Conversation
|
@TomAugspurger what is the right behavior for this: |
|
For end-users doing something like a = pd.DataFrame(columns=['a', 'b'])
b = pd.DataFrame(columns=['b', 'a'])
dd.concat([a, b])we want that warning to come through, to match pandas. Internal to dask.dataframe, we should ensure that we either have aligned columns, or specify sort if we rely on the output order. I can try to fix all these warnings this afternoon. |
This is important for irfft2 dtype analysis
| - *test_and_lint | ||
| - *no_coverage | ||
| - *no_optimize | ||
| - *no_imports |
There was a problem hiding this comment.
Does anyone mind me dropping this CI entry as we add one for 3.7? cc @jakirkham @jcrist
There was a problem hiding this comment.
I do. We generally try to support both Python 2/3 in our codebase as there are inevitably scientific libraries that we want to pull in that haven't made the switch. Though I think this has gotten better in the past year or two, it's still a surprisingly common problem (even with new codes).
There was a problem hiding this comment.
There is still another 2.7 entry. We used to have two. You might want to check the full file in this branch
There was a problem hiding this comment.
@jakirkham does your concern still hold, given that there is another Python 2.7 entry?
There was a problem hiding this comment.
Oops. Sorry. Missed the other one. This is fine.
|
OK, merging this in. |
flake8 dask