[OSM-dev] osm2pgsql Lua tag transformations

[Sarah Hoffmann]

On Thu, Jun 25, 2015 at 11:06:41AM +0200, nebulon42 wrote:
> Thank you for explaining this. Would you see the possibility of
> removing relation members from the geometry as something that could
> be added as feature? Or would it be difficult/not worth/...?
> Otherwise I could follow up on this on the osm2pgsql issue tracker.

I don't know how useful it would be for the wider audience but
it would not be very difficult to include. You can open a feature
request on github but seeing that osm2pgsql is best described as
pityware(*) these days, I doubt that much will happen unless such
a request is accompanied by a pull request.

(*) Pretty much abandonend but occasionally somebody takes pity
    and fixes the worst issues.

Sarah

> For completeness reasons: Does anybody know of another or possible
> way of solving this problem of platforms being part of the route
> (like here:
> https://www.openstreetmap.org/#map=18/48.19117/16.34794&layers=T)?
> 
> Thanks, nebulon42
> 
> Am 2015-06-24 um 21:37 schrieb Sarah Hoffmann:
> >Hi,
> >
> >On Sat, Jun 20, 2015 at 05:47:43PM +0200, nebulon42 wrote:
> >>I'm currently experimenting with Lua tag transformations when
> >>importing data with osm2pgsql and trying to remove members of
> >>type=route relations (public transport routes) that have the role
> >>platform from planet_osm_line. The problem I'm trying to avoid can
> >>be seen here:
> >>https://www.openstreetmap.org/#map=18/48.19117/16.34794&layers=T
> >>(platforms coloured like routes)
> >>
> >>I thought that when I set membersuperseded = 1 for the corresponding
> >>relation members in filter_tags_relation_member this would work, but
> >>apparently it isn't.
> >
> >membersuperseded doesn't really do what you think it does. If a way
> >in a relation is marked as superseded, then the way itself will no
> >longer be considered an object of its own but just a part of the
> >relation. Or to put it the other way around, if membersuperseded is 0
> >then the way may still be included in the output tables as a
> >separate object (naturally depending on if it has interesting
> >tags or not).
> >
> >It is currently not possible to exclude specific members in a
> >relation from taking part in the computation of the final geometry.
> >The geometry computation always uses all ways available.
> >
> >Sarah
> >