So here is my issue. I've managed to install PyPy using conda with the following command:
conda install -c conda-forge pypy3.5
Unfortunately, when I try to create an environment that uses this pypy3 exectuable, I cannot find a way to do it. If I run pypy3
, I get the PyPy shell without any issue and I can also run my programs using pypy3
instead of python
.
Though now, I'd like to be able to create a full environment using PyPy if that's possible. I've tried things like the following in vain:
conda create -n pypy3 python=pypy3
conda create -n pypy3 python=pypy3 -c conda-forge
I've tried specifying pypy3.5, and other variations but nothing works. I can see the pypy3 executable in the bins of my Miniconda installation but I cannot find a way for Conda to use it. I cannot find much on the internet on this since people seem to be asking just for the install of PyPy via Conda, and nothing about creating environments using PyPy.
So here are my questions:
- Is there already a way to create a Conda environment using PyPy instead of a regular CPython?
- Is there a way to force Conda to look first locally instead of checking online for distributions?
- Is there a way to enforce an executable to use as Python when we create an environment with Conda?
- Would there be a dirty workaround possible creating a regular environment and then forcing this environment to point to my pypy3 executable?
I don't know if anybody can help here. Maybe the solution already exists but I couldn't find much on the subject anyway.
EDIT: As suggested by @darthbith I can use the following command:
conda create -n pypy3 -c conda-forge pypy3.5
But that doesn't do what I would expect. I can use pypy3
to get the shell and execute my Python programs but it is not handled as a regular Python version. I'd like to have PyPy to be considered like any version of Python and be able to use pip to install packages (most pure Python packages should work with PyPy).
I understand that a lot of people would advise against what I'm trying to do here, but I see it as just being a faster version of Python that works for anything that doesn't rely on C libraries. Since I'm working on pure Python libraries and many libraries in PyPi are written in pure Python, I don't see why I wouldn't be able to achieve what I'm trying to do here.