REGR: Fix fastparquet 0.7.0 not being able to read a parquet file

[lithomas1]

• closes BUG: pandas 1.3.2 incompatible with fastparquet 0.7.0  #43075
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

My bad on this one, pinning one of the CI to ensure that this works.

[lithomas1]

Well this does work as expected, but I can't pin CI to fastparquet 0.7.0(I forgot that it breaks tests). I'll figure something out tomorrow(probably xfailing the tests).

[alimcmaster1]

Well this does work as expected, but I can't pin CI to fastparquet 0.7.0(I forgot that it breaks tests). I'll figure something out tomorrow(probably xfailing the tests).

Do you want to merge master? Which tests need xfailing - not sure I understand why? thanks!

[lithomas1]

Just rebased. (Sorry for letting this one stale out. I completely forgot about it).

Well this does work as expected, but I can't pin CI to fastparquet 0.7.0(I forgot that it breaks tests). I'll figure something out tomorrow(probably xfailing the tests).

Do you want to merge master? Which tests need xfailing - not sure I understand why? thanks!

Sorry for being vague here. If we want to pin one of the CI's to make sure that this patch works, we'll have to xfail some tests. We need to do this, because fastparquet 0.7.0 broke some tests, and we didn't xfail them yet(Currently, CI is on 0.7.1 which fixed those tests and is green, so no-one bothered to xfail them for 0.7.0).

The other option is accepting this patch untested and I've verified it manually.

Thanks for the review, and sorry about causing the regression.

[alimcmaster1]

Looks like the code checks are failing please can you update the whatsnew

/home/runner/work/pandas/pandas/doc/source/whatsnew/v1.3.3.rst:36: WARNING: Bullet list ends without a blank line; unexpected unindent.
looking for now-outdated files... none found

[alimcmaster1]

Just rebased. (Sorry for letting this one stale out. I completely forgot about it).

Well this does work as expected, but I can't pin CI to fastparquet 0.7.0(I forgot that it breaks tests). I'll figure something out tomorrow(probably xfailing the tests).

Do you want to merge master? Which tests need xfailing - not sure I understand why? thanks!

Sorry for being vague here. If we want to pin one of the CI's to make sure that this patch works, we'll have to xfail some tests. We need to do this, because fastparquet 0.7.0 broke some tests, and we didn't xfail them yet(Currently, CI is on 0.7.1 which fixed those tests and is green, so no-one bothered to xfail them for 0.7.0).

The other option is accepting this patch untested and I've verified it manually.

Thanks for the review, and sorry about causing the regression.

Understood thank you. Yes agree think we are unlikely to ever run CI against fastparquet 0.7.0 now that 0.7.1 is out and fixes the regressions. The usually pattern when making a change like this would be to pin a build temporarily, like you did, post a link on the PR of the desired tests passing in the pipeline then undo the pin.

Happy with the current solution - ping on green. @lithomas1

[jreback]

thanks @lithomas1

[jreback]

@meeseeksdev backport 1.3.x

[lumberbot-app]

Something went wrong ... Please have a look at my logs.