Horton

Re: PoC/WIP: Extended statistics on expressions

Tomas Vondra <tomas.vondra@enterprisedb.com>

View thread View on postgr.es
From: Tomas Vondra <tomas.vondra@enterprisedb.com>
To: Dean Rasheed <dean.a.rasheed@gmail.com>
Cc: Justin Pryzby <pryzby@telsasoft.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2021-03-24T00:51:09Z
Lists: pgsql-hackers

Attachments

Hi,

I've pushed the first two parts, dealing with composite types during
bootstrap. I've decided to commit both, including the array->list
conversion, as that makes the reloads simpler. I've made two tweaks:

1) I've renamed the function to reload_typ_list, which I think is better
(and it used to be reload_typ_array).

2) I've removed the did_reread assert. It'd allow just a single reload,
which blocks recursive composite types - seems unnecessary, although we
don't need that now. I think we can't have infinite recursion, because
we can only load types from already defined catalogs (so no cycles).


I've rebased and cleaned up the main part of the patch. There's a bunch
of comments slightly reworded / cleaned up, etc. The more significant
changes are:

1) The explanation of the example added to create_statistics.sgml was
somewhat wrong, so I corrected that.

2) I've renamed StatBuildData to StatsBuildData.

3) I've resolved the FIXMEs in examine_expression.

We don't need to do anything special about the collation, because unlike
indexes it's not possible to specify "collate" for the attributes. It's
possible to do thatin the expression, but exprCollation handles that.

For statistics target, we simply use the value determined for the
statistics itself. There's no way to specify that for expressions.

4) I've updated the comments about ndistinct estimates in selfuncs.c,
because some of it was a bit wrong/misleading - per the discussion we
had about matching stats to expressions.

5) I've also tweaked the add_unique_group_expr so that we don't have to
run examine_variable() repeatedly if we already have it.

6) Resolved the FIXME about acl_ok in examine_variable. Instead of just
setting it to 'true' it now mimics what we do for indexes. I think it's
correct, but this is probably worth a review.

7) psql now prints expressions only for (version > 14). I've considered
tweaking the existing block, but that was quite incomprehensible so I
just made a modified copy.

I think this is 99.999% ready to be pushed, so barring objections I'll
do that in a day or two.

The part 0003 is a small tweak I'll consider, preferring exact matches
of expressions over Var matches. It's not a clear win (but what is, in a
greedy algorithm), but it does improve one of the regression tests.
Minor change, though.


regards

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

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