refactor: remove order by clause when counting rows to improve performance

[shogo-nakano-desu]

Purpose of this PR

To enhance performance. I have added a process to remove the order by clause in the num_items method of the Paginator trait, which is not needed to count rows.

Performance Measurement Results

Upon executing queries on test data available to me, a performance improvement of approximately three times was observed when the order by clause was removed. (Unfortunately, I cannot provide the data used here, but I hope for your understanding.) Testing environment is PostgreSQL 14.7.

[shogo-nakano-desu]

@tyt2y3
First and foremost, I would like to express my gratitude for creating and maintaining such an excellent crate. Thank you very much.

At your convenience, could you please review this PR? We are currently using sea-orm in our production environment, and the changes proposed in this PR are expected to significantly contribute to resolving performance bottlenecks. Your assistance in reviewing would be greatly appreciated. Thank you for your time and consideration.

[tyt2y3]

Thank you for the contribution. I think this is the right thing to do. Just being curious, what performance gains you saw on a typical vs worst case query?

[shogo-nakano-desu]

Thank you so much for your review.

Table info (Our testing environment)

• About 15 columns
• About 8M rows
• The field used in the ORDER BY clause does not have an index.
  ＊In our production environment, there are approximately 50M rows and the number is increasing rapidly.

Result in our testing environment

Worst case

Execution Time: 2740 ms -> 840 ms

Typycal case

In the production environment, handling a significantly larger volume of data is expected. Therefore, it is anticipated that the worst-case scenario in terms of data quantity in the test environment would be equivalent to a typical case in the production environment. As a result, we have not measured the typical case in the test environment.

[tyt2y3]

Interesting, good to know!

The field used in the ORDER BY clause does not have an index.

That's usually the culprit haha

[shogo-nakano-desu]

Now I am pushing a commit to fix Clippy error. It will be reflected when GitHub status is back to normal. https://www.githubstatus.com/

[tyt2y3]

GitHub is back now

[shogo-nakano-desu]

Great, I pushed my commits. Now I think this PR will pass CI.

[shogo-nakano-desu]

Only these lines are the main change. Others are changes to fix clippy&compile errors.

[shogo-nakano-desu]

I have to fix error: associated function accepts is never used clippy errors in src/driver/sqlx_xxx.rs
Fixed 571b51d

[tyt2y3]

Sad, those clippy warnings seem to be very confusing

[tyt2y3]

Can you reset hard this PR to 3ef07ab ? I will merge that first.
We can work on the clippy fixes later.

[shogo-nakano-desu]

Sure, that's a good idea. I will create another PR to fix clippy errors.

[shogo-nakano-desu]

Can you reset hard this PR to 3ef07ab ?

Done

[github-actions]

🎉 Released In 0.12.11 🎉

Thank you everyone for the contribution!
This feature is now available in the latest release. Now is a good time to upgrade!
Your participation is what makes us unique; your adoption is what drives us forward.
You can support SeaQL 🌊 by starring our repos, sharing our libraries and becoming a sponsor ⭐.