a PHP script that splits America with an high initial node
L> count (of e.g. 1.400.000) and tries to render each resulting tile. If
L> a
L> tile render fails then that area is split again with an ever lower
L> node
L> count until at least two subtiles emerge which are then rendered.

This sounds like a good approach, as least until we have a better way to 
determine where to split tiles.

L> Neighboring tiles sometimes overlap because splitter seems to use a
L> little larger bbox for splitting then what the source file contains.

I'll need to think about this a little. The problem will almost certainly 
be due to the way tile boundaries get rounded so they are correctly aligned 
- the splitter tries to align the tile boundaries appropriately and will 
grow the edges if need be to achieve this. Unfortunately I won't be able 
to look at this and figure out what the best solution is for about another 
3 weeks as I'm heading off on holiday until the 19th. My first thought is 
that tile boundaries should not be allowed to move if they are already aligned 
correctly, rather the location of the split for that tile should be adjusted 
instead.

L> Will routing be a problem because the result of recursively splitting
L> and rendering results in tiles that apparently do not line up
L> perfectly?
L> I've done a few tests and it feels that routing has become worse, but
L> I
L> don't have a definitive answer. If so, what can be done about this?

I don't have a definitive answer either, I'd need to study the problem a 
bit closer before I'd understand exactly what the consequences are. My suspicion 
however is that yes this will cause problems, because nodes on tile boundaries 
won't always match up with boundary nodes on neighbouring tiles, and this 
is a no-no as far as routing is concerned.

Sorry I can't be of more help in the short term.

Chris