[Imports] First questions

[Pieren]

2009/7/14 Iván Sánchez Ortega <ivan at sanchezortega.es>:
> El Lunes, 13 de Julio de 2009, Pieren escribió:
> If everything is connected, I think that the only way to do this properly is
> to have one single .osm file.
>
> However, IIRC, the current state of the bulk uploaders will upload all nodes
> first, then all ways, then all relations. The uploaders should be changed so
> they upload stuff recursively to minimize the amount of non-connectec nodes
> between changesets.
>
>
> Cheers,

Thanks for the reply. I just resent the same post at
imports at osmfoundation.org. I hope I didn't double posted...

About the uploader (bulk_upload.py from svn/0.6) , we already use a
modified version which creates elements in the same sequence as they
are in the .osm file. We also modified the python script to allow XML
parsing for such file size (SAX parsing instead of DOM). Also the way
how the key map (old_id -> new_id) is stored on a file has been
changed for the performance (use python persistency shelve module).
With all these changes, we can handle the 1.4GB xml file with almost
no impact on performance or RAM usage.
Pieren