[OSM-dev] planet_osm2txt with planet-060818?

[Nick Whitelegg]

>Well, this sounds like the same as it is done in planet_osm2txt.pl. It's 
also
>event driven. But reading all nodes into memory sucks a lot of memory for 
the
>current planet.osm File.
>What I'm testing is to only suck in all nodes for a
>given region. This already helped with processing the file in 
check_osm.pl to
>redure the memory imprint.

There is I think a strong need for a utility to only put a given tile of 
planet.osm into a database, that way people can easily set up localised 
planet-based databases. The trouble is a) the checks slow it down 
significantly, (e.g. using REXML and the July planet.osm took a day to put 
the UK OSM data into Freemap) and b) reading in all the nodes into memory 
(necessary for filtering out segments) takes up a lot of memory - by my 
estimation 320MB, requiring a 512MB box, so doable on the average newish 
desktop PC but not, say, a 128MB virtual machine :-|

I guess one way of solving these issues would be to load the whole of 
planet into the database, then delete nodes, segments and ways outside the 
required area using SQL statements. 

Nick