Open lisapro opened 2 months ago
We have to update the images used by Jupyterhub, so that they references aquamonitor-python 0.3.3. https://github.com/NIVANorge/niva_jupyter_hub/issues/30
I have temporarily downloaded older version on jupyter hub with !pip install --user --force-reinstall git+https://github.com/NIVANorge/Aquamonitor-Python.git@6cf4abf since the updated versions cause incompatibility. There are still some messages, but the functions seem to work, I will check more
no, I was wrong, it did not work, because the next commit (that has the error in the func) has the same version, I guess this is the resason why git still installs it. I can make a branch, downgrade libraries and install from the branch, is it ok for you?
I don't see why you should downgrade. The 0.3.3 version works.
becuase current versions of libraries joblib>=1.4 pandas>=2.2 requests>=2.32 yaspin>=3.0 cause incompatibilites in the Hub. And I need this to be up an runnning correctly by Friday morning. So the temporary solution is to downgrade those libraries to avoid incompatibilities.
Is it necessary for AM-Python to have such strict dependencies, @roarbra ? E.g. here I see we now require requests>=2.32
and yaspin>=3.0
. Does the library actually rely on new features introduced in those versions, or would it actually work OK with older versions (requests>=2.24
and yaspin>=2.5.0
)?
Having strict dependency pinning is fine if necessary, but otherwise it makes dependency resolution on the Hub harder, as outlined here.
@lisapro On your Hub, try
pip install --user --no-deps git+https://github.com/NIVANorge/Aquamonitor-Python.git
I've just tested on my Hub and it successfully installs aquamonitor-0.3.3
and the example notebook still runs OK afterwards. Hopefully it'll work for you too as a temporary fix?
@JamesSample the --no-deps option fixed it, yes!
No, not at all. I'm just pushing these versions to be up to date, and avoid warnings about vulnerabilities. I will stop doing that.
I guess vulnerabilities are a good reason to keep updating the pinning. Maybe we can try to find a balance somewhere in the middle?
You can also keep doing exactly as you are, because
pip install --user --no-deps git+https://github.com/NIVANorge/Aquamonitor-Python.git
should work, unless one day AM-Python really does rely on new features only available in the updated requirements.txt
. If this happens, the command above won't fix AM-Python, but it won't break the environment either. Then I just have to spend some time investigating dependencies (which I do anyway during the periodic updates).
Importmal app is using Aquamonitor-Python for getting the data https://github.com/NIVANorge/importmal_app/tree/master I use it though Jupyter Hub
Now it fails on get_projects function, I see that there were recent changes in it. Could you please take a look?