is it possible to disable python dependency inference globally for a specific module? I'm trying to avoid adding

# pants: no-infer-dep

to thousands of lines. There is a library that we depend on that isn't published to pip. I guess I could work around it by mapping some other random package to the module path...

you can point pants at private (non-pip) package repositories

So it's a long story, but essentially this package is just a C++ library and can't be installed via pip because of the requirements. information here: https://graph-tool.skewed.de/installation.html#native-installation

Python does support binary wheels these days so it's entirely possible to bundle C++ code in python packages in a reasonable way, but I understand this library is probably old enough that nobody has done that. I don't have anything useful to add though, I would just point it to some dummy library as you said.

To be clear, I think they don't publish Python wheel not because it's old/outdated, but because the Python wheel installation would fail for call it 95% of users and the user experience sucks so they just push people to use conda instead.

total tangent, but binary wheels have been very good for me lately tbh; I know there used to be a lot of trouble installing things from pip since it needed to compile source, etc, but as basically everyone has started shipping binary wheels it's all been a lot better IME

In this case binary's aren't good enough because of the other c++ deps. I don't full understand them, I just don't think it's like a go binary or a rust binary. I agree though, binary dists are nice and work well, when they work 🙂