Running pip as a Python module
Last weekend I was helping run a Python workshop and one of the technical people helping me out ran into a fairly bizarre installation issue with her environment.
conda install was working but
pip install was failing.
Basically the very quick summary is that her pip command was pointing to a different Python installation than the Python that was on the command line path. These sorts of mismatches with the path can lead to some very strange installation issues. I think this is another good example of why always using environment that are not the system Python is a good idea, since not installing into the base Python prevents a variety of issues like this.
Because I was short on time, due to running a technical workshop, I just went for the safe option of using
python -m pip to invoke pip. This will ensure that the path is the same for both
pip (unless something very pathological is going on, in which case I'd say your install is truly broken).
This was probably a better approach than trying to fix up all the system paths given the limited time.
There's other people who agree and a great blog post that explains why you should prefer this.
The reason I figured I should make a blog post about this is because not a lot of people are aware that they can run modules from the command line with
This can be useful in many situations, specifically I find I use
python -m pytest relatively often in a few projects.
The docs for this command line options can be found here.
Once upon a time you couldn't run pip like this since pip is a separate project and hence wasn't available as a Python module in a standard install.
While pip is still separate from the main Python project, in the sense that it has different release schedules and a different source tree of code behind it, these days you can still run pip via
python -m pip.
Specifically PEP 453 came about to make sure that
pip could be run via python as though it were a module.
If you are interested in learning more about why this was done I'd recommend reading that PEP and about ensurepip.