[OSM-dev] AND upload almost ready

[Brett Henderson]

Martijn van Oosterhout wrote:
> Ok, but what I was thinking of was more --write-api-change. This way
> all the stuff still goes via the API, but by running it directly on
> the server via the loopback interface you don't have the RTT and
> bandwidth limitations.
>
> I was more asking if TomH had any problem with running the program
> osmosis, being written in Java. I could well imagine the
> --write-mysql-change not quite being up to it yet.
>   
No worries.  I'll leave it to you guys to figure out the best way of 
doing it.  If you want to use osmosis and need to create/enhance tasks 
I'm happy to help.
>> It also has one limitation that may
>> rule it out completely at this point. The "create" code assumes that all
>> entities have already been assigned ids unlike JOSM which I believe
>> assigns them a negative id until a real one is allocated.
>>     
>
> This is a SMOP (simple matter of programming). I might do this. At the
> very least the --sort stuff will be useful.
>   
Cool.  If you have any questions on how things hang together or want 
some assistance getting started let me know.  If you wish to create a 
--write-api-change task you'll need to create a new implementation of 
the ChangeSink interface.  XmlChangeWriter is an example of this.  
You'll then need to create a matching factory class to make it creatable 
via the command line (eg. XmlChangeWriterFactory).  Finally you'll need 
to register the factory with a name in TaskRegistrar.

If you want to enhance or fix the existing --mysql-change-writer task, 
it's implementation is unsuprisingly in MysqlChangeWriter.

Cheers,
Brett