[OSM-dev] Osm2pgsql with -slim -no_diffs?

[Peter Körner]

Am 22.06.2011 16:12, schrieb Jukka Rahkonen:
> Hi,
>
> I do not use diff files for updates but because of too little memory I
> still need to run osm2pgsql in slim mode. I have a feeling that the slim
> mode is doing now a lot of unnecessary work for me and it could be much
> faster without all the stuff that is needed for preparing the database for
> diffs.
> Is there some hidden option for running osm2pgsql in a fast memory saving
> mode is there is no need to import diffs in the future?
There isn't.

Frederik Ram wrote some mail some time agao, pointing out that there are 
actually 4 modes of import:


              | Updatable    | Not Updateable
----------------------------+----------------
Slim         |      X       |
None-Slim    |              |        X


osm2pgsql currently can't handle the modes without X


Slim: node-locations are (partially) stored to disk (going over the 
database, using a sparsetable, an mmaped file, etc.)

None-Slim: node-locations are stored completely in RAM


There is another unclear thing: when importing all nodes, even those 
without tags using the -x|--extra-attributes option, the slim-tables are 
somewhat redundant.


I think it might be a good idea to build a new importer, possibly based 
on osmium which does allow to use different node-locations-stores incl. 
google sparsetable, mmap and memory-only and does the multipolygon 
assembly of relations. The decision which ways are stored as polygons 
and which are as linestrings is up to the author of the importer.

Peter