Fix pandas version constraint to prevent numpy binary incompatibility

[KRRT7]

Problem

When installing datacompy in a fresh environment, the dependency resolver installs pandas==2.1.1 alongside numpy>=2.2.6, which causes a binary incompatibility error:

ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

This prevents any tests from running and breaks imports.

Root Cause

The current dependency specification allows pandas>=0.25,<=2.3.3, but pandas 2.1.1 has pre-compiled C extensions that are incompatible with numpy 2.2.6. Pandas needs to be at least version 2.2.0 to work with numpy 2.2+.

Changes

• Updated pandas minimum version constraint from >=0.25 to >=2.2.0 in pyproject.toml
• Removed upper bound to allow flexibility with newer pandas versions
• This ensures binary compatibility between pandas and numpy 2.x series

Testing

Verified that after this change:

• uv sync installs pandas 2.3.3 (compatible version)
• pytest --collect-only successfully collects 269 tests without errors
• No import errors when loading datacompy modules

Environment Tested

• Python: 3.10.18
• Platform: macOS (ARM64)
• Package manager: uv

[CLAassistant]

All committers have signed the CLA.

[fdosani]

@KRRT7 Thanks for flagging this and opening up a PR. I just made a new conda env using the following:

conda create -n test python=3.10 ipython uv pip conda openjdk=8

and then within the env I installed datacompy

> v pip install datacompy                                                                                                                   
Using Python 3.10.19 environment at: miniconda3/envs/dc_test
Resolved 12 packages in 4.56s
Prepared 1 package in 6.89s
Installed 12 packages in 174ms
 + datacompy==0.18.1
 + jinja2==3.1.6
 + markupsafe==3.0.3
 + numpy==2.2.6
 + ordered-set==4.1.0
 + pandas==2.3.3
 + polars==1.33.1
 + pyarrow==22.0.0
 + python-dateutil==2.9.0.post0
 + pytz==2025.2
 + six==1.17.0
 + tzdata==2025.2

It seems to resolve fine for me. I do know the error you mentioned, I've experienced this before myself. We are working on a new 1.0 release here:

datacompy/pyproject.toml

Lines 36 to 42 in cb82bda

	dependencies = [
	"jinja2>=3",
	"numpy>=2,<=2.3.3",
	"ordered-set>=4.0.2,<=4.1",
	"pandas>=2,<=2.3.3",
	"polars[pandas]>=1.19,<=1.31.1",
	]

where we are bumping up releases to the latest. not sure if you have any thoughts on this but maybe end users should be pinning on their end to ensure compatible versions? @ak-gupta @rhaffar any thoughts?

[fdosani]

Capturing this change in https://github.com/capitalone/datacompy/tree/refactor/init-setup. Appreciate the PR!