Python client fails to install via pip about duckdb HOT 17 CLOSED

Interestingly, it does appears that my system is being recognized as macOS 10.16 instead of the expected macOS 14.x, which is likely leading to the compatibility issue.

> python -c "import platform; print(platform.platform())"

macOS-10.16-x86_64-i386-64bit

I gave up on pip and successfully installed via conda with the following commands:

conda update conda
conda update --all
conda install -c conda-forge duckdb

Thanks for both of your help!

from duckdb.

My listed pip version is also 24.0
It seems related to virtual envs? pypa/pip#9621
But they mention that the pip version has something to do with it, which can't be the case

Also seems somewhat helpful:
https://stackoverflow.com/questions/77005016/how-to-update-the-compatible-tags-in-pip-in-conda-environments

from duckdb.

Can you share which version of Python you are compiling for?

I would assume by "it gets stuck ... indefinitely" you mean that building from source takes up to 20 minutes (that is a lot).

It's possibly you might have Python 3.7 (that we don't package wheels for on OSX), that according to https://devguide.python.org/versions/#supported-versions is past end-of-life, or Python 3.13 (not yet released) ?

from duckdb.

Hi @carlopi, thanks for getting to this so quickly!

Python version: 3.11.5

By indefinitely, I mean something like 20 minutes, though I didn't actually measure it.

from duckdb.

This is then surprising, if you download the
duckdb-1.0.0-cp311-cp311-macosx_12_0_arm64.whl from Pypi (https://pypi.org/project/duckdb/#files), say by clicking on the link, then moving the file to your current folder, and then do pip install duckdb-1.0.0-cp311-cp311-macosx_12_0_arm64.whl, does that work?

I would expect this should take a couple of second and install the pre-build wheel. If this work, then the question is why is that not fetched. If that doesn't work, could you possibly try with the universal macos wheel named
duckdb-1.0.0-cp311-cp311-macosx_12_0_universal2.whl ?

Thanks a lot for being available to track this down.

I would expect both to work, and either one of them to be selected before build from source, but unsure what might be off here.

from duckdb.

Thank you for your suggestions. I followed the steps you outlined and tested both the ARM64 and the universal macOS wheel files, but neither worked for my setup. Here are the details of my findings:

> pip install -v duckdb-1.0.0-cp311-cp311-macosx_12_0_arm64.whl
ERROR: duckdb-1.0.0-cp311-cp311-macosx_12_0_arm64.whl is not a supported wheel on this platform.

> pip install -v duckdb-1.0.0-cp311-cp311-macosx_12_0_universal2.whl
ERROR: duckdb-1.0.0-cp311-cp311-macosx_12_0_universal2.whl is not a supported wheel on this platform.

I've also updated pip to the latest version to ensure there are no compatibility issues from that side.

> pip install --upgrade pip
Requirement already satisfied: pip in /Users/kevinlacaille/miniconda/lib/python3.11/site-packages (24.0)

It seems like the pre-built wheels are not recognized as compatible with my environment. Any further guidance or troubleshooting steps would be greatly appreciated.

Thanks again for your assistance in resolving this!

from duckdb.

Can you possibly pipe the output of pip debug --verbose to a file, and add it here?

Basically pip debug --verbose > pip_debug.output and then attach it here (maybe checking there are no sensitive informations) ?

from duckdb.

Great idea, here's the output attached: debug_verbose.txt

from duckdb.

I can't go much further, it's possible that conda install duckdb might work, unsure, very wild guess based on the fact that your python installation is in the miniconda folder.

Someone else will pass by with more knownledge hopefully.

from duckdb.

You stated that you are on MacOS 14.3.1, but the output from debug --verbose seems to indicate that's not the case?
Seeing as the only Compatible tags listed are macosx_10_...

I would expect to see some entries with macosx_14_... in the list
My output contains those entries on a MacOS 14.1.2 M2 pro

from duckdb.

You're right, that's really strange, because here's what I'm seeing for my OS version:

> sw_vers
ProductName:		macOS
ProductVersion:		14.3.1
BuildVersion:		23D60

Let me poke around pip and I'll report back.

from duckdb.

I think this can be considered closed, mix and match of conda with pip is not working great, but that's fine / not really on duckdb.

from duckdb.

I'm actually experiencing the same issue and I'm not using Conda.
Macbook Pro M1 Pro, MacOS 14.4.1
pip version 24.0

from duckdb.

@AlonSam, can you possibly follow the same steps as above, in particular:

python -c "import platform; print(platform.platform())"

and

pip debug --verbose > pip_debug.output

?

from duckdb.

@AlonSam: can you provide the information needed to help solve this? Thanks

from duckdb.

I'm not sure how, but somehow it was simply solved by itself.. working now.
Sorry for re-opening the issue.
Thanks

from duckdb.

Cool, thanks.

from duckdb.