I think while I've been away from

pants

, the dependency analyzer got too good 🙂 Silly problem: I'm trying to extract a subset of a library which tried to hide its dependencies by putting imports under

if

statements, which I think used to work (but I may be wrong) ie something like:

def hide_imports(library: str):
    if library == "numpy":
        import numpy as result
    elif library == "pandas":
        import pandas as result
    return result

but

pants

cannot be fooled:

❯ ./pants dependencies --no-dependencies-transitive src/.../test.py
//:root_requirements#numpy
//:root_requirements#pandas
...

(the situation is more complex with internal dependencies, but you get the idea; what if we wanted to build a library which only included

numpy

as a dependency and would be ok if someone called

hide_imports("pandas")

?) Is there any way to build a

python_distribution

or

pex_binary

with "only obvious" dependencies enclosed or expressly excluding some dependencies? Trying to build a subset of a library for a client without exposing more IP than necessary.

You have options... 1. In the source code on the import line you don't want inferred add

# pants: no-infer-dep

2. In the BUILD file declare a dependency for the file and use an exclusion: e.g.

!//:root-reqs#pandas

(note the exclamation mark) 3. Muck up your source code import so pants doesn't know what's going on (mostly a joke 😌)

Awesome, and thanks for the clarification; I had even seen the

no-infer-dep

business earlier but somehow missed the comment tag (and knew about the exclusion but thought it would just break things somehow). Thanks very much for the help!