What’s new in 2.3.0 (June 4, 2025)

These are the changes in pandas 2.3.0. See Release notes for a full changelog including other versions of pandas.

Upcoming changes in pandas 3.0

pandas 3.0 will bring two bigger changes to the default behavior of pandas.

Dedicated string data type by default

Historically, pandas represented string columns with NumPy object data type. This representation has numerous problems: it is not specific to strings (any Python object can be stored in an object-dtype array, not just strings) and it is often not very efficient (both performance wise and for memory usage).

Starting with the upcoming pandas 3.0 release, a dedicated string data type will be enabled by default (backed by PyArrow under the hood, if installed, otherwise falling back to NumPy). This means that pandas will start inferring columns containing string data as the new str data type when creating pandas objects, such as in constructors or IO functions.

Old behavior:

>>> ser = pd.Series(["a", "b"])
0    a
1    b
dtype: object

New behavior:

>>> ser = pd.Series(["a", "b"])
0    a
1    b
dtype: str

The string data type that is used in these scenarios will mostly behave as NumPy object would, including missing value semantics and general operations on these columns.

However, the introduction of a new default dtype will also have some breaking consequences to your code (for example when checking for the .dtype being object dtype). To allow testing it in advance of the pandas 3.0 release, this future dtype inference logic can be enabled in pandas 2.3 with:

pd.options.future.infer_string = True

See the Migration guide for the new string data type (pandas 3.0) for more details on the behaviour changes and how to adapt your code to the new default.

Copy-on-Write

The currently optional mode Copy-on-Write will be enabled by default in pandas 3.0. There won’t be an option to retain the legacy behavior.

In summary, the new “copy-on-write” behaviour will bring changes in behavior in how pandas operates with respect to copies and views.

1. The result of any indexing operation (subsetting a DataFrame or Series in any way, i.e. including accessing a DataFrame column as a Series) or any method returning a new DataFrame or Series, always behaves as if it were a copy in terms of user API.

2. As a consequence, if you want to modify an object (DataFrame or Series), the only way to do this is to directly modify that object itself.

Because every single indexing step now behaves as a copy, this also means that “chained assignment” (updating a DataFrame with multiple setitem steps) will stop working. Because this now consistently never works, the SettingWithCopyWarning will be removed.

The new behavioral semantics are explained in more detail in the user guide about Copy-on-Write.

The new behavior can be enabled since pandas 2.0 with the following option:

pd.options.mode.copy_on_write = True

Some of the behaviour changes allow a clear deprecation, like the changes in chained assignment. Other changes are more subtle and thus, the warnings are hidden behind an option that can be enabled since pandas 2.2:

pd.options.mode.copy_on_write = "warn"

This mode will warn in many different scenarios that aren’t actually relevant to most queries. We recommend exploring this mode, but it is not necessary to get rid of all of these warnings. The migration guide explains the upgrade process in more detail.

Enhancements

Other enhancements

• pandas.api.interchange.from_dataframe() now uses the PyCapsule Interface if available, only falling back to the Dataframe Interchange Protocol if that fails (GH 60739)

• The semantics for the copy keyword in __array__ methods (i.e. called when using np.array() or np.asarray() on pandas objects) has been updated to work correctly with NumPy >= 2 (GH 57739)

• Series.str.decode() result now has StringDtype when future.infer_string is True (GH 60709)

• to_hdf() and to_hdf() now round-trip with StringDtype (GH 60663)

• Improved repr of NumpyExtensionArray to account for NEP51 (GH 61085)

• The Series.str.decode() has gained the argument dtype to control the dtype of the result (GH 60940)

• The cumsum(), cummin(), and cummax() reductions are now implemented for StringDtype columns (GH 60633)

• The sum() reduction is now implemented for StringDtype columns (GH 59853)

Deprecations

• Deprecated allowing non-bool values for na in str.contains(), str.startswith(), and str.endswith() for dtypes that do not already disallow these (GH 59615)

• Deprecated the "pyarrow_numpy" storage option for StringDtype (GH 60152)

Bug fixes

Numeric

• Bug in Series.mode() and DataFrame.mode() with dropna=False where not all dtypes would sort in the presence of NA values (GH 60702)

• Bug in Series.round() where a TypeError would always raise with object dtype (GH 61206)

Strings

• Bug in Series.__pos__() and DataFrame.__pos__() where an Exception was not raised for StringDtype with storage="pyarrow" (GH 60710)

• Bug in Series.rank() for StringDtype with storage="pyarrow" that incorrectly returned integer results with method="average" and raised an error if it would truncate results (GH 59768)

• Bug in Series.replace() with StringDtype when replacing with a non-string value was not upcasting to object dtype (GH 60282)

• Bug in Series.str.center() with StringDtype with storage="pyarrow" not matching the python behavior in corner cases with an odd number of fill characters (GH 54792)

• Bug in Series.str.replace() when n < 0 for StringDtype with storage="pyarrow" (GH 59628)

• Bug in Series.str.slice() with negative step with ArrowDtype and StringDtype with storage="pyarrow" giving incorrect results (GH 59710)

Indexing

• Bug in Index.get_indexer() round-tripping through string dtype when infer_string is enabled (GH 55834)

I/O

• Bug in DataFrame.to_excel() which stored decimals as strings instead of numbers (GH 49598)

Other

• Fixed usage of inspect when the optional dependencies pyarrow or jinja2 are not installed (GH 60196)

Contributors

A total of 24 people contributed patches to this release. People with a “+” by their names contributed a patch for the first time.

• ChiLin Chiu +

• Irv Lustig

• Isuru Fernando +

• Jake Thomas Trevallion +

• Joris Van den Bossche

• Kevin Amparado +

• LOCHAN PAUDEL +

• Lumberbot (aka Jack)

• Marc Mueller +

• Marco Edward Gorelli

• Matthew Roeschke

• Pandas Development Team

• Patrick Hoefler

• Richard Shadrach

• SALCAN +

• Sebastian Berg

• Simon Hawkins

• Thomas Li

• Will Ayd

• William Andrea

• William Ayd

• dependabot[bot]

• jbrockmendel

• tasfia8 +