Hey everyone !
Bumping an older thread, I've read the GIN readme in the code base and have skimmed through the implementation, it made me wonder -
would it be possible to use the index to query for distinct / count distinct values, at least for some key types ?
For instance, say I have a GIN index on a single, highly cardinal but non-unique text column (a-la 'name'); from my very limited understanding, would it be possible to query for distinct / count distinct values via roughly:
- Traversing the GIN entry tree tuples
- Gathering the key data from said tuples
- Discarding keys with no / empty posting lists, as they aren't discarded from the tree
- Traversing the pending list for indices with fastupdate & merging the result set
I'm probably off by quite a lot, but I'd really appreciate your great insight on the above.
Thanks !
Danny
