It can be done using flex output byt filtering ways by tags and then removing (or ignoring) the unused nodes.
It can greatly improve the usability since it would not be necessary to use osmium.
However, for large maps, it could take a long time to import (or it would not fit into memory). In such cases the user can always use osmium,
We can also add filtering of relations and include (only) the relation members. See these two links for examples: link1, link2. Focus on function select_relation_members that returns node and way ids that should be processed in the second stage (can be checked via osm2pgsql.stage). Unfortunately, I did not find an equivalent for ways.
It can be done using flex output byt filtering ways by tags and then removing (or ignoring) the unused nodes.
It can greatly improve the usability since it would not be necessary to use osmium.
However, for large maps, it could take a long time to import (or it would not fit into memory). In such cases the user can always use osmium,
We can also add filtering of relations and include (only) the relation members. See these two links for examples: link1, link2. Focus on function select_relation_members that returns node and way ids that should be processed in the second stage (can be checked via osm2pgsql.stage). Unfortunately, I did not find an equivalent for ways.