Bug Report: Missing Support for Latest NumPy Versions (>=2.2 or 2.3.x)

[xgboosted]

Current Behaviour

ydata-profiling cannot be installed alongside the latest versions of numpy (>=2.2, 2.3.x) and pandas (>=2.3.x) due to strict dependency requirements. The current release restricts numpy to <2.2, while pandas 2.3.x requires numpy>=1.26.0, and the latest numpy is 2.3.3.

Expected Behaviour

ydata-profiling should support the latest numpy and pandas versions so it can be used in modern data science environments.

This issue blocks adoption in projects that require up-to-date scientific Python stacks. Please update ydata-profiling to support the latest numpy releases (>=2.2, including 2.3.x) and ensure compatibility with current pandas versions.

Thank you!

Data Description

N/A

Code that reproduces the bug

Steps to reproduce the behavior:
1. Create a new Python environment
2. Run:
   
   pip install numpy==2.3.3 pandas==2.3.3 ydata-profiling==4.17.0
   
3. Observe the dependency conflict error:
   
   ERROR: Cannot install -r requirements.txt (line 21), -r requirements.txt (line 3) and numpy==2.3.3 because these package versions have conflicting dependencies.
   The conflict is caused by:
       The user requested numpy==2.3.3
       pandas 2.3.3 depends on numpy>=1.26.0; python_version >= "3.12"
       ydata-profiling 4.17.0 depends on numpy<2.2 and >=1.16.0

pandas-profiling version

4.17.0

Dependencies

- Python Version: 3.12.10
 - ydata-profiling Version: 4.17.0
 - numpy Version: 2.3.3
 - pandas Version: 2.3.3

OS

Ubuntu 24.04

Checklist

• There is not yet another bug report for this issue in the issue tracker
• The problem is reproducible from this bug report. This guide can help to craft a minimal bug report.
• The issue has not been resolved by the entries listed under Common Issues.

[jack-mcivor]

Why so many upper-bounds? I see from this discussion there was plans to remove them? #1293 (comment)

You may have seen, but there is a strong argument for not including upper-bound constraints here https://iscinumpy.dev/post/bound-version-constraints/

[xgboosted]

That was a good decision to not include the upper bound for key Python packages, but then why was it not implemented yet?

numpy, scipy, pandas, matplotlib; these 4 are the essential libs that should have the bleeding edge support in every aspiring Python project.

[SathvikBRao108]

Facing Issue ,where langchain-aws and ydata-profiling cannot be downloaded together due to numpy version issues

[fabclmnt]

Hi,

thank you for your feedback. You’re always welcome to support the project and contribute. 🙂

Regarding versions of important libraries like numpy, scipy, pandas, and matplotlib: while these are essential, they also introduce frequent breaking changes that can affect the project’s stability. Since this is an open-source solution, updates happen as often as the community contributes.

For transparency, the NumPy update was also blocked by a dependency issue: upgrading numpy beyond 2.2 requires upgrading Numba as well, but the corresponding numba versions (around 0.60–0.61) had compatibility bugs. Until those issues were resolved upstream, moving to the newer numpy versions wasn’t safe for the project.

The good news is that we are finally able to upgrade numpy, and a new version will be made available soon.

[fabclmnt]

#1790