I can't seem to bootstrap 2.13.x correctly:

File "pydantic/dataclasses.py", line 52, in init pydantic.dataclasses
ImportError: cannot import name dataclass_transform

When running

pex3 lock export

what's the order of the packages? Would it possible to sort into alphabetical order for easier diffing?

locked_requirements

in the JSON seems ordered, which is great

diffing like https://github.com/pantsbuild/pants/pull/17347 ? 😉

it’s available as a cli tool for standalone use too: https://pypi.org/project/lockfile-diff/

Nothing beats a PR diff though 😛

oh, those… 😛

hey folks, I'm working on helpers to generate the lockfile rules. For testing, I'd like to set a query rule for the lockfile sentinel (eg

Flake8LockfileSentinel

). Problem is, that class now exists only at runtime. Is there a simple way to go from the tool (

class Flake8(PythonToolBase):

) to whatever the concrete class of the UnionRule is ( whatever was used as the 2nd argument in`UnionRule(GenerateToolLockfileSentinel, Flake8LockfileSentinel)`) ? I could fish it out of the generated rules, but that doesn't seem like something Pants wants anyone to do.

Is there any way to stop coursier from running when I don't need the deps it provides?

⏺ oss/pants-pyright % ./pants --changed-since=origin/main fmt                                                                                ⎇ 17557-npx-dep-version*
⠠ 107.83s Fetching with coursier: org.scala-lang:scala-reflect:2.13.7
⠤ 107.73s Fetching with coursier: org.jline:jline:3.20.0
⠤ 107.73s Fetching with coursier: org.scalameta:scalafmt-interfaces:3.2.1
⠤ 107.73s Fetching with coursier: org.scala-lang.modules:scala-parallel-collections_2.13:1.0.4
⠤ 107.73s Fetching with coursier: org.scala-lang:scala-compiler:2.13.7

That went up to 3000 seconds before I cancelled it yesterday, and I'm formatting 3 python files in the nodejs backend

Where did we land last time on the pants package being an implicit vs explicit namespace package? (This is breaking the immut PR as we're trying to add files to the immutable source tree) CC @ancient-vegetable-10556 @hundreds-father-404 @witty-crayon-22786

I can’t remember what the end-result was, but whatever’s there now is what we decided to go with. Let me check 🙂

# NB: Mark this as an explicit namespace packages
__path__ = __import__('pkgutil').extend_path(__path__, __name__)

Fixes my tests, naturally. But wanted to see if that was verboten

OK. Yeah, there’s no

__init__.py

, which makes it an implicit namespace package. As I recall, marking it explicitly did not fix the thing I was poking around with, so we decided not to change. PyOx is no longer a priority, since SCIE looks like a much better fit, so if you’ve found a good use case for explicit namespace packaging, then that sounds great. @happy-kitchen-89482 was the one working with me on trying to get to the bottom of that.

tl;dr: no impediment to making that change, in my opinion.

We're not the only ones who might be in that situation though: might need to exclude those sources from being linked? Not sure if that's all Python sources, or only in some situations.

@witty-crayon-22786 I don't think I followed any of that 🙈

Whether or not pants itself uses namespace packages, other consumers might be using them, right?

Consumers of...? The reason it's breaking the immut PR is because our test config is trying to

touch

an

__init__.py

file in the source tree

ahhh.

got it, sorry.

yea, fine with me.

Which some Pants users might be trying to write to their source tree. In that case... ummmm... I guess we didn't say that was kosher? Unclear

we may end up needing the “force mutable” flag exposed during the release cycle, yea.

Even still, users don't have that toggle so I fear this should be fixed (if at all) differently.

could dogfood a “force mutable” flag here, or independently change the namespace package thing… i think that i lean toward the latter given how weird the test situation is.

Yeah I personally like explicit namespace packages because "explicit is better than implicit" 😉

And then we figure out the immutable behavior completely separately

New 🧵 for immutable inputs symlinking source dirs

Is there a way to test subsystems (and only subsystems)?

class CowsayTool(NpxToolBase):
    options_scope = "cowsay"
    name = "Cowsay"
    # Intentionally older version.
    default_version = "cowsay@1.4.0"
    help = "The Cowsay utility for printing cowsay messages"

I want to ensure that the version option is present, but the

create_subsystem

utility doesn't seem to do any checking to see if fields exist or not:

cowsaytool = create_subsystem(CowsayTool, versionXYZ="cowsay@1.5.0")

My hope was to do something like this:

rule_runner.set_options(["--backend-packages=['pants.backend.experimental.javascript']", "--cowsay-version='cowsay@1.5.0'"])

and then somehow get the subsystem and verify that the version was picked up correctly. Would I only be able to do this via the

run_pants

mechanism (which seems heavy for what I want)