Open Pierre-Sassoulas opened 1 year ago
Makes perfect sense to me!
I thought about this for pyreverse
as well.
Do you know if we could publish it on PyPI under this name?
There already exists a package with this name, as it seems that anno 2005 pyreverse
was once standalone and afterwards migrated into pylint
.
As for the dependencies between pylint
and pyreverse
separating them should be possible.
The ImportsChecker
supports creating import graphs in dot format, but it uses pylint/graph.py
to do so.
graph.py
was once also used by pyreverse
, but after we have added the new output formats and refactored all Printer
related stuff it no longer has any connections to it.
I think being able to get the PyPI
name should indeed be a blocker. Andreas you can try emailing either Logilab or the other maintainer?
I'm not sure that's a true blocker, we could also use pylint-pyreverse
. It would be nice to have pyreverse
though. I created pylint-pyreverse
, pylint-spelling
, pylint-core
and pylint-testutils
just to guarantee those are not squatted. (I contacted Emmanuel Breton to ask about this via LinkedIn, it was a task well suited for a french person 😄).
Thanks for taking care of this Pierre!
Providing the following relevant link if we want to pursue the pyreverse namespace in the future: https://peps.python.org/pep-0541/#how-to-request-a-name-transfer. Not sure if applicable here as I'm pretty sure I contacted the current owner on LinkedIn (they did not answer yet).
Some pylint users do not use pyreverse, the dependencies of pylint-spelling are not handled well, and most of our user are not lib maintainer requiring our test utils. Now that we have an organization dedicated to pylint's development it's not unexpected that we would have a multi-repositories approach. Moving them to another repository would mean:
pylint-spelling
(Right now we catch import errors and do a lot of effort to not fail when enchant is not available)We can also filter the git history like what was done for epylint in the emacs orphanage (or testutils in a private PyCQA repository). The only negative I see with that is that we would need to do more releases (because right now everything is embedded in pylint). Releases aren't that hard to do and we would not need a maintenance branch for pyreverse, spelling or testutil, imo. We also need to decide how the transition is going to happen (optional way to install pyreverse/testutil like we did for spelling ?).
Opening this issue to gather feedback.