So I just spent several hours rewriting hundreds of imports Pants #general

So I just spent several hours rewriting hundreds o...

fierce-fall-70380

02/04/2025, 6:39 PM

So I just spent several hours rewriting hundreds of imports in our monorepo to be relative to the outer source root. I'm now down to a handful of inference warnings that I'm struggling to resolve. 🧵

fierce-fall-70380

02/04/2025, 6:47 PM

Here is one inference error:

Copy code

13:28:57.24 [WARN] Pants cannot infer owners for the following imports in the target fooserv/itest_fooserv/fooserv_itest_helper.py:

  * email_first (line: 50)

If you do not expect an import to be inferrable, add `# pants: no-infer-dep` to the import line. Otherwise, see <https://www.pantsbuild.org/2.24/docs/using-pants/troubleshooting-common-issues#import-errors-and-missing-dependencies> for common problems.

Here are the imports on line 50:

Copy code

from web import email_first, login, mfa

Note:

Copy code

ACAM-M-RWLH: acam$ pants list fooserv/web/email_first.py fooserv/web/login.py fooserv/web/mfa.py
fooserv/web/email_first.py
fooserv/web/login.py
fooserv/web/mfa.py
ACAM-M-RWLH: acam$

I've confirmed the contents of email_first.py don't matter (I still get the inference error if I replace it with a zero byte file). There are also no other modules with the same name over the whole monorepo.

microscopic-knife-5995

02/04/2025, 6:58 PM

Could you try adding

__init__.py

? That might fix it...

fierce-fall-70380

02/04/2025, 7:10 PM

Yep there's already a

web/__init__.py

...

microscopic-knife-5995

02/04/2025, 7:12 PM

Maybe try

-ldebug --keep-sandboxes=always

to see if that gives any insight?

fierce-fall-70380

02/04/2025, 7:14 PM

I'll try it. By the way does

--keep-sandboxes

work when using just the pants binary / scie-pants? Or do you have to run it from the git repo for that to work? I've tried that option before but never received any output confirming the sandbox would be kept or where it's located.

microscopic-knife-5995

02/04/2025, 7:15 PM

I think it works for most cases... I haven't found a case when it doesn't...

👍 1

fierce-fall-70380

02/04/2025, 7:16 PM

Ok this is interesting. I changed the import line to add another (unused) import from that same directory, and now that's the one it complains about:

Copy code

-from web import email_first, login, mfa
+from web import logo, email_first, login, mfa

Copy code

15:14:05.54 [WARN] Pants cannot infer owners for the following imports in the target fooserv/itest_fooserv/fooserv_itest_helper.py:

  * logo (line: 50)

If you do not expect an import to be inferrable, add `# pants: no-infer-dep` to the import line. Otherwise, see <https://www.pantsbuild.org/2.24/docs/using-pants/troubleshooting-common-issues#import-errors-and-missing-dependencies> for common problems.

This really smells like a bug ...

microscopic-knife-5995

02/04/2025, 7:17 PM

Ah... I think your import path is incorrect. Does this work?

Copy code

from fooserv.web import logo

fierce-fall-70380

02/04/2025, 7:17 PM

No because

is not a source root.

fierce-fall-70380

02/04/2025, 7:18 PM

Ok yeah, I rearranged the imports, and now it's complaining about

mfa

instead:

Copy code

-from web import email_first, login, mfa
+from web import mfa, email_first, login

🤔

microscopic-knife-5995

02/04/2025, 7:19 PM

I think it just complains at the import it first hit.

fierce-fall-70380

02/04/2025, 7:19 PM

It is not that simple because

from web import login, mfa

works without error!

fierce-fall-70380

02/04/2025, 7:24 PM

So strange. Have to run, back in awhile ... thanks for the help @microscopic-knife-5995

fierce-fall-70380

02/04/2025, 8:07 PM

> This really smells like a bug ... 🚨 If I move the

web

imports line closer to the top of the import statements, I don't get the inference error. That definitely seems like a bug.

fierce-fall-70380

02/04/2025, 8:17 PM

Maybe try
-ldebug --keep-sandboxes=always
to see if that gives any insight?

According to the docs, when you run with

--keep-sandboxes

you are supposed to see a message like:

Copy code

21:26:13.55 [INFO] preserving local process execution dir `"/private/var/folders/hm/qjjq4w3n0fsb07kp5bxbn8rw0000gn/T/process-executionQgIOjb"` for "Run isort on 1 file."

That message is never printed for me when using the

pants dependents

goal, though.

happy-kitchen-89482

02/05/2025, 1:18 AM

FWIW

--keep-sandboxes

should always work, and log the locations of the sandboxes to the console

happy-kitchen-89482

02/05/2025, 1:19 AM

pants dependents

doesn’t run any subprocesses, so there are no sandboxes to speak of

happy-kitchen-89482

02/05/2025, 1:21 AM

And yes, if the order of imports is affecting the result then that does smell buggy

happy-kitchen-89482

02/05/2025, 1:21 AM

I assume

fooserv/

is your package root? so that

web

is a top-level package?

happy-kitchen-89482

02/05/2025, 1:23 AM

Would you be able to create a small toy repo that reproduces the issue? It will be hard to debug without a consistent repro. I fear that since rearranging the imports affects the behavior, this is going to be difficult to reproduces independently…

happy-kitchen-89482

02/05/2025, 1:23 AM

Alternatively we can walk you through how to run Pants from source to debug this

fierce-fall-70380

02/05/2025, 1:39 AM

@fast-nail-55400 pointed me at that new debug option in the other thread; going to try that now.

fierce-fall-70380

02/05/2025, 2:36 AM

@happy-kitchen-89482 Thanks for the info. Turns out it was PEBKAC (I botched a manual edit to the import statements in this one module after a rebase conflict, whoops).

6 Views

Open in Slack

Previous Next