Dave,

As I think, in general, task --ttc should be used next way:
1. create filtered osm file with --tt task and given rules: transformed.osm,
2. update transform.osm file with --ttc task (source.osc) and given rules.
(simply keep transformed.osm up-to-date with upcoming changesets)

Here, we don't expect that deleted items are same as in transform.osm -
they're already transformed, so we need to make them equal before applying
changes.

Is there any other cases acceptable for --ttc task? My opinion here, that we
should transform _all_ incoming changes (Add, Modify,Delete) because they
will not be merged.


Alexander Chmyr
Senior Software Engineer
Cogniance, Inc
--------------------------
skype: alex.smudgy



On Tue, Dec 8, 2009 at 12:45 PM, Dave Stubbs <osm.l...@randomjunk.co.uk>wrote:

> On Tue, Dec 8, 2009 at 10:16 AM, Anton Popov <apo...@cogniance.com> wrote:
> > Hello Dave,
> >
> > We can split the logic for sure. I will refactor these two tasks.
> >
> > Speaking about delete actions, I can say I've ignored them expressly,
> > because there is no big sense to transform(!) tags in entity, that will
> be
> > deleted. I'm new to Osmosis and if you will place arguments when my
> > assumption can be wrong I can make changes for the code.
> >
>
>
> Yeah, they don't need transforming, that's fine. The issue is whether
> they are being dropped from the ongoing change stream? The change
> stream comes into the task via the process method, and the ongoing
> stream leaves it via the call to the Sink.process method. So creates
> and modifies will pass through the transform and only get piped to the
> sink if the transform didn't drop it, but the deletes are never sent
> to the sink, no matter what.
>
> As I said this was just me looking at the code, I'm fairly sure it's
> wrong but haven't had time to test -- just put it through this use
> case: create a changeset that includes a delete, then using a pass
> through transform you should get the same changeset out the other end,
> including the unaltered deletes.
>
> Dave
>
_______________________________________________
osmosis-dev mailing list
osmosis-dev@openstreetmap.org
http://lists.openstreetmap.org/listinfo/osmosis-dev

Reply via email to