While hll_hash_timestamptz wouldn't be terribly useful, hll_hash_date certainly could be. And it shouldn't be hard to implement; you can transform a date to an INT. However, since the range of INTs for a valid date would be different from 0-MAXINT, should we adjust the hash function somehow?
As long as you can get it to an integer, it should be safe to hash. Consecutive ranges of the integer space should not have worse hashing properties than a random subset of the space.
While hll_hash_timestamptz wouldn't be terribly useful, hll_hash_date certainly could be. And it shouldn't be hard to implement; you can transform a date to an INT. However, since the range of INTs for a valid date would be different from 0-MAXINT, should we adjust the hash function somehow?