notes_on_dshash.md

Notes on dshash

When to use it?

Unlike dynahash, dshash is always in shared memory (dynahash can be put in shared memory) it allows the hashtable to grow (though be careful, it can't shrink).

Differences with dynahash

While the behaviours below are documented in dshash, it's worth highlighting them as they are notably different from those of dynahash.

Sequence iteration

dshash_seq_term must be called when the scan is finished (in dynahash, its hash_seq_term counterpart must only be called if the scan is abandoned before the termination of the scan).

Note

Postgres versions before 15 do not have support for sequential scans. Current version's comments suggest that it still doesn't: "Future versions may support iterators" but this appears to be untrue considering presence of dshash_seq_ family of functions

Release entry locks

When you are done with an entry returned by dshash_find or similar functions, make sure you release the lock held by entry using dshash_release_lock.