pandas-dev/pandas

Overview

pandas provides labeled, relational-style data structures for Python, chiefly Series and DataFrame, and is aimed at practical data analysis and manipulation. It handles missing values across floating-point and non-floating-point data, supports automatic and explicit alignment, and covers split-apply-combine grouping, label-based slicing and boolean subsetting, merge/join, reshape and pivot, hierarchical axis labels, and time-series operations such as date-range generation, frequency conversion, moving windows, and shifting/lagging README lines 40–93.

It also includes I/O for CSV and other delimited text, Excel, SQL queries, and HDF5, with an HDFStore interface for fast save/load README lines 40–93. The library is useful to analysts, data scientists, and Python developers who need to load, clean, combine, summarize, and reshape tabular or labeled data without building those operations themselves.

Install and first steps

Install pandas with either conda or pip. conda users can install from conda-forge:

conda install -c conda-forge pandas

Or install the released package from PyPI:

pip install pandas

The released package is the normal entry point; source builds should follow the separate instructions for setting up a pandas development environment rather than treating the release commands as a source-build recipe installation section.

After installation, the official user guide is the place to start. For a minimal workflow, the README's feature list points to guides for missing-data handling, indexing, grouping, merging, reshaping, and time series; for example, open the documented pages on missing data, grouping, or CSV I/O and run the examples against your own data. The README itself does not include a complete end-to-end code walkthrough, so the precise first example should come from the stable documentation rather than an inferred snippet.

Requirements and limits

pandas requires Python and NumPy; python-dateutil is also a dependency, and tzdata is required only on Windows/Emscripten dependencies. Minimum supported versions for required, recommended, and optional dependencies are listed in the full installation instructions. Building from source has its own contributor environment steps and is not covered by the binary install commands.

The project is released under the BSD 3-Clause license license. Questions and usage help are directed to Stack Overflow and the pydata mailing list, while development work happens in the GitHub issue tracker and pandas-dev channels Getting Help. pandas is now a large, long-lived API, so version and release notes matter when code needs to keep working across upgrades.

HotRepo's take

The project appeared on HotRepo's Python daily list on 2026-11-06 with 157 stars gained and a best position of 9; that record is a single-day list snapshot, not evidence of a broader trend. The surrounding HN threads show what readers actually discuss: the pandas 2.0 thread in 2023 was dominated by comparisons with Polars—one commenter said Polars "smokes Pandas in basically every way," while the Polars author welcomed Arrow-backed pandas as better interoperability but still argued Polars remains far faster for whole queries HN discussion. A later thread about OVH and documentation hosting is about infrastructure, not pandas internals HN discussion. The useful read is that pandas is the default labeled-data layer whose ecosystem role is stable, while performance-oriented alternatives are discussed alongside it rather than as replacements for every use case.