pg_analytics failed to create heap table from parquet table about paradedb

[philippemnoel]

Hi @DeoLeung! Thank you for reporting this, we'll get a fix going.

I'm curious, what is your use case for creating heap tables from parquet tables?

[DeoLeung]

we ingest the data into parquet table directly, and today we want to do some benchmark against heap table :) not a normal production operation

[philippemnoel]

we ingest the data into parquet table directly, and today we want to do some benchmark against heap table :) not a normal production operation

Makes sense. Are you benchmarking analytical query performance?

[DeoLeung]

yes, but not very successful today due to this issue and the slow insert from select

from our staging server, users tend to upload some tableau data, which hard to observe pg_analytics's performance gain. and I think it has some small overhead, like counting 300 row table(maybe the insert from select bug separate them into too many small parquet?), it took 10ms while heap table with a seq scan <1ms