Open liavkoren opened 7 years ago
Hi @liavkoren Thanks for opening this issue! The Jedi completion runs indeed in another thread. Would the following fix work: https://github.com/jonathanslenders/ptpython/pull/224 ?
It's not the cleanest solution, but otherwise I have to keep adding exceptions for everything that Jedi will throw.
Oh, just saw this. Would you like me to pull the commit and check it locally?
Okay, I'm tots confused. I cloned the repo and pip installed -e, but I can't reproduce the jedi error even on master
. Numpy, prompt-toolkit and jedi versions all match.
Ok. No problem. I'll probably merge it anyway. These Jedi errors are pretty annoying, and right now, I don't have a good way to log what's going wrong without disturbing the end user.
Thanks!
Note: the exception happens immediately on typing the period in
(a - b > 0).
numpy==1.10.4 prompt-toolkit==1.0.15 jedi==0.11.0
I deeply love ptipython (it's my default python repl) and, you know, god[s]/time-management willing, I'm happy to try and help.
It's not super rare for me to run into Jedi exceptions. Overall, those aren't actually a big deal, but what is kind of a drag is then not having access to autocomplete for the rest of the life-span of the parent ptpython session. I'm assuming Jedi is being run in a child thread, and when it hits an unhandled exception, it takes autocomplete out. How much of a PITA would it be for a new Jedi to be created when that happens?