What’s new in 1.4.3 (June 23, 2022)

These are the changes in pandas 1.4.3. See Release notes for a full changelog including other versions of pandas.

Behavior of concat with empty or all-NA DataFrame columns

The behavior change in version 1.4.0 to stop ignoring the data type of empty or all-NA columns with float or object dtype in concat() (Ignoring dtypes in concat with empty or all-NA columns) has been reverted (GH 45637).

Fixed regressions

• Fixed regression in DataFrame.replace() when the replacement value was explicitly None when passed in a dictionary to to_replace also casting other columns to object dtype even when there were no values to replace (GH 46634)

• Fixed regression in DataFrame.to_csv() raising error when DataFrame contains extension dtype categorical column (GH 46297, GH 46812)

• Fixed regression in representation of dtypes attribute of MultiIndex (GH 46900)

• Fixed regression when setting values with DataFrame.loc() updating RangeIndex when index was set as new column and column was updated afterwards (GH 47128)

• Fixed regression in DataFrame.fillna() and DataFrame.update() creating a copy when updating inplace (GH 47188)

• Fixed regression in DataFrame.nsmallest() led to wrong results when the sorting column has np.nan values (GH 46589)

• Fixed regression in read_fwf() raising ValueError when widths was specified with usecols (GH 46580)

• Fixed regression in concat() not sorting columns for mixed column names (GH 47127)

• Fixed regression in Groupby.transform() and Groupby.agg() failing with engine="numba" when the index was a MultiIndex (GH 46867)

• Fixed regression in NaN comparison for Index operations where the same object was compared (GH 47105)

• Fixed regression is Styler.to_latex() and Styler.to_html() where buf failed in combination with encoding (GH 47053)

• Fixed regression in read_csv() with index_col=False identifying first row as index names when header=None (GH 46955)

• Fixed regression in DataFrameGroupBy.agg() when used with list-likes or dict-likes and axis=1 that would give incorrect results; now raises NotImplementedError (GH 46995)

• Fixed regression in DataFrame.resample() and DataFrame.rolling() when used with list-likes or dict-likes and axis=1 that would raise an unintuitive error message; now raises NotImplementedError (GH 46904)

• Fixed regression in testing.assert_index_equal() when check_order=False and Index has extension or object dtype (GH 47207)

• Fixed regression in read_excel() returning ints as floats on certain input sheets (GH 46988)

• Fixed regression in DataFrame.shift() when axis is columns and fill_value is absent, freq is ignored (GH 47039)

• Fixed regression in DataFrame.to_json() causing a segmentation violation when DataFrame is created with an index parameter of the type PeriodIndex (GH 46683)

Bug fixes

• Bug in pandas.eval(), DataFrame.eval() and DataFrame.query() where passing empty local_dict or global_dict was treated as passing None (GH 47084)

• Most I/O methods no longer suppress OSError and ValueError when closing file handles (GH 47136)

• Improving error message raised by DataFrame.from_dict() when passing an invalid orient parameter (GH 47450)

Other

• The minimum version of Cython needed to compile pandas is now 0.29.30 (GH 41935)

Contributors

A total of 18 people contributed patches to this release. People with a “+” by their names contributed a patch for the first time.

• Alex-Blade +

• Andrew Hawyrluk

• JHM Darbyshire

• Jeff Reback

• Joris Van den Bossche

• Marc Garcia

• Marco Edward Gorelli

• Matthew Roeschke

• MeeseeksMachine

• Pandas Development Team

• Patrick Hoefler

• Richard Shadrach

• Robert de Vries

• Simon Hawkins

• Thomas Li

• Tim Swast

• Wenjun Si

• Yuanhao Geng