Dear all, I was struggling with

pyright

in Pants. I could verify that it breaks when using the target

::

on a big repository, at about

1,400

files, and ours has even more files. By chatting with @better-van-82973, it turned out that an additional partitioning on the files level could help in this case. I would be interested in trying to implement this myself but I'm completely lost on how to configure my IDE, visual studio code, to debug on a rule. Maybe I'm missing something, but this is what I do: 1. Once the environment is ready with rust/python, I run

pants

in the root folder. I use

pants

from

homebrew

. The process apparently succeeds and I could also verify by running

pants test src/python/pants/backend/python/typecheck/pyright/rules_integration_test.py

2. Then I link the venv to visual studio code by using the setting:

"python.defaultInterpreterPath": "~/.cache/pants/pants_dev_deps/edb7e9b2fbe8ee3a512a6fe15c1d92e1628269ed.venv/bin/python",

. It's worth noticing that the pattern is different from what is mentioned in the docs, supposed to include the architecture. I also tried to

export

in

dist

and link in the same way, but same result. However, visual studio code seems to recognize the environment corretly. 3. The I tried the following commands in the terminal: a.

PANTS_DEBUG=true pants --no-pantsd

b.

pants test --debug-adapter src/python/pants/backend/python/typecheck/pyright/rules_integration_test.py

By running with remote debugging on the file

rules_integration_test.py

I cannot make it stop at any breakpoint in

pyright_determine_partitions

. Is it because it is into an

async

function? What would it be the best way to debug this?

@bitter-ability-32190 I have this vague memory you've done a bunch of work on the debug-adapter functionality. Any insight?

Trying to debug pants tests is kinda cursed I think because there's 2 Pantss at play. I'd probably just try and run with PANTS_DEBUG and poke from there.

Thank you, I ended up by writing tests directly with partial rules to isolate the output. It’s not super efficient be it works. One question: can I potentially clone the pyright backend into a plugin? I assume I’d have to make some renaming, other than registering the plugin itself. It would help to test the modifications on our monorepo, if I cannot use pants from source.

I think

pyright

is pip installed from the resolve you provide? If so, use a VCS URL?

I think the problem is that it’s not the underlying Pyright package that needs to change for this, it’s the backend code in Pants. Previously I recommended modifying the backend in a fork and running using

PANTS_SOURCE

, but I’m curious if there’s another way