[mkgmap-dev] [PATCH] Inter-Tile routing - alpha code for testing

Tue Feb 17 18:35:46 GMT 2009

Hi Clinton,

Thanks very much for the report.

>...
> mkgmap then gave a lot of messages similar to the following:
> 
> Way K 7006's last point at 52.91016/12.02696 is a boundary node
> Way K 7005's last point at 52.91016/11.92830 is a boundary node
> Way K 7005's first point at 52.91016/11.88691 is a boundary node
> Way K 7005's last point at 52.91016/11.88111 is a boundary node
>

Good, that shows it's found what it thinks are boundary nodes on the edges of
the clipped regions.

Can you please check that the coordinates do actually correspond to points
that lie on the edges of the clipped regions.

> This was tested on a Windows Vista machine (sorry, that's what I had available).
>
> Am I perhaps missing an option or similar? I believe that splitter.jar
> correctly creates boundary boxes, so that should not be a factor.

I haven't actually used the splitter myself so I don't know much about
its behaviour/usage. The boundary nodes are actually created within
mkgmap when it clips the ways to the region.

> To test the inter-tile routing I used Mapsource (this has generally
> been a reliable indicator of routing success), and first created a
> simple route on an Autobahn within a tile. This worked correctly. I
> then added to the route another point on the same Autobahn, not far
> away, but across a tile boundary. Mapsource took longer to calculate
> the route, but then added a straight line between the last two points,
> instead of routing along the Autobahn.

Hmm, your test sounds good but the result wasn't.

I don't know if the boundary node code has ever actually been used
before. Perhaps Steve or Robert can confirm whether it is known to work
or not.

> I hope this helps you in some way.

Yes, thanks a very useful report. If you have the time, please check if
(some of)those boundary points are actually where you think they should be.

Cheers,

Mark