* Merlin Moncure (mmoncure@gmail.com) wrote:
> Yeah -- the most common case I see is outlier culling where several
> repeated low non-deterministic selectivity quals stack reducing the
> row count estimate to 1. For example:
> SELECT * FROM foo WHERE length(bar) <= 1000 AND length(bar) >= 2;
This is exactly the issue that I've seen also- where we end up picking a
Nested Loop because we think only one row is going to be returned and
instead we end up getting a bunch and it takes forever.
There was also some speculation on trying to change plans mid-stream to
address a situation like that, once we realize what's happening. Not
sure that's really practical but it would be nice to find some solution.
Thanks,
Stephen