Re: Optimizing nbtree ScalarArrayOp execution, allowing multi-column ordered scans, skip scan

On Mon, Apr 22, 2024 at 4:00 AM Alexander Lakhin <exclusion@gmail.com> wrote:
> Please look at another case, where a similar Assert (but this time in
> _bt_preprocess_keys()) is triggered:
> CREATE TABLE t (a text, b text);
> INSERT INTO t (a, b) SELECT 'a', repeat('b', 100) FROM generate_series(1, 500) g;
> CREATE INDEX t_idx ON t USING btree(a);
> BEGIN;
> DECLARE c CURSOR FOR SELECT a FROM t WHERE a = 'a';
> FETCH FROM c;
> FETCH RELATIVE 0 FROM c;
>
> TRAP: failed Assert("so->numArrayKeys"), File: "nbtutils.c", Line: 2582, PID: 1130962

I'm pretty sure that I could fix this by simply removing the
assertion. But I need to think about it a bit more before I push a
fix.

The test case you've provided proves that _bt_preprocess_keys's
new no-op path isn't just used during scans that have array keys (your
test case doesn't have a SAOP at all). This was never intended. On the
other hand, I think that it's still correct (or will be once the assertion is
gone), and it seems like it would be simpler to allow this case (and
document it) than to not allow it at all.

The general idea that we only need one "real" _bt_preprocess_keys call
per btrescan (independent of the presence of array keys) still seems
sound.

--
Peter Geoghegan