COUNT DISTINCT takes much longer than DISTINCT and crashes OSM Planet
#1082
Comments
hannahbast
pushed a commit
to ad-freiburg/qlever-petrimaps
that referenced
this issue
Sep 4, 2023
The previous query used `COUNT(DISTINCT ...)`, which amounts to a `GROUP BY`, which is rather slow and crashed the latest version of OSM Planet. Now just use the "all distinct geometries" query (which we have to compute anyway), as subquery of a simple `COUNT`. See also ad-freiburg/qlever#1082
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This query materializes all OSM geometries. The query execution tree is an
INDEX SCANfollowed by aDISTINCT. It takes around 2 seconds on OSM Germany and around 50 seconds on OSM Planet.https://qlever.cs.uni-freiburg.de/osm-germany/jO4LF9
This query returns the size of that result. The query execution tree is an
INDEX SCANfollowed by aGROUP BY. It takes around 20 seconds on OSM Germany and crashes the server for OSM Planet.https://qlever.cs.uni-freiburg.de/osm-germany/r4ICBq
Here is a workaround, which just takes the first query as subquery and counts the result.
https://qlever.cs.uni-freiburg.de/osm-germany/7H9cJp
The text was updated successfully, but these errors were encountered: