Pandas 1.5.3 Release Highlights: New Features, Bug Fixes, and Deprecations

Pandas is a Python data‑analysis library that provides fast, flexible, and expressive data structures for handling relational or labeled data.

This release (1.5.3) includes new features such as installing optional dependencies via pip extras, allowing the Index to hold NumPy numeric dtypes, a new dtype_backend parameter for returning pyarrow‑backed or NumPy‑backed nullable dtypes, and improvements to copy‑on‑write behavior when writing data.

Bug fixes include correcting overflow behavior in DataFrameGroupBy.cumsum() and DataFrameGroupBy.cumprod() without lossy conversion to float.

Backward‑incompatible API changes cover disallowing unsupported resolutions in datetime64/timedelta64 dtypes, renaming the result name of value_counts to count, prohibiting astype conversion to unsupported datetime64/timedelta64 dtypes, defaulting UTC and fixed‑offset time zones to standard library tzinfo objects, giving empty DataFrames/Series a default RangeIndex, introducing a new rendering engine for DataFrame‑to‑LaTeX, and standardising datetime parsing.

Deprecations include the removal of parsing datetime strings with system local time zones via tzlocal (use tz or tz_localize instead), deprecating the infer_datetime_format argument in to_datetime() and read_csv(), changing the behaviour of to_datetime() with the unit argument, and deprecating several index helper functions such as Index.is_boolean(), Index.is_integer(), and pandas.io.sql.execute().