Using Multiple Versions of Python about pycall.jl HOT 4 CLOSED

If you omit the -l option, it doesn't read the .profile file. This is the whole point of running which python under bash: software like Anaconda installs Python under a new directory that is only added to the PATH in ~/.profile.

from pycall.jl.

I agree that the whole point of adding the -l option is to read the .profile file.
Suppose that I have both Enthought and Anaconda distribution installed in my machine and I have to switch between them. Choosing the default python by reading the .profile file will need me to edit my .profile file every time I want PyCall to pick up different version of python. Omitting the -l option enable PyCall to pick up whichever python that is in my PATH now.

Unless you are not using Julia in a shell, the current active PATH will point to whatever python distribution you are using, including the directory Anaconda Python live in. My point is that why read the PATH from the .profile file? Why not the current active PATH since we can change it easily?

Sorry if I miss some obvious point and thanks for taking time to response :)

from pycall.jl.

The problem is that a lot of people don't run Julia from a shell in MacOS, they run Julia from a prepackaged binary, and with a prepackaged binary of Python like Anaconda. If we don't read the PATH in ~/.profile, it is very hard for those users to figure out how to tell Julia to find the right Python version. (See also JuliaLang/IJulia.jl#67)

from pycall.jl.

I should add that the right solution here is for the prepackaged Mac Julia binaries to source ~/.profile to get the user's PATH. That way I can go back go using just python and things will work, with no need for bash -lc 'which python' hackery.

But that's a bug report that should be filed against Julia, not here; please comment at JuliaLang/julia#4222, for example.

from pycall.jl.