[mkgmap-dev] minor enhancement to tile-splitter

Sun Mar 8 22:03:14 GMT 2009

On Sun, 2009-03-08 at 21:01 +0000, Steve Ratcliffe wrote:
> Hi
> 
> > The attached patch makes the tile splitter look at the width of the area
> > at both the min & max latitude. This slightly reduces the number of
> > tiles needed for the planet and makes the splitting slightly faster.
> 
> I have applied the patch - thanks.
> 
> > Mkgmap does not manage to generate a map from the output though, it
> > fails with an out of heap space error. I'll try adding something which
> > limits the maximum bounds of each tile too.
> 
> Yes that should help.
> 
> In some cases it might be just enough to pull the boundary of the tile in
> so that it encloses the points that exist.  There is no need to tile the
> ocean or places where there is no data.

I believe you are right, the best solution would probably be to discard
these empty tiles. I tried adding code to split the tiles based on
lat/lon and ran into some null pointer exceptions when splitting tiles
containing no data. I have attached the patch I used, but it isn't in a
state to be applied.

If we wanted to make the processing more efficient, I was wondering if
we should import the data to a coarse grid of areas to start with. Then
either sub-divide those afterwards, or divide them dynamically as each
one hits the node limit. This avoids the first few splits which would
normally need to process all the imported data.

	Jon

-------------- next part --------------
A non-text attachment was scrubbed...
Name: lots-of-changes.patch
Type: text/x-patch
Size: 4577 bytes
Desc: not available
Url : http://lists.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20090308/b9a6a39b/attachment.bin 