[OSM-dev] Full history extracts

[Brett Henderson]

On 24/12/2011, at 8:19 PM, Frederik Ramm <frederik at remote.org> wrote:

> Hi,
>
> On Sat, 24 Dec 2011 13:47:48 +1100
> Brett Henderson <brett at bretth.com> wrote:
>> Osmosis doesn't currently support the full history data format.  It
>> only supports the alternative osmChange format.
>
> Suppose I were to write a simple Perl script that streams the full
> history file (which is sorted by type then id then version), and wrap
> every "version 1" entry in <create>...</create> and every other entry
> in <modify>...</modify> unless it has visible=false in which case I'd
> choose <delete>...</delete>.
>
> Do you think this would then work with Osmosis?

Yes it would. But the diff application logic isn't very efficient
because it runs a fair few queries per entity to be updated.  Several
to delete any existing identical items, plus the various history and
current inserts. It would take forever to complete the load.

I think a better approach would be to create dump files to be loaded
in bulk, but not even osm files can be loaded into apidb using osmosis
currently. The existing apidb code is a minimal update to previous
MySQL code and doesn't take advantage of pgsql bulk loading like the
pgsnapshot code does.  Just a simple matter of coding ;-)

>
> Bye
> Frederik
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev