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

Wed Feb 11 15:23:36 GMT 2009

Hi Paul,

> > Thanks a lot for the feedback. From what you report, it sounds like 
> > the new map worked quite well with no really bad problems.
> > 
> >> Given that I'm comparing gmapsupp.img files for combined England,
> >> Scotland and Wales previously created by using osm2mp and mkgmap with
> >> .mp files and now directly with mkgmap from .osm files then the initial
> >> observations are:-
> >>
> >> Using the osm2mp method gives a final file of 119.8MB where the direct
> >> conversion gives a final file of 103.7MB
> > 
> > Hmm, did it only include 90% of the roads?
> 
> I don't know how you'd tell!!

I think others are seeing similar shrinkage and nobody, yet, has reported
any missing roads.

> > 
> >> Support for roundabouts seems less well implemented. On approaching a
> >> roundabout I would previously receive an instruction to "Take 3rd exit
> >> at roundabout". Approaching the same roundabout I was simply told to
> >> "Keep left to Bloggs Road" even though I needed the 3rd exit.
> > 
> > OK - the attached new lines file should fix that (this goes in
> > resources/styles/default).
> 
> Helped but is still not as it was but I note that r877 has roundabout
> updates. A quick glance at the text description for a route using r877
> has the "Take 3rd exit at roundabout" direction so this may already have
> been fixed. I'll see what the graphical map does when I go home

Yes, it now knows about exits but, as you know, there is still a problem
with the text citing the wrong exit.

> >  
> >> Slightly different routes are now chosen. These are not wrong - Just
> >> different to before. e.g. Leaving work I always ignore the Garmin route
> >> as it's not the best. When it recalculated it previously chose the route
> >> I actually take. Using the new file it picked a different route which I
> >> had to ignore a second time before it chose the route I actually take.
> > 
> > Was the new route completely silly or merely sub-optimal? If the later,
> > how could it have improved the route?
> > 
> 
> Sub-optimal. The best route home for me is to use 2 motorways (think the
> 2 shortest sides of a triangle) whereas the Garmin has always tried to
> send me along the hypoteneuse. Unfortunately whilst this is shorter this
> is an A road which goes through three middle sized towns and therefore
> has traffic lights B roads etc all the way along so takes considerably
> longer. 

I believe it is a common gripe that the Garmin does not pay any attention
to traffic lights.

> Previously after the first recalculation the Garmin would send
> me along the motorway. Now it tries to get me back onto the A road until
> the second recalculation.

OK - don't know why that should be.

> > Tweaking of the road classes and speeds in the lines file should alter the
> > behaviour so you could try adjusting those values. I thought I had set
> > them to the same values used by osm2mp.
> >  
> >> I also noticed today that with both versions of the .img file any
> >> routing from the Rochdale area along the M62 fails at around the point
> >> that the M621 splits from the M62 (approximately J27). I realise there
> >> could be many reasons for this but I thought I'd mention it.
> > 
> > Does a split between map tiles occur there?
> > 
> Yes the split in the tiles would appear to be here. Further testing
> *"suggests"* that I can't route anywhere off the starting tile. I could
> try increasing the overlap in splitter to see if this helps but I also
> don't seem to be able to route into Scotland or Wales. I've also noticed
> these places show in the "Find Cities" list without distance or direction.

Yes, inter-tile routing is not yet supported. It's in the pipeline...

Cheers,

Mark