Recently we have a new

pyproject.toml

in our repo (along with other docker, an original

pyproject.toml

project + other things). the two python projects both use numpy, and in their

pyproject.toml

are using the same version. both BUILD files, load the requirements from

poetry

poetry_requirements(
    name = "reqs",
    resolve = "python-default",
    source = "pyproject.toml",
)

and both, use the same resolver. I am now seeing that pants complains of inability to resolve

numpy

on both projects. One is noisier than the other,; numpy is the only library commonly used between both. It seems wrong to need to explicitly declare in each project, which of the pair to use

BUILD projecta,

    dependencies = [ "lib/projecta:reqs#numpy" ]

dependencies = [ "lib/projectb:reqs#numpy" ]

Should I be using a separate resolve for each project ?

I guess it depends on what you want going forward - you mention that

numpy

is the only shared dependency between the two projects now, but do you anticipate there being more shared dependencies in the future? IMO, if you would want both projects to share dependencies/code, go with a single resolve. If you want them to be fully separate and not share dependencies or code, go with separate resolves.

do you find (those that run large repos with many py projects) that different resolves occur. We have 4 main pythoin projects, 3 are fleet of foot (can take latest - have less dependencies) and 1 that has a complex dependency tree - which causes us to wait for many upstream libs to catch up before we can version++.