Horton

Re: PoC/WIP: Extended statistics on expressions

Justin Pryzby <pryzby@telsasoft.com>

View thread View on postgr.es
From: Justin Pryzby <pryzby@telsasoft.com>
To: Tomas Vondra <tomas.vondra@enterprisedb.com>
Cc: Dean Rasheed <dean.a.rasheed@gmail.com>, hackers <pgsql-hackers@postgresql.org>
Date: 2021-08-24T13:13:20Z
Lists: pgsql-hackers 
On Mon, Aug 16, 2021 at 05:41:57PM +0200, Tomas Vondra wrote:
> > This still seems odd:
> > 
> > postgres=# CREATE STATISTICS asf ON i FROM t;
> > ERROR:  extended statistics require at least 2 columns
> > postgres=# CREATE STATISTICS asf ON (i) FROM t;
> > CREATE STATISTICS
> > 
> > It seems wrong that the command works with added parens, but builds expression
> > stats on a simple column (which is redundant with what analyze does without
> > extended stats).
> 
> Well, yeah. But I think this is a behavior that was discussed somewhere in
> this thread, and the agreement was that it's not worth the complexity, as
> this comment explains
> 
>   * XXX We do only the bare minimum to separate simple attribute and
>   * complex expressions - for example "(a)" will be treated as a complex
>   * expression. No matter how elaborate the check is, there'll always be
>   * a way around it, if the user is determined (consider e.g. "(a+0)"),
>   * so it's not worth protecting against it.
> 
> Patch 0001 fixes the "double parens" issue discussed elsewhere in this
> thread, and patch 0002 tweaks CREATE STATISTICS to treat "(a)" as a simple
> column reference.

0002 refuses to create expressional stats on a simple column reference like
(a), which I think is helps to avoid a user accidentally creating useless ext
stats objects (which are redundant with the table's column stats).

0002 does not attempt to refuse cases like (a+0), which I think is fine:
we don't try to reject useless cases if someone insists on it.
See 240971675, 701fd0bbc.

So I am +1 to apply both patches.

I added this as an Opened Item for increased visibility.

-- 
Justin

Commits

  1. Disallow extended statistics on system columns

  2. Identify simple column references in extended statistics

  3. Don't print extra parens around expressions in extended stats

  4. Change position of field "transformed" in struct CreateStatsStmt.

  5. Add transformed flag to nodes/*funcs.c for CREATE STATISTICS

  6. Stabilize stats_ext test with other collations

  7. Extended statistics on expressions

  8. Reduce duration of stats_ext regression tests

  9. Allow composite types in catalog bootstrap

  10. Convert Typ from array to list in bootstrap

  11. Disallow CREATE STATISTICS on system catalogs