Sarah Hoffmann

Pypy3 now seems to work but at least one test is slow to the point of being unusable. CI also has issues when running with a self-compiled shapely. Both needs...

I'm closing this as a won't fix. The new Python frontend drops batch processing in favour of using Nominatim as a library.

To add some context: the problem here is that some normalized versions of words have a lot of tokens in our word table now. In this case the problematic ones...

Tokens with the same transliteration are now combined and only result in a single search being emitted.

See https://github.com/osm-search/Nominatim/blob/5751686fdce33f6b69a93dd1641ceacb2239791f/lib-sql/functions/placex_triggers.sql#L1034 to learn where local names are added. See https://github.com/osm-search/Nominatim/blob/5751686fdce33f6b69a93dd1641ceacb2239791f/lib-sql/functions/placex_triggers.sql#L91 for where the place names from the linked place are linked in.

Here is how this could work: * add a new import mode `nominatim import --prepare-database`, which executes all commands that need privileged mode (create database, create extensions, anything else?) *...

Please go ahead.

This is now implemented on master since #3204.

We now have the following working: https://nominatim.openstreetmap.org/ui/search.html?street=23&city=Gries+im+Sellrain https://nominatim.openstreetmap.org/ui/search.html?street=23+Gries https://nominatim.openstreetmap.org/ui/search.html?street=23+Gries+im+Sellrain For the Viaduktboden address, thing still work like this: https://nominatim.openstreetmap.org/ui/search.html?street=Viaduktbogen+13&city=Innsbruck That's probably as good as it gets. So closing as resolved.

Finally closing this. We have a list of postcode regexps now. We just need to integrate them in search somewhat differently.