[BUG FIX] In-place updates with loc or iloc don't work correctly when the LHS has more than one column

[skirui-source]

Fixes: #7377

This PR enables to setitem using a scalar value, dataframe or array/list iterable in both DataframeLocIndexer and DataFrameIlocIndexer . Only the following cases are currently supported in cudf:

• Scalar value: follows the original code path, assigns column- values via specified key (row-label)
• Dataframe : checks for column-alignment in LHS and RHS, then uses a scatter map of the indices to assign column-values accordingly. Substitute NA for columns not found in the RHS
• All other cases (array, list, range value, etc) : first conversion to cupy array followed by special handling:
  • If 2d array: If the inner dimension is 1, it's broadcastable to all columns of the dataframe.
  • Otherwise the value must be a 1d array (scalar values are handled in case 1 above), there are 2 subcases:
    • If the key on column axis is a scalar, meaning the user is indexing a single column; Therefore 1d value should assign along the columns.
    • Otherwise, the key on column axis is a 1d array. In this case, the key on row axis can be a scalar or 1d and in both cases of row key, the ith element in value corresponds to the ith row in the indexed object. If the key is 1d, a broadcast will happen.

[github-actions]

This PR has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this PR if it is no longer required. Otherwise, please respond with a comment indicating any updates. This PR will be labeled inactive-90d if there is no activity in the next 60 days.

[shwina]

This looks great to me! There's a lot that can generally be improved in the indexing code but very much out of scope for this PR.

Appreciate the attention to testing here 👍 Nice work @skirui-source and @isVoid!

[galipremsagar]

I went ahead and pushed a commit to internalize the shape_mismatch_error to _shape_mismatch_error. Other than that this PR looks good to go. 🎉

Dismissing since the review has been addressed.

[skirui-source]

@gpucibot merge