feat(query): add topn runtime filter in native storage format

[sundy-li]

I hereby agree to the terms of the CLA available at: https://databend.rs/dev/policies/cla/

Summary

Summary about this PR

If we push down the sort + limit to storage level, we can have TopN opt.

But it must work with runtime predication (prewhere) + Page index optimization , so currently we only have it in native storage format.

Query select * from hits_page where URL like '%google%' order by EventTime desc limit 10; can works 5 times faster.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Comments	Updated
databend	⬜️ Ignored (Inspect)			Jan 27, 2023 at 6:54AM (UTC)

[BohuTANG]

Good idea.
Can we use OrderBy(min-max)+ Limit in reading fuse block data to reduce the block size?

[sundy-li]

Good idea. Can we use OrderBy(min-max)+ Limit in reading fuse block data to reduce the block size?

Unless we have prewhere and page index.