Faster query with a simpler plan by joining with the CTE instead of making IN query

[ddorian]

I set a benchmark doing 200+ updates/second on the queue (using the current behaviour), and ran both queries 1 after the other in a separate session to get their plans. My change removes the HashAggregate. The same plan was with LIMIT 10.

Using PostgreSQL 15.4 on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0, 64-bit.

limit_1_before.sql.txt
limit_1_after.sql.txt

The limit 10 was done with a bigger queue doing 40 read/s of batch_size=100.

limit_10.sql.txt

[ChuckHend]

Thanks @ddorian - I'll take a look in next couple days so we can keep this moving forward.

[ChuckHend]

👍

[ChuckHend]

@ddorian , can you bump the version in ./core/Cargo.toml? This will get the last CI check to pass and publish the release to that core crate.