Обсуждение: BUG #17819: The assert failed #17334 has not been repaired

The following bug has been logged on the website:

Bug reference:      17819
Logged by:          xin wen
Email address:      xinwen@stu.scu.edu.cn
PostgreSQL version: 15.2
Operating system:   Ubuntu 20.04
Description:

The assert failed reported in #17334
(https://www.postgresql.org/message-id/17334-135f485c21739caa%40postgresql.org)
has not been repaired.
When executing the following query:

CREATE TABLE point_tbl(f1 point);
CREATE INDEX gpointind ON point_tbl USING gist(f1);
INSERT INTO point_tbl SELECT ('0,0') FROM generate_series(1, 185);
INSERT INTO point_tbl VALUES ('0,NaN');
SET enable_seqscan=off; SELECT f1 <-> point '(0,0)' AS dist FROM point_tbl
ORDER BY dist;

The FailedAssertion("box->low.y <= box->high.y", File:
"/backend/access/gist/gistproc.c") will still occur.

PG Bug reporting form <noreply@postgresql.org> writes:
> The assert failed reported in #17334
> (https://www.postgresql.org/message-id/17334-135f485c21739caa%40postgresql.org)
> has not been repaired.

If supporting NaNs in geometric types excites you, then by all means
step up and work on it, but it's not a high-priority thing for most
of us.  We've put some effort into the area before (e.g. [1]) but
arriving at self-consistent definitions is hard, and implementing
them is messy, and TBH the geometric types are a development backwater
in the first place.

At this point I'd be about ready to forbid NaNs in geometric types.
But even making that happen is work that nobody's volunteered for.

            regards, tom lane

[1] https://www.postgresql.org/message-id/flat/CAGf+fX70rWFOk5cd00uMfa__0yP+vtQg5ck7c2Onb-Yczp0URA@mail.gmail.com