On Fri, 4 Feb 2022 at 13:07, Les <nagylzs@xxxxxxxxx> wrote: > >> >> > PostgreSQL uses seq scan for O, F, N, T letters, but it uses index scan for A, I, C letters (with the "like" query). >> >> That's interesting. >> >> Does it help if you create an additional index on relpath with the >> text_pattern_ops modifier, e.g. >> >> CREATE INDEX ... USING btree (relpath text_pattern_ops); > > > It does not help. What if you try applying the C collation to the values from the table: where fi.is_active and fi.relpath collate "C" ^@ 'A'
