We have a Python repo that uses a

conda

environment. This means we have to add

python_requirement

entries for all third-party dependencies used by goals like

test

and

package

to our root

BUILD

file or

python_infer

throws

UnownedDependencyError

. I’d like to run

./pants dependencies

on our tests to check that Pants can infer all the third-party dependencies successfully. What is the best way to do this?

There is none that I'm aware of better than your process. We've had no conda user yet to my knowledge contribute in any significant way to Pants and none of the existing contributors have meaningful conda experience.

Gotcha. I’ve chimed in on the GitHub discussion about

conda

. I can try to dig into

conda

integration when I find some time. In the meantime, is there a way to tell

./pants dependencies

to only run on tests and imported modules? I thought about adding

tags=["tests"]

to

python_sources

in test directories, but my understanding is that this will only check dependencies for ancillary Python files in those directories. Is it possible to pass

python_tests

as a target?

See

./pants help filter

You can always mix in

./pants --filter-xxx=yyy <goal>

Thanks, will do!

Maybe a lie, but the option exists. There is fuzziness in my head about when it applies / pre vs post.

Looks like

--filter-target-type

does the trick.