Repository navigation
Preserve comprehension scopes in full_name - #2124
Conversation
Signed-off-by: Don Yin <Don_Yin@outlook.com>
|
How much LLM is this? Do you understand what you have done here? |
|
Fair question, and quite a lot. I've been trying to give back to the Python tools I rely on (jedi sits under pi-lens which is a code checker I contribute to and use quite heavily, through jedi-language-server) I search and check the issues where a maintainer has already said what fix they'd accept. I used my agent to find #1944 and draft the patch and tests by your words I then reviewed the change myself and checked the comparison image in the description against a real run, which you saw above And yes, I understand it But to be honest this issue hasn't been causing anyhthing noticible on the side of the pi-lens for me, and not a big deal if you'd rather close this; I'm happy to |
|
Thanks! I would personally recommend you to use Zuban instead of Jedi, because that's where most of my work flows into nowadays. The language server is in a much more solid state, offers diagnostics, auto-import and is much faster than Jedi. It's essentially Jedi 2.0 (if you're not using IPython in which case Jedi is still the only option). |
I reproduced this with the
pig_latinexample in #1944. I changed comprehension contexts to keep the enclosing scope, so both occurrences ofwordnow report__main__.pig_latin.<comprehension>.word, as suggested. I added that example as a regression test; the separate question about resolvingtranslateremains.Fixes #1944
I ran the full pytest suite, flake8 and zuban; the regression tests and lint passed, with the existing readline test failure and
utils.pytype errors reproduced without the change.