[OSM-dev] Fastest way to extract a bounding box

[marqqs at gmx.eu]

Frederik is right.
The big question is: do you need _all_ OSM information of the chosen area?
If not, the processing time will decrease a lot - regardless of extracting the 5 x 5 km box from a file or from a database.


Database:

You could store only that information you need. If necessary, create a second database for this purpose. I'm not an expert in databases but certainly others will have ideas to improve performance in this case.


File:

Prepare your file by deleting every OSM object and every tag you do not need.
Meanwhile I did some tests - I was curious... I chose a 5 x 5 km area with a lot of data in it: city of Bielefeld. Then I ran some commands on my (weak) computer:

$ time osmosis/bin/osmosis --read-pbf file="germany.osm.pbf" --bounding-box left=8.5 bottom=52 right=8.575 top=52.045 cascadingRelations --write-xml file="5x5o.osm"

real	6m3.796s
user	6m19.484s
sys	0m6.788s

$ time osmconvert germany.osm.pbf -b=8.5,52,8.575,52.045 >5x5.osm

real	2m16.491s
user	1m59.091s
sys	0m2.868s

$ osmconvert germany.osm.pbf --out-o5m >germany.o5m
$ time ./osmconvert germany.o5m -b=8.5,52,8.575,52.045 >5x5a.osm

real	1m43.370s
user	1m22.617s
sys	0m3.924s


Conclusion:

If you decide to extract this 5 x 5 km area form a file, it is better _not_ to take Osmosis and _not_ to use .pbf format. It seems, osmconvert is much faster, especially if you have converted your data into .o5m format.


Something else:

The file Osmosis has generated seems to be incomplete. E.g, there is a relation 445596 but its parent relation 63226 is missing.
http://www.openstreetmap.org/browse/relation/445596
http://www.openstreetmap.org/browse/relation/63226
Did I do something wrong with the option "cascadingRelations"? Can anybody help, please...?

Markus



-------- Original-Nachricht --------
> Datum: Wed, 01 Jun 2011 19:10:01 +0200
> Von: Benjamin Meier <b.meierr at googlemail.com>
> An: dev at openstreetmap.org
> Betreff: Re: [OSM-dev] Fastest way to extract a bounding box

> Hi, you're of course not bothering me :-)
> 
> I didn't expect that the difference is so much depending on those things.
> 
> I don't need relations especially if they dramatically slow down the 
> process.
> 
> Ways with nodes inside and outside of the bbox shall be included but if
> it's faster and/or possible the ways may end at the bbox-border.
> 
> As a rough rule of thumb I can say that I would prefer the fastest way 
> possible.
> 
> Benjamin
> 
> Am 01.06.2011 18:36, schrieb marqqs at gmx.eu:
> >> What I would like to know is if there's a significantly faster way to
> do
> >> that.
> > "That depends..." ;-)
> >
> > Which option do you need? Do you need relations? Do you need relations
> which have relations as members which have ways which have nodes inside the
> box?
> >
> > Example: There is a bus line which crosses the bbox. This and other bus
> lines have the same operator which is represented by a relations. Do you
> need this information? Do you need the bus line?
> >
> > What is to be done with ways which have some nodes inside the bbox and
> some nodes outside of it? Shall these ways be included? Or excluded? If
> included, what to do with the nodes? What do do with the noderefs, if some of
> the nodes are not included because they are outside the bbox?
> >
> > Sorry to bother you with these Questions... but there are a lot of ways
> to extract a bounding box, and the time you need depends on the option you
> want use.
> >
> > Markus
> >
> > -------- Original-Nachricht --------
> >> Datum: Wed, 01 Jun 2011 18:22:12 +0200
> >> Von: Benjamin Meier<b.meierr at googlemail.com>
> >> An: dev at openstreetmap.org
> >> Betreff: [OSM-dev] Fastest way to extract a bounding box
> >> Hi,
> >>
> >> I'm looking for the fastest way to extract a bounding box from
> osm-data.
> >> The intention is to get a section (a square of about 5km * 5km)
> >> around a user's current position in a few seconds.
> >>
> >> The input-data shall be of the size "germany", "central-europe" or even
> >> the whole planet.
> >>
> >> Up to now I made some experiments with unsatisfying results.
> >>
> >> I extracted such a bounding-box from "germany.osm.pbf" with osmosis in
> >> about
> >> 110 seconds. After that I installed a Postgis/Postgresql-DB, filled it
> >> with the
> >> pbf-file (as described in the wiki:
> >> http://wiki.openstreetmap.org/wiki/Osmosis_PostGIS_Setup)
> >> and extracted the bounding-box with osmosis from the DB in about 53
> >> seconds.
> >>
> >> The hardware is a core2duo e7400 with 2GB RAM.
> >>
> >> What I would like to know is if there's a significantly faster way to
> do
> >> that.
> >>
> >> Thanks,
> >> Benjamin
> >>
> >> _______________________________________________
> >> dev mailing list
> >> dev at openstreetmap.org
> >> http://lists.openstreetmap.org/listinfo/dev
> > _______________________________________________
> > dev mailing list
> > dev at openstreetmap.org
> > http://lists.openstreetmap.org/listinfo/dev
> 
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev