[OSM-dev] [OSM-talk] Osmosis error with multiple bouding boxes

Karl Newman siliconfiend at gmail.com
Thu Oct 25 19:09:15 BST 2007


Crossposting to dev.

On 10/25/07, Lambertus <osm at na1400.info> wrote:
> Nice, autorouting is also next on my list. You're using cGPSmapper right?

Yes, I have to use cGPSMapper to accomplish autorouting, because it
seems that the routing fields in the IMG format are not publically
known/documented. I actually created a version of the osmgarminmap XSL
sheet that would create routable polish format (MP) files, but it only
worked on tiny OSM files because XSL is the wrong tool for the job. If
anyone wants to play with it, though, you can look at it at
www.migratingcoconuts.com/pub/osmarminmap/ You can test the routing in
GPSMapEdit, too. It wasn't a total waste of time, though--it gave me
some good ideas for a "production scale" version. I particularly like
the rules file that was adapted from osmarender. I think it will be a
great tool to allow people to create customized GPS maps (say, if you
only want footpaths or something like that).

Karl

>
> Osmosis needs some work with regard to tee and many outputs still, because
> it's quitting here with 100+...on out of heap space errors (The server has
> 4GB), with 50 pipes it worked. But other than that it works many times
> faster than processing the planet all over for each tile. So I'm happy with
> it.

Maybe the task could be tweaked with an option to buffer all the data
into a file, then send it on to each pipe sequentially. You could also
set up a hierarchical/recursive series of Osmosis calls (sort of lame,
I know).

>
> ----- Original Message -----
> From: "Karl Newman" <siliconfiend at gmail.com>
> To: "Lambertus" <osm at na1400.info>
> Cc: "Brett Henderson" <brett at bretth.com>; "OSM Talk"
> <talk at openstreetmap.org>
> Sent: Thursday, October 25, 2007 19:11
> Subject: Re: [OSM-talk] Osmosis error with multiple bouding boxes
>
>
> On 10/25/07, Lambertus <osm at na1400.info> wrote:
> > I've tested 0.21 now with the 'tee' command where 4 outPipes are created
> > and
> > 4 bounding boxes that use these outPipes and this is the result:
> > <snip>
> > So apparently t1 and t2 are functional, but t3 isn't. Scaling down the
> > test
> > to only two outPipes results in succes. Probably a software coded limit on
> > outPipes or somesuch?
>
> You forgot the outputCount=4 argument to the --tee task.
>
> I'm working on a Garmin auto-routing map generator, and I'll also need
> the planet chopped up into tiles/segments. Maybe we can come up with
> an Osmosis task that will do it more efficiently. In my case, though,
> I'll need to keep the ways inside each segment complete (include all
> the nodes in each tile from ways that cross tile/segment boundaries)
> in order to properly set up routing between neighboring tiles.
>
> It's a little difficult, but not impossible. It's easy enough to split
> out nodes into individual tiles/segments, but more difficult to make
> the ways complete across tiles. The new RandomAccessObjectStore might
> be helpful here as a temporary storage for all nodes. I'll think about
> a possible structure for this. Maybe Brett has some ideas, too.
>
> Karl
>
>
>




More information about the dev mailing list