[mkgmap-dev] Commit: r868: Implement routing for .osm input files.

Wed Feb 11 22:46:16 GMT 2009

Hi Mark,

Mark Burton schrieb:
>> - It seams that restrictions are not implemented. Nuvi tried to route me 
>> through a "destination only" road. Didn't happen with polish version.
>>     
>
> I guess you mean ways that are tagged access=destination?
>
> The attached patch fixes that case. It's too strict in that it will stop
> all routing using the tagged way but I don't know how else to handle it.
This road is actually tagged with "motorcar=destination" It seams that 
it was pure luck that this road was not used with the polish map version 
because I can't find any check of the destination tag in there. I think 
it was not taken because of max_speed 30 tag. This road is really 
shorter but residential and a lot of drivers ignore the restriction. 
They don't know that a policemen lives in that street ... I'd better 
take the major road around it.  Thanks for the patch I will check the 
effects of the restriction stuff ...

>> - Map generation with bigger maps seams to be rather time consuming. 
>> Generation of my map (160km x 160 km) takes now 19 minutes.  With the 
>> old method it took 3 minutes for osm2mp and 4 minutes for mkgmap. I ran 
>> hprof but I didn't find the time consuming portion until now.
>>     
>
> hprof showed that quite a lot of time is going in Coord.distance(). The
> orginal road name pois code used that method a lot to find the nearest
> city to each road. So that will get improved when your more efficient code
> is integrated.
>
>   
You are totally right. This is caused by the road-pois and not by the 
new routing stuff. I had the same problem with my pois when I had the 
linear search. Search times reduced by factor 10 or so when I switched 
to tile bases search. Sorry about that false alert.

Berni.