[Tile-serving] [osm2pgsql] Excluding linestrings from polygon tables with multi backend (#270)

[Paul Norman]

> So basically, your conclusion is that from the tagging it is always possible to figure out the expected geometry?

In typical OSM fashion, this has four answers: no, maybe, yes, and it doesn't matter.

Some objects have contradictory tagging, such as `type=multipolygon` `area=no`. It's not possible to figure out the expected geometry, even if you know if the members are closed or not.

Leaving aside contradictory tagging, reproducing the current osm2pgsql behavior requires knowledge of if the way is closed or not. The multi-backend can't reproduce the pgsql backend, and still won't be able to with this change. But having `landuse=residential`, `power=line` or `power=sub_station`  be a linear feature or an area depending on closed or not is absolutely insane and we don't **want** to rely on that. So maybe.

I have yet to see any tagging in the wild that would be either a line or an area based on if its closed or not. If it's tagging that indicates an area and it's not closed, it's a tagging or geometry error. This matches mappers expectations, who often think of areas and lines of being different.

And lastly, for this specific bug, it doesn't matter, because in the context of a multi-backend, you are putting specific known features into tables, and you know if they're areas or not, and having non-areas in polygon tables violates your expectations.

---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/issues/270#issuecomment-148975909
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20151017/450a4398/attachment.html>