Installation error building wheels for numpy.

[WPennock]

I was trying to install aguaclara within a newly-installed Anaconda environment. When I did pip install aguaclara within the Anaconda Prompt, I got:

ERROR: Could not build wheels for numpy, which is required to install pyproject.toml-based projects

I looked it up, and this seems to be a backwards compatibility issue with Numpy. My current installation has Numpy 1.26.4, but the pyproject.toml for aguaclara has version 1.22.1 specified. I am thinking this means that the version of the aguaclara package needs to be updated to current package versions. I have seen others facing this issue in Windows 11. For reference, my environment is:

• Operating system: Windows 11
• Python version: 3.11.7

@fletchapin @monroews Is this package still being maintained? Is there a way I can help?

[monroews]

We don't have anyone in charge of maintaining this code base. I can happily give you editing permission to update the code!

[WPennock]

I would be happy to do this, since it falls under the scope of my work and our lab group and my classes use it fairly frequently, so I get plenty of feedback on issues. However, it has been some years since I looked at how the testing, versioning, and documentation. Is there any resource I can consult or person I can talk to in order to make sure I carry this on effectively?

[WPennock]

There was a developer guide (guide-dev), but it appears that link is now broken. Is there somewhere I can find this, @monroews?

[Edit: I found it here]

[monroews]

if you proceed to update the code please fix the link too!

[WPennock]

Thank you, @monroews!

I am afraid I am struggling to set up my local environment as per the guide, and I believe it is the issue of dependencies again. I have no trouble creating the local Git repository, but I run into issues when I try to make the local aguaclara repository the package Python on my machine will run.

If I try to do this globally (pip install --editable . -U --user), I get

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
numba 0.54.1 requires numpy<1.21,>=1.17, but you have numpy 1.22.1 which is incompatible.

For context, aguaclara 0.3.4 requires numpy 1.22.1, so this seems to be an inconsistency in the dependencies although I can confirm that aguaclara works for me. I do want to make sure I am using version 0.3.4. Is there a way to test that in Python? For other packages like numpy, you can enter print(np.__version__). Is there a similar approach for aguaclara?

If I try to do this for the virtual environment (pipenv install --dev -e .), I get

 ResolutionFailure(message=str(e))
[pipenv.exceptions.ResolutionFailure]: Warning: Your dependencies could not be resolved. You likely have a mismatch in your sub-dependencies.
  You can use $ pipenv run pip install <requirement_name> to bypass this mechanism, then run $ pipenv graph to inspect the versions actually installed in the virtualenv.
  Hint: try $ pipenv lock --pre if it is a pre-release dependency.
ERROR: No matching distribution found for scipy

I'm hoping this is just that I am not familiar with doing this, but my understanding is that the dependencies aguaclara 0.3.4 requires are several years out of date and not compatible with my current python environment. However, I understand pyenv is supposed to help work around this.

@HannahSi and @fletchapin would you be willing to point me in the right direction so that I can begin keeping this package updated?

[monroews]

my only advice is to use chatGPT to guide you. I think they would provide good advice for this problem.

[fletchapin]

Hi @WPennock sorry I am just seeing this now but I can take a look this weekend at the dependencies! I don't think it has been actively maintained but once the dependencies are updated if you want to take it over that would be great.

[fletchapin]

When I just installed the package with pip install -e . after launching a virtual environment with pipenv shell I did not get an error. I believe the issue is you are not using a fresh environment since the dependency clash is with numba==0.54.1 and numba is not an aguaclara dependency.

To avoid this always make sure to create a new virtual environment for every package!

[fletchapin]

@WPennock if you have any remaining issues let me know and I will re-open it, but otherwise I am closing this issue

[WPennock]

Thank you, @fletchapin! I have trouble even with pipenv shell entered first, although I hadn't been doing that, and I think that would be good to add to the developer guide. For reference, I get references to ssl and urllib3, although I am sure I will eventually be able to work out the specifics.

WARNING:pipenv.patched.pip._internal.models.search_scope:pip is configured with locations that require TLS/SSL, however the ssl module in Python is not available.

...

[ResolutionFailure]:       raise ResolutionFailure(message=str(e))
[pipenv.exceptions.ResolutionFailure]: Warning: Your dependencies could not be resolved. You likely have a mismatch in your sub-dependencies.
  You can use $ pipenv run pip install <requirement_name> to bypass this mechanism, then run $ pipenv graph to inspect the versions actually installed in the virtualenv.
  Hint: try $ pipenv lock --pre if it is a pre-release dependency.
ERROR: No matching distribution found for urllib3

I can keep working on getting my local repository working, but the larger reason for this issue is that I do not believe people running Windows 11 can pip install aguaclara, and I wonder if that may be true for other operating systems as well. I think the dependencies need to be updated for the package.

Would that basically require looking up current versions of packages that are stable and changing those version numbers in setup.py? I found this page and plan to work through it, but I wanted to make sure I'm working on this in a way that is consistent with what has been done in the past.

[fletchapin]

When I searched the SSL error, I saw that it may be an issue with the way your SSL certificate is installed on Windows: https://stackoverflow.com/questions/41328451/ssl-module-in-python-is-not-available-when-installing-package-with-pip3

I have installed the package for Linux, Mac, and Windows (albeit Windows 10 and Python 3.10) so I would be surprised if it does not work on your machine. Perhaps I can try to install it on a VM with Windows 11 and Python 3.11 to double check

[monroews]

I'd be very interested in knowing if the AguaClara code still works well in Colab or other online versions of Jupyter notebooks.

[fletchapin]

@WPennock like you said, we can change the version of numpy it requires to unpin it from 1.22.1. After looking at the Colab notebook it installs, but with a warning, so maybe some packages for the newer versions of Python are incompatible with that old numpy version.

With that said, the import works correctly despite the warning and you are able to use the aguaclara package in Colab currently.

[WPennock]

@fletchapin Thank you for looking into all of this! Okay, I am glad that aguaclara is working in all of those environments.

I am not sure why my student and I had trouble on Windows 11 machines. The errors that were thrown related to the version of numpy, so I suspect that requiring a more recent version of numpy or using * to not specify a version may solve the issues we had.

I still cannot install on my laptop directly, and I believe that this is because I have Python 3.12. I get similar errors to those shown here. The answer suggested that numpy 1.26.4 is required for Windows 12. I am able to install aguaclara in a virtual environment (thank you for your help with this!) if I take out the numpy version requirement. I'm even able to run pipenv run pytest (which incidentally has failures for functions in onshape_parser.py and procoda_parser.py, which I suspect are due to the new numpy version). I will see if I can eliminate these errors and create a pull request for changing the requirement to 1.26.4.

[fletchapin]

Hi @WPennock I've been busy on vacation but I just wanted to check if you had any further questions about this issue or help creating a PR