[OSM-dev] [OSM-newbies] Should osm2pgsql create the planet_osm tables?

[Dave Stubbs]

On Jan 22, 2008 10:15 AM, Alexander Wright <silverfish at wright-family.me.uk>
wrote:

> On Tuesday 22 January 2008 09:55:23 Dave Stubbs wrote:
> > On my box this takes several hours for the latest planet, and spends at
> > least an hour with no output as postgres is indexing the data.
>
> I've not seen the insert code, so please ignore me if this is done
> already!
> I believe that if you are importing large amounts of data, its usually
> better
> to drop indexes on a table, insert the data, the recreate the indexes
> afterwards.
>
> I'm quite prepared to believe there is a good reason for not doing this,
> though.
>
>
I think this is effectively what it's doing (it uses a postgres COPY
statement too). Unfortunately there are about 10 million ways in the DB
being added to GIST index, and it takes a while on my box as there's only
1GB RAM.

I haven't looked at that part of the code in much detail either... I'm
guessing it isn't doing anything particularly silly though, and before the
tiger import it worked in a very reasonable amount of time.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20080122/04e75505/attachment.html>