Re: misleading error message in ProcessUtilitySlow T_CreateStatsStmt

Álvaro Herrera <alvherre@kurilemu.de>

From: Álvaro Herrera <alvherre@kurilemu.de>

To: jian he <jian.universality@gmail.com>

Cc: Tom Lane <tgl@sss.pgh.pa.us>, PostgreSQL-development <pgsql-hackers@postgresql.org>, Kirill Reshke <reshkekirill@gmail.com>

Date: 2025-08-29T12:48:21Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

CREATE STATISTICS: improve misleading error message
- cbd3732cbbaa 14.20 landed
- 4529861cc408 15.15 landed
- f30c04682cdf 16.11 landed
- f9b41f3e1ecd 17.7 landed
- 38883916e41c 18.0 landed
- 9ec0b29976b6 19 (unreleased) landed
- f9a7622483c3 15.15 landed
- f225473cbae2 19 (unreleased) landed
- e9fb314c7ad8 17.7 landed
- d84a6c3dad1f 16.11 landed
- b830f9d67463 14.20 landed
- 3eea4dc2c7cf 18.0 landed
Glossary: improve definition of "relation"
- 16a9165ce4a4 19 (unreleased) landed
Extended statistics on expressions
- a4d75c86bf15 14.0 cited

On 2025-Aug-29, jian he wrote:

> On Fri, Aug 29, 2025 at 5:46 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:

> > WFM, although I think you could shorten it to "tables, materialized
> > views, and foreign tables".  We generally expect that partitioned
> > tables are included when saying "tables", no?  I'm not dead set on
> > that either way, though.
> 
> https://www.postgresql.org/docs/current/sql-copy.html
> use "COPY TO can be used only with plain tables, not views, and does
> not copy rows from child tables or child partitions"

I'm inclined to think that we should only mention partitioned tables
specifically when they for some reason deviate from what we do for
regular tables, i.e., what Tom is saying.  I don't think we've had an
explicit, consistent rule for that thus far, so there may be places
where we fail to follow it.

Anyway, I have pushed the error message change.

> > Alternatively, maybe the right fix is to move the parse-analysis
> > work into CreateStatistics altogether.  I think there is not a
> > very good argument for ProcessUtilitySlow doing all that stuff.
> > It's supposed to mainly just be a dispatching switch(), IMO.
>
> seems doable.
> transformStatsStmt, CreateStatistics both used only twice, refactoring
> arguments should be fine.
> please check the attached POC, regress tests also added.

Yeah, I like how this turned out.  I found out this was introduced in
commit a4d75c86bf15.

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/