Support construction of a DataFrame from a Mapping

[mrkn]

This pull-request adds support for using a Mapping in the construction of a DataFrame. By applying this change, the following Julia code will work.

julia> using PythonCall
julia> pd = pyimport("pandas")
julia> df = pd.DataFrame(Dict("a" => [1, 2, 3], "b" => [4, 5, 6]))
Python:
   b  a
0  4  1
1  5  2
2  6  3

I have confirmed that there is no performance degradation when constructing a DataFrame from a dict.

$ asv continuous -E virtualenv -b ^frame_ctor.FromDict origin/main HEAD
· Creating environments
· Discovering benchmarks
·· Uninstalling from virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
·· Installing 684a22fb <support_mapping_in_dataframe> into virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
· Running 16 total benchmarks (2 commits * 1 environments * 8 benchmarks)
[ 0.00%] · For pandas commit 2aa155ae <main> (round 1/2):
[ 0.00%] ·· Building for virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[ 0.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[ 3.12%] ··· Running (frame_ctor.FromDicts.time_dict_of_categoricals--)........
[25.00%] · For pandas commit 684a22fb <support_mapping_in_dataframe> (round 1/2):
[25.00%] ·· Building for virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[25.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[28.12%] ··· Running (frame_ctor.FromDicts.time_dict_of_categoricals--)........
[50.00%] · For pandas commit 684a22fb <support_mapping_in_dataframe> (round 2/2):
[50.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[53.12%] ··· frame_ctor.FromDicts.time_dict_of_categoricals                                327±8μs
[56.25%] ··· frame_ctor.FromDicts.time_list_of_dict                                    16.8±0.09ms
[59.38%] ··· frame_ctor.FromDicts.time_nested_dict                                      16.1±0.2ms
[62.50%] ··· frame_ctor.FromDicts.time_nested_dict_columns                              16.3±0.3ms
[65.62%] ··· frame_ctor.FromDicts.time_nested_dict_index                                13.2±0.1ms
[68.75%] ··· frame_ctor.FromDicts.time_nested_dict_index_columns                       12.9±0.08ms
[71.88%] ··· frame_ctor.FromDicts.time_nested_dict_int64                                29.1±0.1ms
[75.00%] ··· frame_ctor.FromDictwithTimestamp.time_dict_with_timestamp_offsets                  ok
[75.00%] ··· ======== ============
              offset
             -------- ------------
              <Nano>   7.97±0.1ms
              <Hour>   10.9±0.2ms
             ======== ============

[75.00%] · For pandas commit 2aa155ae <main> (round 2/2):
[75.00%] ·· Building for virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[75.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[78.12%] ··· frame_ctor.FromDicts.time_dict_of_categoricals                                330±2μs
[81.25%] ··· frame_ctor.FromDicts.time_list_of_dict                                     17.2±0.3ms
[84.38%] ··· frame_ctor.FromDicts.time_nested_dict                                     16.2±0.08ms
[87.50%] ··· frame_ctor.FromDicts.time_nested_dict_columns                              16.2±0.1ms
[90.62%] ··· frame_ctor.FromDicts.time_nested_dict_index                                13.2±0.2ms
[93.75%] ··· frame_ctor.FromDicts.time_nested_dict_index_columns                        13.5±0.2ms
[96.88%] ··· frame_ctor.FromDicts.time_nested_dict_int64                                29.7±0.3ms
[100.00%] ··· frame_ctor.FromDictwithTimestamp.time_dict_with_timestamp_offsets                  ok
[100.00%] ··· ======== =============
               offset
              -------- -------------
               <Nano>   8.16±0.07ms
               <Hour>   11.2±0.06ms
              ======== =============


BENCHMARKS NOT SIGNIFICANTLY CHANGED.

• partially closes ENH: generalize __init__ on a dict to abc.collections.Mapping and __getitem__ on a list to abc.collections.Sequence #58803 (only dataframe construction part)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

[twoertwein]

Looks good to me! Can you please update the documentation of DataFrame.__init__ (around line 517) to mention Mapping instead of dict?

It might be good to check whether DataFrame.from_dict can also accept a mapping.

[mrkn]

@twoertwein I noticed that creating a DataFrame from a Mapping of Mappings wasn't covered, so I added support for it. Could you please review this new change?

[twoertwein]

Looks good to me!

(DataFrame.from_dict is documented to allow also array-like types that contain Mappings but that can be a separate PR, pandas-dev/pandas-stubs#928 and pandas-dev/pandas-stubs#929)

[phofl]

Please wait until the discussion on the issue reaches consensus

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this.

[mroeschke]

Thanks for the PR here, but let's centralize the discussion in the issue regarding whether pandas should make this more general. Closing for now, but we can reopen if there is consensus to change in the issue