[OSM-dev] simple tool for finding minimal set of osmosis diff files

[Timo Juhani Lindfors]

Hi,

Brett Henderson <brett at bretth.com> writes:
> Osmosis has a task named --read-change-interval which will download all 
> changes from the last time it ran (it maintains a timestamp file), and 
> merge them into a single changeset.  That changeset can then be used to 
> patch a file, or imported into a database, etc.

Aha, due to java issues I was using the java5 compatible version which
does not have that option:

$ grep -ri read-change-interval osmosis-0.24.1-java5
$

> For example, I use it to maintain a local copy of all data in Myanmar.
> http://www.bretth.com/myanmar/
> The main file of interest in that directory is replicate_osm_file.sh 
> which is launched from cron hourly.  configuration.txt defines which set 
> of diffs to download, I'm using hourly diffs but minute and day diffs 
> will also work.

Interesting, I didn't know you could apply diffs if you only have some
small bounding box of data.

> There is one big problem with your script which should be easy to fix.  
> DON'T use the bz2 daily files, they use a different timezone offset 
> (BST) to the gz daily files (UTC) so will contain slightly different 
> data.  I'm going to disable them very soon in favour of the new gz 
> files.  The gz files are a bit larger but fix timezone issues and 
> drastically improve the reliability of daily files.

Thanks for the warning :-) I fixed the script.

> One nice thing about your script is that it uses daily files and hourly 
> diffs, osmosis won't do that for you.  It's probably faster to use your 
> approach because less files are downloaded, although the total bandwidth 
> consumption will be almost identical.

Yep, I am experimenting with

http://svn.openstreetmap.org/applications/utils/export/tiledata/import

and trying to make it support minutely diffs so I really need more
than just hourly diffs. I need to add support for the case where
starttime is not at midnight.

best regards,
Timo Lindfors