Messed around with dependency inference yesterday....
# developmentc
clean-city-64472
10/07/2021, 2:45 PMMessed around with dependency inference yesterday. I'm wondering if there is anyway you could print off a list of inferred deps and compare it to what is written in the BUILD file. It's kind of odd that you can't actually look at a BUILD file and say authoritatively whether it is fully leveraging inference or if it has overlaps with explicit requirements.
➕ 1
👍 2
h
hundreds-father-404
10/07/2021, 3:24 PMYou can! Pants 2.8 adds it to ./pants peek output
🙏 1
h
happy-kitchen-89482
10/07/2021, 3:42 PMhappy-kitchen-89482
10/07/2021, 3:44 PMI recall that you liked having explicit BUILD file deps because it helped enforce dependency hygiene? You could instead write a custom goal that uses this peek data to enforce your dependency rules, so that you get the hygiene without the boilerplate.
👀 1
➕ 1
c
clean-city-64472
11/10/2021, 6:46 PMJust coming back to this. I might be missing something but I don't see how this allows you to test if a dep would have been inferred had you not explicitly included it. When I add an explicit dep it shows up in both
dependenciesdependencies_rawh
hundreds-father-404
11/10/2021, 6:47 PMinferred = dependencies - dependencies_raw
c
clean-city-64472
11/10/2021, 6:50 PMthat doesn't seem to be true...
clean-city-64472
11/10/2021, 6:50 PMoh
clean-city-64472
11/10/2021, 6:51 PMif I put something that owuld have been inferred -it shows up in raw
clean-city-64472
11/10/2021, 6:51 PM^ this is true
clean-city-64472
11/10/2021, 6:59 PMThe meta problem is just how to move from explicit dependencies to inferred dependencies in a sane way with some sort of CI enforcing a standard.
➕ 1
h
hundreds-father-404
11/10/2021, 7:03 PMOh, I see now what you mean. So yeah, I only think this query will be helpful as a migration step. You can do something like save the ouptut of
./pants peek ::👍 1
hundreds-father-404
11/10/2021, 7:03 PMWhat are you hoping to do with CI?
c
clean-city-64472
11/10/2021, 7:04 PMI think if you explicitly set a dep that would have been inferred that should cause a CI failure.
h
hundreds-father-404
11/10/2021, 7:05 PMOhh that's interesting. So the idea there is you really want to lean in to dep inference? For example, to avoid the issue of forgetting to update BUILD files when you change imports, whereby you are depending on something unnecessarily?
c
clean-city-64472
11/10/2021, 7:06 PMEssentially. If you are migrating off of dep inference on a large codebase with many contributors you want to be able to lock in your gains as you go by adding CI coverage on BUILD files that no longer have explicit deps (other than those that aren't being inferred correctly).
➕ 1
h
hundreds-father-404
11/10/2021, 7:08 PMThat makes sense! Would you want to open a feature request? https://github.com/pantsbuild/pants/issues/new/choose
I'm thinking about whether this should be global vs. per-language. Right now we have the
[python-infer]c
clean-city-64472
11/10/2021, 7:08 PMSure - will do
clean-city-64472
11/10/2021, 7:16 PM