Autocompletion not working for python

[dtransposed]

Description

Autocompletion is not working for a code written in a script, however it works for Inferior Python.

Reproduction guide

• Start Emacs
• Open new python script(.py)
• Activate virtual environment, which contains the correct module (pyenv-mode-set, tested module is numpy, although I tried also with time)
• Write a sample python script:

  import numpy as np
  a = np.

• Try to get a completion list (hit TAB)

Observed behaviour Completions do not work. company-diag returns the log:

However, when performing same import and autocompletion in inferior mode (run-python), autocompletion works as expected. company-diag returns the log:

System Info

OS: darwin20.1.0 Emacs: 27.1 Spacemacs: 0.300.0 Spacemacs branch:

commit df8249a0522c2bda83cef476d41d55de3cfbec0f (HEAD -> develop, origin/develop, origin/HEAD)
Author: JAremko <w3techplayground@gmail.com>
Date:   Mon Apr 5 06:15:21 2021 +0300

Distribution: spacemacs Editing style: vim Completion: helm Layers:

I have tried solutions listed in following issues: https://github.com/syl20bnr/spacemacs/issues/11785 https://github.com/syl20bnr/spacemacs/issues/10638

Namely:

1. Putting this at your dotfile user-config:

   (eval-after-load "company"
   '(add-to-list 'company-backends 'company-anaconda))

2. Adding

dotspacemacs-additional-packages '(jedi)

Nothing works for me so far.

[lebensterben]

not related, but this should be the example for other bug reports. it's well described.

[dtransposed]

I have noticed super interesting thing. Every time I run: pip install --upgrade jedi json-rpc service_factory in the peen environment, the autocompletion starts suddenly working! Then I restart emacs and autocompletion breaks! I run the pip install... again and things are working once again. This is weird.

@lebensterben Thank you!

[lebensterben]

Do you have virtual environment?

[dtransposed]

Yes. I am using pyenv. For the purposes of this investigation I have created new environment testenv

1. I open emacs.
2. Before I open any file I activate this environment through pyenv-mode-set.
3. When I edit the file I can see that the name of virtual environment is present on the power bar.
4. pip install --upgrade jedi json-rpc service_factory is being run inside this virtual environment.

Edit: anaconda-show-doc also behaves funny, exactly like the autocompletion engine.

[alexey0308]

@dtransposed you use anaconda backend, which requires inferior python buffer created for auto-completion. Hence, first C-c C-p, then everything should work.

EDIT: Unfortunately, I can not find it in the anaconda documentation and it looks like it should work right away, but at the times I was using it, I had such problem, until the inferior-python-buffer is started.

[dtransposed]

@alexey0308 Thank you! It works like charm! Could you explain why this is the case? Do other backends have different behaviour?

[lebensterben]

https://github.com/syl20bnr/spacemacs/blob/3371a8220c8e48503dc6dd4e2d6ac532ba9946ad/layers/%2Blang/python/packages.el#L97-L109

Line 102 probably need to be changed to :modes python-mode inferior-python-mode for anaconda backend.

[alexey0308]

LSP starts its server automatically in the background, but you need to restart it if you activate the environment, so it catches up with the new configuration. I don't use anaconda anymore and I don't remember, where actually I read, that one needs to have a python process for anaconda to work. Maybe check the source code or docs or take it as a folklore knowledge :)

[lebensterben]

But it's actually handled here: https://github.com/syl20bnr/spacemacs/blob/3371a8220c8e48503dc6dd4e2d6ac532ba9946ad/layers/%2Blang/python/funcs.el#L64-L71

[lebensterben]

(If we drop support of all non-lsp backends, life would be much easier for all of us)

[alexey0308]

https://github.com/syl20bnr/spacemacs/blob/3371a8220c8e48503dc6dd4e2d6ac532ba9946ad/layers/%2Blang/python/packages.el#L97-L109

Line 102 probably need to be changed to :modes python-mode inferior-python-mode for anaconda backend.

@lebensterben this is for inferior buffer only: it is always on anaconda backend, regardless of lsp or not for the python-mode. For python-mode, you already found anaconda configuration in your second message.

[alexey0308]

(If we drop support of all non-lsp backends, life would be much easier for all of us)

Anaconda can work with tramp AFAIK, I don't know if lsp as well?

[lebensterben]

Anaconda can work with tramp AFAIK, I don't know if lsp as well?

A valid point. Setting up LSP over remote connection is doable, and I've actually seen someone mentioned this a few days ago. But it requires some setups.

See https://emacs-lsp.github.io/lsp-mode/page/remote/

But I really wish we can drop non-lsp support, for its many benefits:

• There are too much code just to deal with which backend the user is using, which makes Spacemacs bloated.
• With only LSP, a large part of layer configurations would be similar. And it's possible for us to abstract common setup routines out.
• Everything is handled by either LSP (and lsp-MAJOR-MODE), or a minor mode. We no longer needs to worry about key bindings collisions in many cases, since we don't need to maintain two sets of functions for the same functionality.

[alexey0308]

I understand, it's more work to support it, especially when there is no dedicated maintainer for every layer. But it's rather a political decision: you see people using anaconda for some reason, or e.g. I use ess because I like to have completion based on the inferior buffer state as well, and lsp infers only based on the code base, right? Trade-off between maintaining and users ¯_(ツ)_/¯

[smile13241324]

One point against LSP is that it normally takes much more system load than traditional emacs modes. Still the ecosystem definitely goes into the direction of LSP. I dimly remember we already have layers supporting only LSP right now.

[lebensterben]

Please retry with one of the LSP backend.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Please let us know if this issue is still valid!