Re: [OSM-dev] osm2pgsql hstore support

2010-03-20 Thread Sven Geggus
Sven Geggus wrote: > Currently the patch has not been tested in conjunction with > incemental database updates. OK, in the meantime I can confirm, that incemental database updates work fine. I just fixed a problem with polygons (SVN r20565). When using hstores we need a way to mark polygon tags

Re: [OSM-dev] osm2pgsql hstore support

2010-03-16 Thread Peter Körner
Very cool this. What's the speed to access the hstore column, compared to the generic columns we now have? I did some benchmarks and the answer is - sometimes :) Without any indexes using a seqscan hstore is slower (see hstore-vs-columns-noindex.txt) hstore does not support (at least the doc

Re: [OSM-dev] osm2pgsql hstore support

2010-03-16 Thread Jukka Rahkonen
Sven Geggus fuchsschwanzdomain.de> writes: > > Hello, > > I just commited a patch for osm2pgsl for optional generation of a > hstore column (hstore new). Hi, Sounds great. I had been thinkin about splitting geometries and tags to a pairs of tables (one with columns geometry and osm_id, anothe

Re: [OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Peter Körner
[1> On the one hand, this has the potential to greatly increase the size of > the db, as you will now get all kinds of tags you're never interested > in. On the other hand, one can now quickly use keys one never used > before, without the need to reimport the whole db. When you just want to render

Re: [OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Sven Geggus
Lennard wrote: > Very cool this. What's the speed to access the hstore column, compared > to the generic columns we now have? I did not yet try to measure this. > On the one hand, this has the potential to greatly increase the size of > the db, as you will now get all kinds of tags you're nev

Re: [OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Lennard
Sven Geggus wrote: > Now we would just need something like this: > > select way,(tags->'man_made') as man_made FROM planet_osm_point where > ((tags->'man_made') is not NULL); > > What am I getting wrong here? Nothing. This will work. Mapnik just expects the name used in a filter to correspond

Re: [OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Sven Geggus
Martijn van Oosterhout schrieb am Sonntag, den 14. März um 17:06 Uhr: > Very cool. Now we just need to find a way to use it. I thought about > it earlier but I didn't see a way to make mapnik use it so left it. This won't work?! I didn't bother with this yet, but I just thougt that all I would n

Re: [OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Martijn van Oosterhout
Very cool. Now we just need to find a way to use it. I thought about it earlier but I didn't see a way to make mapnik use it so left it. Have a nice day, On Sun, Mar 14, 2010 at 4:47 PM, Sven Geggus wrote: > Hello, > > I just commited a patch for osm2pgsl for optional generation of a > hstore co

[OSM-dev] osm2pgsql hstore support

2010-03-14 Thread Sven Geggus
Hello, I just commited a patch for osm2pgsl for optional generation of a hstore column (hstore new). For those of you who don't know about hstore colums yet: Hstore is for sets of key/value pairs. As associative array datatype, just like a hash in perl or dictionary in python. This should come