And, even better, is there no way to use stubs like this and not have to make them an explicit dependency? The docs for stubs (https://www.pantsbuild.org/docs/python-typecheck-goal#type-stubs-pyi-files) seem to say I shouldn't need to do that but they're not discovered unless I add them as an explicit dependency.

So, to get it working automatically, you would need to set

module_mapping

because pants thinks the module right now is `numpy_stubs`: https://www.pantsbuild.org/docs/python-third-party-dependencies#inline-requirements But then you would have a problem, both the requirements

numpy

and

numpy_stubs

would export the same module name of

numpy

, and it would now be ambiguous when we have imports of

numpy

which you want to use. We do the safe thing of defaulting to neither when there is ambiguity, which is necessary because it would be an error for example if you had

numpy==1.2

and

numpy==1.4

in your project, Pants inferred a dep on both, and we tried installing both which is illegal