-
-
Notifications
You must be signed in to change notification settings - Fork 17.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
INFO: Please pin your numpy < 2 if installing pandas via pip #55519
Comments
This shouldn't be necessary anymore. The next release of pandas (pandas 2.2.0) will be compatible with numpy 2. |
Ugh, it looks like the numpy 2.0 rc is delayed until February, so we won't be able to have numpy 2.0 compat until pandas 2.2.1. |
Looks like the numpy v2.0.0.b1 release candidate was made available 5 days ago, so compatibility didn't make it into pandas 2.2.1 (which came out 3 weeks ago). Does this mean that the aim is now that pandas 2.2.2 be the first one to be compatible with 2.2.2? (P.S. Thanks for all your work!) |
We are basically ready for numpy 2.0's release (as in no code changes are required for us to be compatible). However, since we use numpy's C API, we must compile against numpy 2.0rc1 or later to be compatible with numpy 2.0. This is because numpy 2.0rc1 is the first numpy pre-release to be ABI stable. Since numpy 2.0's development timeline has shifted a bunch, I am currently planning on just releasing bug fix updates of pandas on the regular cadence, but I will do a special bug fix release of pandas 2.2.x when numpy 2.0rc1 does come out. (so pandas 2.2.2 will not necessarily support numpy 2.0, but some pandas 2.2.x will) |
Just a quick update, we're targeting pandas 2.2.2 for numpy 2.0 compatibility. Expect a release in the coming weeks (probably next week, sometimeish) |
It'd be really helpful to do this as soon as is possible/reasonable for you, because Pandas is a dependency of so many other packages. There's a very long chain/tail of package releases that need to be made for numpy 2.x compatibility, so I'm hoping the key packages in the PyData stack don't already take weeks to do their releases. |
pandas 2.2.2 is out! |
I'm running pandas 2.2.2 and I still get:
The documentation says it should be compatible; maybe it's an issue with pip? |
Can you open a new issue? There's a lot of potential factors at play here (e.g. this could be coming from another library as well). |
For pandas versions before 2.2, pin numpy to version 1 [1]. This commit should fix CI for Python 3.9. [1] pandas-dev/pandas#55519
numpy 2.0 is the next major release of numpy (scheduled to come out sometime in December), and as such contains several breaking API changes.
Because of this, as of right now, there is NO released pandas version compatible with numpy 2.0.
We are currently aiming to be compatible with the release of pandas 2.2 (the next minor version) for pandas, however this is not guaranteed (we are still working through a couple of issues).
So, to avoid issues when installing pandas with pip, you should pin numpy < 2.
Note: This does not affect conda users, as conda has a special mechanism to retroactively pin numpy on older released pandas versions.
The text was updated successfully, but these errors were encountered: