[OSM-dev] New, faster, planet dump tool

Brett Henderson brett at bretth.com
Thu Sep 27 00:01:59 BST 2007


Robert (Jamie) Munro wrote:
> Presumably if you run the changeset that covers the time over which the
> planet dump was made, it will be able to apply all the changes that the
> dump missed, and ignore all the ones it got. That way, you will have a
> consistent planet replica going forward.
>   
That should work.  One thing to note with all of this is that osmosis 
won't create identical history on the destination database.  When 
extracting changes it will summarise all the changes within a time 
interval.  In other words, it will only use the most recent change 
within the interval for each entity.  This was a decision I made when 
writing the --mysql-read-change task to reduce the amount of data for 
large time intervals and to make it easier to apply changes to xml 
files.  Most of osmosis could work with complete history though so this 
can be changed in the future if required.

If you create overlapping changesets as described above the current 
tables in the destination database should end up correct so long as you 
apply the changes in the correct order .

One issue with the current osmosis code (and I intend to fix it) is that 
it doesn't create user ids on the destination database.  It creates a 
single osmosis user and assigns all changes to that user.  I only 
recently added user attribute support to most tasks but haven't added it 
to the mysql writing tasks.  It's not a big job to add.




More information about the dev mailing list