Allow cudf.concat to process a dictionary of frames #15115

[amanlai]

Closes #15115.

Unlike pandas.concat, cudf.concat doesn't work with a dictionary of objects. The following code raises an error.

d = {
    'first': cudf.DataFrame({'A': [1, 2], 'B': [3, 4]}),
    'second': cudf.DataFrame({'A': [5, 6], 'B': [7, 8]}),
}

cudf.concat(d, axis=1)
cudf.concat(d)

This commit resolves this issue.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[shwina]

/ok to test

[shwina]

/ok to test

[shwina]

/ok to test

[shwina]

Can we parametrize to include a few additional test cases?

• A single entry in d
• A third entry in d with a different number of elements, e.g.,:

  d = {
      'first': pd.DataFrame({'A': [1, 2], 'B': [3, 4]}),
      'second': pd.DataFrame({'A': [5, 6], 'B': [7, 8]}),
      'third': pd.DataFrame({'C': [1, 2, 3]})
  }

[amanlai]

Aren't single entry in d and a third entry in d with just 1 key-value pair the same thing? Are you saying there should be a test for d = {'first': cudf.DataFrame({'A': [1, 2], 'B': [3, 4]})}?

[shwina]

Sorry - I mistyped! I've edited my comment

[shwina]

Should we also add tests for concatenating dict-of-Series and dict-of-Index objects?

If those cases don't work currently - that's OK. We should at least raise a NotImplementedError.

[amanlai]

I can add tests for these.

[shwina]

Overall, this is looking pretty good! Just left a comment asking for more tests.

[shwina]

Looks like the PR is failing style checks. Can you run locally:

pre-commit run --all

and fix up any style issues reported?

[amanlai]

@shwina I screwed up and deleted the branch by mistake which I think closed the PR. How do I start this up again?

[shwina]

Here's what I would do; the following creates a local branch called fea-concat-dict-axis-1 starting from branch-24.04, and applies your two commits on top of it:

git checkout branch-24.04
git checkout -b fea-concat-dict-axis-1
git cherry-pick --ff b16713a4022487af9e8cba162d2da4fc20309cda
git cherry-pick --ff f7ed7f98e7e4a3d27565246d8c5cb4c09fb099d7

Then you can push the branch fea-concat-dict-axis-1 and create a PR from it.

Does that help? Feel free to ask if you have any questions!

[amanlai]

@shwina just to clarify, do I create a new pull request?

[shwina]

Yes I think that's best

[amanlai]

@shwina I ended up making a new pull request at #15160.