Andres Freund <andres(at)2ndquadrant(dot)com> writes:
> Craig just mentioned in an internal chat that there's no btree or even
> hash opclass for the new pg_lsn type. That restricts what you can do
> with it quite severely.
> Imo this should be fixed for 9.4 - after all it was possible unto now to
> index a table with lsns returned by system functions or have queries
> with grouping on them without casting.
Sorry, it is *way* too late for 9.4.
regards, tom lane