jl.pytruth(np.bool_(False)) == true

[schlichtanders]

Affects: Both

Describe the bug
A clear and concise description of what the bug is. What did you expect? What happened instead?

numpy has its own boolean values see its docs.
Using python, the numpy bools look exactly like normal bools, but are different types.

When using PythonCall/JuliaCall unfortunately numpy bools behaviour entirely differently from normal Python bool. That is very confusing. Especially

from juliacall import Main as jl
jl.pytruth(np.bool_(False))  # returns true

I guess this is just a bug with pytruth which actually should return true in this case.

Your system
Please provide detailed information about your system:

• The operating system: Nixos
• Julia 1.9.0 PythonCall/JuliaCall 0.9.13

[cjdoris]

Are you sure you're on 0.9.13? This look a lot like an issue that was solved in 0.9.13, and I can't reproduce it:

>>> from juliacall import Main as jl
>>> import numpy as np
>>> jl.pytruth(np.bool_(False))
False
>>> jl.pytruth(np.bool_(True))
True

Also, what's the actual issue you're having? Because it never really makes sense to do jl.pytruth(x) from Python when you can do bool(x) instead.

[schlichtanders]

yes I can confirm I am on 0.9.13

In the same vscode notebook jl.pytruth(np.bool_(False)) returns true

[schlichtanders]

My usecase is that I am compiling a super small intro to PythonCall which is going to be presented today evening, where by chance I run into this point.

[cjdoris]

I really can't see how you can get this issue on the current version. Are you definitely not on a dev version of PythonCall? What are juliacall.__version__, juliapkg.status(), juliapkg.state.STATE and pip show juliacall?

Also, when I asked about your usecase, I more meant in terms of code. Like I said, doing jl.pytruth(anything) is weird because you can do bool(anything) instead. Are you actually using pytruth from within a Julia function? If so, you should wrap that function using jl.pyfunc to create a Python callback function which avoids the automatic conversion causing the issues you're having.

[schlichtanders]

thank you so much for not giving up!! Everything works now.

indeed the python kernel I used was not the one I thought I would be using. So I got 0.9.12 there for juliacall.version

and yes indeed I was originally using pytruth on the julia side of things, but running a python kernel using the python way was better for sharing here on github.