[Tile-serving] [osm2pgsql-dev/osm2pgsql] Non moving node updates should not update related way/relation geometries (#1046)

[Jochen Topf]

I did some statistics on node changes using the planet history file. Of the about 2 billion node changes (i.e. new versions of an existing non-deleted node) about one quarter did not change the location. Many of these will be for POIs which are not in a way or relation, so this is not so relevant here. But even if you only look at untagged nodes (no tags in old and new version), in about 10% of cases the location did not change. Presumably these are all members of a way or relation so they will trigger further processing unnecessarily.

I did some experimentation in a [branch](https://github.com/joto/osm2pgsql/tree/EXP-node-not-moved-issue-1046), but it is not totally straightforward to change the current code, so this would need a lot more testing and especially benchmarking. It makes the code more complex, so this has to be evaluated thoroughly. If anybody want to play around with it, go ahead. At the moment I don't have the time to work on that.

This is also part of a larger issue: How do we handle changed objects in a better and more flexible way instead of just deleting and inserting. And how do we detect early that changes in the OSM data do not actually result in changes in the database, thus allowing us to elide further processing through the processing chain from nodes to ways to relations, to generalization to tile invalidation. This needs a lot more thoughts.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/osm2pgsql-dev/osm2pgsql/issues/1046#issuecomment-2294772753
You are receiving this because you are subscribed to this thread.

Message ID: <osm2pgsql-dev/osm2pgsql/issues/1046/2294772753 at github.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20240817/dbdca3b1/attachment.htm>