Open jeroenhabets opened 4 years ago
hi @jeroenhabets thanks for the feedback. Main concern of pip install is that cli dependencies mess up with users' python environment and existing installed packages, all those unexpected python packages dependencies version issues. That's why we recommend user install via package manager per OS
I just want to add that being able to install azure-cli
as a dependency via pip
in the usual Pythonic way is required when azure-cli
is used within another Python package. Here is a related issue.
We have an internal Python package which uses azure-cli
, which is currently broken (specifically, installation of the internal package fails due to a dependency conflict with azure-storage-blob
).
My requirements.txt
looks like this:
azure-cli==2.5.1
azure-identity==1.2.0
azure-keyvault-secrets==4.1.0
azureml-sdk==1.5.0
snowflake-connector-python[pandas]==2.2.8
Which results in this error:
ERROR: azure-cli 2.5.1 has requirement azure-storage-blob<2.0.0,>=1.3.1, but you'll have azure-storage-blob 12.3.2 which is incompatible.
Presumably due to incorrect dependency resolution between azureml-sdk
and azure-cli
.
Python version: 3.7.6
Operating system and processor architecture: Darwin-19.4.0-x86_64-i386-64bit
@IRLeif thanks for bringing another use-case. Regarding your issue, that bit me as well and actually motivated me to log this feature request. I had already logged the azure-storage-blob conflict issue, see : #13876
For Python the standard way of installing packages is pip.
The pip installation works already but is not promoted nor supported. If it were we could just add a
azure-cli
line to ourrequirements.txt
and it would automatically get installed and updated.Additional context Is there any reason (left) that requires alternative installation options like MSI on Windows and package managers on Linux?