pandas · product momentumWe ship fast.
Here’s the proof.
Pandas now executes common operations significantly faster through targeted optimizations across data access, merging, groupby, and string handling. The library has eliminated unnecessary memory allocations, streamlined Cython code paths, and improved internal algorithms to deliver measurably quicker performance without any API changes.
Pandas has resolved dozens of edge cases and bugs that caused crashes, data corruption, or unexpected behavior in indexing, merging, groupby, datetime handling, and extension arrays. These fixes ensure reliable operations across complex scenarios like duplicate columns, MultiIndex levels, rolling calculations, and categorical data.
Pandas documentation and API have been significantly clarified with improved docstrings, better examples, and explicit explanations of parameter behavior and edge cases. Type hints have been comprehensively added across key classes, enabling better IDE support and static type checking for all pandas users.
Pandas is simplifying its API and preparing for version 3.1 by deprecating legacy features including automatic type conversions, outdated parameters, and unused methods. Users should review these deprecations and update their code to use recommended alternatives before these features are removed in future versions.
We've standardized the allow_duplicates parameter in insert and reset_index methods to use False as the default instead of an internal placeholder, making the API more consistent and easier to understand.
PR #65146We fixed an issue in our documentation build process where the PANDAS_VERSION environment variable wasn't being set correctly, which could cause build failures or inconsistent documentation generation.
PR #65281We've cleaned up internal datetime handling code to make future maintenance easier. This change has no impact on user-facing functionality.
PR #65276Fixed a crash that occurred when using the rolling method with table mode for sum, mean, median, min, or max calculations. These operations now work reliably without errors.
PR #65273We've updated our documentation build process to use Pixi for more reliable and reproducible builds. This ensures documentation is generated consistently across all environments.
PR #65075The set_eng_float_format function is now deprecated and will be removed in a future version. We recommend using alternative formatting methods for your numerical display needs.
PR #64717We've enhanced our test base class to better support testing extension behavior around copy keyword handling. This makes it easier for developers to write comprehensive tests for custom extensions.
PR #65179We've reorganized how frequency management code is structured internally to improve code maintainability. This change has no impact on user-facing functionality.
PR #65266We've improved the performance of membership checking operations by reducing unnecessary memory allocations. This change makes these operations faster and more efficient, especially when working with large datasets.
PR #64434Fixed an issue where assigning a list of tuples to an object-dtype column using loc would fail or produce incorrect results. The operation now works as expected.
PR #65264