Hi everyone, hope you're well. I recently started using Pants within my organisation and I am facing some issues with dependencies for tests. I have generated a lockfile using pants generate-lockfiles that generated a lockfile based on my pyproject.toml. However, whenever I run a test I get an error that says "No module named numba", despite numba being within the locked_requirements section of my pants lockfile. Below is my setup, was wondering if anyone has any insight on to this, or can tell me what I'm doing wrong? pants.toml

[python]
interpreter_constraints = ["==3.10.*"]
enable_resolves = true
default_resolve = "poetry-lockfile"

[python.resolves]
poetry-lockfile = "poetrylock.lock"

BUILD (in top-level with pyproject.toml)

poetry_requirements(name='poetry')

BUILD (in same directory as tests)

python_sources(
    name='repo_sources',
    sources=[
        'repo/**/*.py',
    ],
    resolves='poetry-lockfile'
)

python_tests(
    name='tests',
    sources=[
        'tests/**/*.py'
    ],
    resolve="poetry-lockfile"
)

Does anything actually import numba in your test?

Yep, what @gorgeous-winter-99296 is getting at is that Pants uses dependency inference to create sandboxes for your code. If there is no explicit import of

numba

then Pants won’t know to pull it into the sandbox. You can work around this by adding an explicit

dependencies=

to your

python_sources

. Or even better, add it via an

overrides

, so that you’re adding it as a dep of just the file that actually uses it, instead of every file you’re globbing over…

Thank you both, that worked