Declare the optional dependency and then import it unconditionally if the optional is installed. You can conditionally import a file you control based on asking whether an optional specifier was installed, either by probing the filesystem, a CLI argument, an environment variable, or whatever folks come up with for https://github.com/pypa/packaging-problems/issues/215. The "if: ... import" laziness-defeating clause only applies to your stub module, and the imports in that module are unconditional and thus will work with analyzers/laziness/etc.
I think this import flow tends to be the canonical one.
I might recommend doing something like importing from something like `numpy.version` (or some other "random" very small utils package) so that the work done on file load is still fairly small.
Common? Yes. Canonical/ideal? No. Optional dependency specifiers in your package metadata is a much better way to go. You can use those even if you don't publish your package as an artifact; `pip install -e .[optional-thing]` should work, or can be made to, easily.
In my preference order: an explicit CLI option/envvar which requires the optional dep, a flag file on the filesystem, or an ImportError attempt if you absolutely must. Use a stub/trampoline file in the conditionals so that static analyzers and laziness analysis on that file itself still work.
Yep it's not clear, and a bit of a case-by-case thing. Fortunately most codebases only need to do this in a handful of spots so you can really "just" look at a package and find a file that probably is safe.
For example, numba._version only imports standard library stuff so is probably good enough here[0]
It would be nice to have general querying capabilities here, of course. I just think that in practice there's at least a halfway-decent workaround in almost any real situation
