On 26/2/2024 09:52, Andrei Lepikhov wrote:
> On 25/2/2024 20:32, Tender Wang wrote:
>> I think in prepare_probe_slot(), should called datumCopy as the
>> attached patch does.
>>
>> Any thoughts? Thanks.
> Thanks for the report.
> I think it is better to invent a Runtime Memory Context; likewise, it is
> already designed in IndexScan and derivatives. Here, you just allocate
> the value in some upper memory context.
> Also, I'm curious why such a trivial error hasn't been found for a long
> time
Hmmm. I see the problem (test.sql in attachment for reproduction and
results). We only detect it by the number of Hits:
Cache Key: t1.x, (t1.t)::numeric
Cache Mode: logical
Hits: 0 Misses: 30 Evictions: 0 Overflows: 0 Memory Usage: 8kB
We see no hits in logical mode and 100 hits in binary mode. We see 15
hits for both logical and binary mode if parameters are integer numbers
- no problems with resetting expression context.
Your patch resolves the issue for logical mode - I see 15 hits for
integer and complex keys. But I still see 100 hits in binary mode. Maybe
we still have a problem?
What's more, why the Memoize node doesn't see the problem at all?
--
regards,
Andrei Lepikhov
Postgres Professional