Skip to content

Improve speed of datafusion::fuzz fuzz_cases::pruning::test_fuzz_utf8 test #13946

@alamb

Description

@alamb

After #12978 the test_fuzz_utf8 test takes over a minute to run on my machine

This ticket tracks improving the speed

          > Thanks @adriangb -- I think this PR is ready to go

One thing I noticed is that the fuzz test takes over a minute on my machine:

        SLOW [> 60.000s] datafusion::fuzz fuzz_cases::pruning::test_fuzz_utf8
        PASS [  65.772s] datafusion::fuzz fuzz_cases::pruning::test_fuzz_utf8
------------
     Summary [  72.749s] 47 tests run: 47 passed (1 slow), 0 skipped
andrewlamb@Mac:~/Software/datafusion$

Is there some way to make it faster? Maybe with multiple threads or crank down the number of things to teset?

Yeah this is what I was hinting at in #12978 (comment).

I'm happy to throw threads at it for a start, and restricting the search space might be necessary but I think requires a more careful eye to minimize how much valuable testing is discarded. The other thing that I think we can do is speed up the tests themselves, in particular minimizing unnecessary round trips to Parquet, but I'm not sure where the right places to hook in would be that still give us a realistic test but remove the need to re-parse the same data over and over again.

Originally posted by @adriangb in #12978 (comment)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions