Complete / integrate sqlite sqllogictest test scripts integrattion

[alamb]

Is your feature request related to a problem or challenge?

Part of #13811

@Omega359 has prototyped integration with the sqlite test suite ❤️

This ticket tracks the remaining work to get this runner merged into datafusion main.

Describe the solution you'd like

I would like:

1. The runner merged into datafusion main
2. Clear instructions on how to run the suite

Bonus points would be some way to regularly run these tests

Describe alternatives you've considered

From #13470 (comment)

I've updated my test branch with the latest changes including adding skipif for Postgresql queries that fail to run. What is left is at least 19 legitimate query result mismatches where tickets were filed.

I've submitted a PR to the sqllogictest-rs repo for my changes but so far they haven't responded. If they do not get around to merging it for whatever reason we have two choices:

1. Rewrite the .slt files to be columnwise vs valuewise and change ordering to only be rowsort and remove the valuesort. I've actually written a lua script to handle the valuewise -> columnwise conversion, the second part will require some more work.
2. Fork the repo and use that.

For now I'm going to slow down work on this.

Additional context

No response

[alamb]

@Omega359 do I have this work described correctly?

[Omega359]

Thanks for putting this together @alamb. Yes, the work is described correctly. CI builds currently are failing because of a clippy warning that I need to look into but otherwise it's good for others to look at.

I find that, at least on my machine, running the sqlite tests in 'normal' test mode takes ...forever. I typically run now with --release. Slower to build, faster overall.

cargo test --release --test sqllogictests -- --include-sqlite

if you do not run in release mode you may need to increase the stack size or it'll fail part way in:

export RUST_MIN_STACK=3048576

[Omega359]

One issue that I haven't fixed is that some of the sqlite test files are LARGE. In total the files are 763M. I am unsure that we would want them in the main repo ... and I am pretty sure they will fail a CI check with some being > 1MB anyways.

Perhaps we need a new datafusion_testing repository where these files would go and be linked into df via .gitmodules

[alamb]

Perhaps we need a new datafusion_testing repository where these files would go and be linked into df via .gitmodules

I think this is a good idea -- or maybe a script that checks out the original sqllite files or something

I also think we need to find a way to run it regularly (but likely it can't be on all PRs), see

• Improve efficiency of CI checks (so we can add MORE!) #13845

@wiedld from InfluxData expressed interest in helping move this project along. I will let her comment on what she next plans, etc

[Omega359]

Yes, this seems like a perfect thing for a nightly ci run.

The thing that is holding me back atm is the sqllogictest-rs changes. Without those in the official repo means either redoing the output of the sqlite results (possible but not particularly easy) or using my fork. I'm hoping to get back to seeing if I can do the rewrite this weekend by comparing the sortmode results during a complete run (with my super hacked up version of sqllogictest-rs) and if they are the same write it out as columnmode. That would mean that the current runner would work.

I don't think attempting to pull in the original sqlite files will work as what I am producing is very different from what they have. datafusion_testing repo I think would be best.