error: Distribution kaleido==0.2.1.post1 @ registry+https://pypi.org/simple can't be installed because it doesn't have a source distribution or wheel for the current platform

[jamesdeluk]

New to uv, sorry if PEBKAC.

I'm trying to install pycaret (uv add pycaret) to a new project on Windows, Python 3.11.

Get this error:

error: Distribution `kaleido==0.2.1.post1 @ registry+https://pypi.org/simple` can't be installed because it doesn't have a source distribution or wheel for the current platform

The repo (https://pypi.org/simple/kaleido/) has:

kaleido-0.2.1-py2.py3-none-win_amd64.whl
kaleido-0.2.1.post1-py2.py3-none-manylinux2014_armv7l.whl

So, no kaleido-0.2.1.post1 for Windows.

Checking the PyPi website, there is https://pypi.org/project/kaleido/0.2.1.post1/

pip install kaleido==0.2.1.post1

Although that applies to the manylinux version.

Yet using conda, it works fine; it installs:

Downloading kaleido-0.2.1-py2.py3-none-win_amd64.whl (65.9 MB)

Is there a way for uv to use the compatible version like conda does?

[konstin]

The problem is that kaleido 0.2.1 and 0.2.1.post1 are technically different versions, and the latter has only a kaleido-0.2.1.post1-py2.py3-none-manylinux2014_armv7l.whl (and due to universal resolution, uv doesn't know which wheels it will need eventually). Setting required platforms is planned in #10067, currently you can try something like

[tool.uv]
environments = [
	"platform_system != "Windows",
	"platform_system == "Windows"
]

or banning 0.2.1.post1 in a constraint.

[jamesdeluk]

Thanks @konstin .

[tool.uv]
environments = [
	"platform_system != "Windows",
	"platform_system == "Windows"
]

The same error occurred with this (note it's "platform_system != 'Windows'" not "platform_system != "Windows" for anyone else who tried this and initially had an error).

or banning 0.2.1.post1 in a constraint.

How can I do this?

EDIT: I tried

dependencies = ["kaleido<0.2.1.post1"]

But then got a weird error:

  × Failed to build `sklearn==0.0.post12`
The 'sklearn' PyPI package is deprecated, use 'scikit-learn'

sklearn 0.0.post12 is ancient!

dependencies = ["kaleido<=0.2.1"]

Seems to be working, but it's taking ages to install - far longer than both pip and conda, making me think something isn't working correctly (warning: Failed to hardlink files; falling back to full copy. This may lead to degraded performance.)

[konstin]

With this input:

[project]
name = "foo"
version = "0.1.0"
requires-python = ">=3.11"
dependencies = [
  "pycaret",
]

[tool.uv]
constraint-dependencies = ["kaleido!=0.2.1.post1"]

I get a resolution that does not contain sklearn==0.0.post12 and - at least on linux python 3.11 - successfully installs.

Seems to be working, but it's taking ages to install - far longer than both pip and conda, making me think something isn't working correctly (warning: Failed to hardlink files; falling back to full copy. This may lead to degraded performance.)

This usually happens when the cache and the install directory aren't on the same drive, but should not have a massive impact. Usually, something really is some package not having a wheel for the target environment, and that build takes a long time. You can get more information running with -v.

[jamesdeluk]

With this input:

To confirm, you mean edit the toml then uv add pycaret? I did that and it seems to have worked - thanks! Or is there a way to "run" the toml directly, similar to building from a requirements file?

This usually happens when the cache and the install directory aren't on the same drive, but should not have a massive impact. Usually, something really is some package not having a wheel for the target environment, and that build takes a long time. You can get more information running with -v.

Turns out it was because I was working within Google Drive, so it was the annoying upload/download causing the delay.

[konstin]

Inside of docker run --rm -it -v $(pwd):/io ghcr.io/astral-sh/uv:python3.11-bookworm bash, i can create a pyproject.toml with:

[project]
name = "foo"
version = "0.1.0"
requires-python = ">=3.11"

[tool.uv]
constraint-dependencies = ["kaleido!=0.2.1.post1"]

and run

uv add pycaret

I get a lockfile with sklearn-compat 0.1.3, but no sklearn==0.0.post12.