Re: Selectivity estimation for intarray with @@

On Tue, Jul 21, 2015 at 5:14 PM, Teodor Sigaev <teodor(at)sigaev(dot)ru> wrote:

> Some notices about code:
>
> 1 Near function transformOperator() there is wrong operator name "@<"
>

Fixed.

> 2 int_query (and intquery) should be replaced to query_int to be
> consistent with actual type name. At least where it's used as separate
> lexeme.
>

Fixed.

> 3 For historical reasons @@ doesn't commutate with itself, it has a
> commutator ~~. Patch assumes that @@ is self-commutator, but ~~ will use
> 'contsel' as a restrict estimation. Suppose, we need to declare ~~ as
> deprecated and introduce commutating @@ operator.

I think deprecating ~~ is a subject of separate patch. I fixed patch
behavior in the different way. @@ and ~~ are now handled by the same
function. The function handles "var @@ const" and "const ~~ var", but
doesn't handle "const @@ var" and "var ~~ const". It determines the case
by type of variable: it should be int[].

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company