We used an old GDAL Version (3.2.1) to load data to a postgis db with ogr2ogr. There was an automatic spatial index generation taking place, which seemed to have a logic to avoid duplicated index names.
Examples with GDAL 3.2.1:
For a table with the name
"ch059_liegenschaften_grundstueckpos_hilfslinien" a index named
"ch059_liegenschaften_grundstueckpos_hilfs_wkb_geometry_geom_idx" got
created.
For a table with the name
"ch059_liegenschaften_grundstueckpos_hilfslinien_proj" a index named
"ch059_liegenschaften_grundstueckpos_hilfslinien_proj_wkb_geomet" got
created.
Examples with GDAL 3.9.1:
For a table with the name
"ch059_liegenschaften_grundstueckpos_hilfslinien" a index named
"ch059_liegenschaften_grundstueckpos_hilfs_wkb_geometry_geom_idx" got
created.(equal like GDAL 3.2.1)
For a table with the name
"ch059_liegenschaften_grundstueckpos_hilfslinien_proj" it tries to
create an index named "ch059_liegenschaften_grundstueckpos_hilfs_wkb_geometry_geom_idx" which
is equal to the one of the table above and therefore not valid in postgres.
I am aware that the the maximum identifier length in postgres is 63 Bytes and therefore the index names get cutted.
But why is this change in the index naming happening between GDAL Version 3.2.1 and 3.9.1?
We don't want to downgrade to the old GDAL Version, but Is there a way to go back to the old naming system which avoided duplicated names?
Thanks and best regards
Hanskaspar Frei
