Is there an interface to access the information in pyproject.toml from Python?
In particular, I'd like to access the dependencies. It doesn't seem hard to do
toml.load("pyproject.toml")['project']['dependencies']
and then parse that list. But that would be rewriting logic that must already be available somewhere.
Apparently there is importlib.metadata.requires(), which would be better than the line above, but that seems to give a similar list of strings that still requires manual parsing:
>>> metadata.requires("scipy")
['numpy<1.27.0,>=1.19.5', 'pytest; extra == "test"', ...
(As for the XY problem, I was thinking about writing a test that tries to install a package with the lowest versions of its dependencies to see whether the package still works.)
In both cases, whether project metadata (such as dependency requirements) is obtained from pyproject.toml's standardized [project] section or via importlib.metadata the values should be the same. At least, as far as I know, the dependency requirements have the same format.
Probably I would use the 3rd party packaging library, in particular its packaging.requirements.Requirement(), to parse the dependency requirements. This library can handle other values of project metadata as well, read the documentation. As far as I know this is the library that pip itself uses internally (so it should be always reliable and up-to-date).
The issue with obtaining project metadata from pyproject.toml, is that some fields can be declared dynamic. Which means their value is not set in pyproject.toml.
If you are fine with that, then the TOML parser from Python's standard library tomllib could be good enough. You could also give pyproject-parser a try, it seems like a correct and reliable library.