Greetings,
* Олег Самойлов (splarv@ya.ru) wrote:
> Ah, thanks. I am not a developer of PostgreSQL. I am a developer in PostgreSQL. :) And I see two hash indexes on the
samedata and one of them 43 times bigger then other, this looked like something terribly wrong. Just free idea how to
considerablyimprove your product.
Ultimately, it might be less improvement overall and more of a
regression though, in the common cases.
This use-case isn't common for a hash index. If we changed all hash
indexes to instead be gin indexes of jsonb values, we'd end up with the
common hash index use-cases being much, much worse.
In the end, this comes down to choosing the right index for your data
and your queries, which is actually a rather difficult problem to handle
in some kind of automated way as there are lots of different trade-offs
to consider.
Thanks!
Stephen