> Perhaps this is what you mean by "deterministic", but isn't it > possible for some collations to treat multiple byte sequences as equal > values? And those multiple byte sequences wouldn't necessarily occur > sequentially in C collation, so it wouldn't be possible to work around > that by having the grouping node use one collation but the sorting > node use the C one. > > If my memory is incorrect, then this sounds like an intriguing idea.
Yes, as per doc (https://www.postgresql.org/docs/12/collation.html#COLLATION-NONDETERMINISTIC) some collations can result in symbols(chars? codes? runes?) to be equal, while their byte representations is not. This optimisation should check for source table collation and do not change sorting collation if columns being sorted use non deterministic collation. Luckily in practice it is probably to be very rare, all builtin collations are deterministic.
