Direct dependency is incorrectly treated as indirect dependency

[konstin]

We incorrectly treat numpy as a transitive instead of a direct dependency, which means we don't emit the correct diagnostic about a missing lower bound:

[project]
name = "dummy"
version = "0.1.0"
description = "Add your description here"
requires-python = ">=3.9"
dependencies = [
  "numpy",
]

Relevant log line from RUST_LOG=uv=debug uv lock:

DEBUG Adding transitive dependency for dummy==0.1.0: numpy*

We need to change this check to consider workspace packages:

uv/crates/uv-resolver/src/resolver/mod.rs

Line 2163 in 30e11c6

if let Some(for_package) = for_package {

[mstepan]

I was trying to reproduce this issue using posted pyproject.toml file but seems it's working as expected:

uv/target/debug/uv pip compile example/pyproject.toml -o example/requirements.txt
...

DEBUG Adding direct dependency: numpy*
INFO add_decision: example @ 0a0.dev0 

[charliermarsh]

Try using uv lock instead of uv pip compile.

[Aditya-PS-05]

@konstin
I am solving it and setting this type of diagnostic message,

Is this the required behavior or somethingelse

[konstin]

We should treat numpy as a direct dependency here, which means we should show the The direct dependency {package} is unpinned. message here for numpy: We should have an additional check that the for_package isn't a workspace member.

[Aditya-PS-05]

How should I check if it's a workspace member? Should I need to explicitly define that these are the workspace members or something else?

[konstin]

ResolverState::workspace_members has this information