Re: PoC/WIP: Extended statistics on expressions

On 3/25/21 1:30 AM, Justin Pryzby wrote:
> On Thu, Mar 25, 2021 at 01:05:37AM +0100, Tomas Vondra wrote:
>> here's an updated patch. 0001 should address most of the today's review
>> items regarding comments etc.
>
> This is still an issue:
>
> postgres=# CREATE STATISTICS xt ON a FROM t JOIN t ON true;
> ERROR: schema "i" does not exist
>

Ah, right. That's a weird issue. I was really confused about this,
because nothing changes about the grammar or how we check the number of
relations. The problem is pretty trivial - the new code in utility.c
just grabs the first element and casts it to RangeVar, without checking
that it actually is RangeVar. With joins it's a JoinExpr, so we get a
bogus error.

The attached version fixes it by simply doing the check in utility.c.
It's a bit redundant with what's in CreateStatistics() but I don't think
we can just postpone it easily - we need to do the transformation here,
with access to queryString. But maybe we don't need to pass the relid,
when we have the list of relations in CreateStatsStmt itself ...

regards

--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company