Hey y’all just wanted to do a sanity check, I’m finishing my migration to pants v2 so there’s a couple of reorganizing that I’m doing.. 1. For the dependency inference part, it’ll automatically “know” the transitive dependencies? I don’t have to explicitly define any internal dependency anymore? So far the

package

and

run

are working fine so I’m assuming and hoping that I still get that sweet

--changed-dependees=transitive

out of the box too… 2. This one I’m not sure if I’m tripping or not but I have an internal ML lib with classifiers, featurizers and all that weird cool stuff. But some of them have huge dependencies such as Tensorflow others dont.. so I was wondering: • Should I move the internal modules/packages into their own “library” such as:

# current one
libraries/
  ml_stuff/
    __init__
    BUILD
    preprocessing/
      __init__
      BUILD
    classifiers/
    ...

# new one
libraries/
  ml_stuff_preprocessing/
  ml_stuff_classifiers/

• Or…

pants

can get those specific inside packages in case I only need..

ml_stuff.preprocessing

? And if that requires Tensorflow it would get it, otherwise it wouldnt (?)

dependencies are inferred at the file level, and so if it’s inferred, it should only be inferred for the files that need it

which is what I had for v1, but now I’m using

BUILD

files everywhere with a simple

python_library()

inside and I noticed that the behavior changed

so: re 1.: to confirm that individual files (or targets) have the dependencies you’d expect, you can use

./pants dependencies --transitive $file

or

./pants dependencies --transitive $target

Yup, thanks for the double check. I sort of had a feeling that that’s what was happening but now I understand why. 🙂