Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Test numeric edge case for repartition with npartitions. #5433

Merged

Conversation

@amerkel2
Copy link
Contributor

amerkel2 commented Sep 25, 2019

We had an issue with Dask 1.2.0 where repartitioning a certain number of source partitions into a certain number of target partitions led to one partition missing from the result. In the end, we found that this is because of a numeric instability when determining divisions (dividing and multiplying again with the same number might not always yield exactly the original number, but slightly less, which causes a problem here: https://github.com/dask/dask/blob/master/dask/dataframe/core.py#L5516).

Example:

>>> (15/11)*11
14.999999999999998
>>> int((15/11)*11)
14

With Dask >=2, the problem was fixed by introducing this check: c4b6770#diff-492da7893fa5fa6ad3a6ef6cdef985f2R4821

Nevertheless, it would be safer to preclude any chance of the same bug being introduced again by adding a test, as done by this PR.

  • [ x ] Tests added / passed
  • [ x ] Passes black dask / flake8 dask
@mrocklin mrocklin merged commit 2817533 into dask:master Sep 25, 2019
2 checks passed
2 checks passed
continuous-integration/appveyor/pr AppVeyor build succeeded
Details
continuous-integration/travis-ci/pr The Travis CI build passed
Details
@mrocklin

This comment has been minimized.

Copy link
Member

mrocklin commented Sep 25, 2019

Thanks for the contribution @amerkel2 . Adding this test seems reasonable to me :)

@amerkel2 amerkel2 deleted the amerkel2:contribution/test_repartition_numeric_edge_case branch Sep 26, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
2 participants
You can’t perform that action at this time.