Dumb question. Reading the docs I couldn’t find a ...
# generalb
breezy-apple-27122
08/02/2023, 9:11 AMDumb question. Reading the docs I couldn’t find a way to reference a dependency that is installed by another dependency. For example I have fastapi that references pydantic. If I import something from pydantic, pants will warn that the dependency couldn’t be inferred. How can I point it correctly without adding to the pyproject file? I’m using poetry_requirements
f
fast-nail-55400
08/02/2023, 9:37 AMIf you are importing a dependency in first-party code, then it is now a direct dependency of that first-party code and should be referenced in your pyproject file. The fact that it is also a transitive dependency of
fastapifast-nail-55400
08/02/2023, 9:38 AMIn my view, this is just good "dependency hygiene."
b
breezy-apple-27122
08/02/2023, 9:39 AMYes but this can also lead to incompatibilities
breezy-apple-27122
08/02/2023, 9:39 AMLike my previous post. Pants hanged because dependencies were incompatible
f
fast-nail-55400
08/02/2023, 9:42 AMAnd making sure the version in pyproject matches the one chosen by
fastapib
breezy-apple-27122
08/02/2023, 9:44 AMFrom the developer perspective i would prefer to let fastapi rule this. This would avoid confusions and save a big CI bill with hanging builds
breezy-apple-27122
08/02/2023, 9:45 AMPackage update PRs come from renovate. So it’s not coming from developers
b
bitter-ability-32190
08/02/2023, 11:28 AMModern resolvers shouldn't have an issue with that, but if you want you can slap a
# pants: no-infer-deph
high-yak-85899
08/02/2023, 2:12 PMAre you not generating lock files?
high-yak-85899
08/02/2023, 2:13 PMAnd totally agree about specifying it as a direct dependency. If you rely on fastapi to vendor the package, you have no idea if your usage is valid with the package version they select. The only way to do that is to pin a range you know your usage is valid for. Then you guarantee compatibility with there’s by generating lock files.
2 Views