Python-language-server: Language Server seemingly stuck at analysis stage

[DonJayamanne]

The status bar does not change from 695 items left....

Here's the log file:
log.log

[aniaan]

I also have the same problem

[zhen8838]

Ubuntu 18.04 languageServer.0.2.41 have same problem
two hours later ...:

[faassen]

In at least my case it's not just seemingly stuck but it also takes up a lot of CPU.

[AlexanderSher]

@DonJayamanne , what module is opened in your case?

[DonJayamanne]

Sorry don't have that info anymore.

[marconfus]

Same here on Ubuntu 16.04.
If I open an empty project with system Python 2.7.12 it looks ok. Just typing some import lines (os or requests for example) it's also ok.
But as soon as I select a VirtualEnv with Python 3.6 and do the same, the Microsoft.Python.LanguageServer process seems to get stuck in an endless loop and uses 100% cpu.
I guess this started after upgrading to the latest version.

[scorphus]

I think I probably have reproducible steps:

1. Create and activate a new Python 3.7.2 virtualenv (I used pyenv)
2. Install packages as detailed in this requirements.txt file (pip install -r requirements.txt)
3. Open vscode and select the interpreter created on step 1 above
4. Create and save a new file with this one line: import boto3

Sorry for a rather big requirements.txt file and for not being able to nail it down to a few packages at the moment. I'll update this issue if that changes.

If anyone could confirm those steps reproduce the issue, I'd be grateful.

Thanks!

[jakebailey]

We have a repro for this (but thanks either way!); it's just idna. https://github.com/Microsoft/python-language-server/issues/832#issuecomment-479609316

[jakebailey]

0.2.42 has been built and pushed to all of the channels (containing #873); it should fix the hangup introduced in 0.2.41.

[blackrobot]

@jakebailey this has fixed the issue for me in Microsoft/vscode-python#5100. Thank you!

[andrew-weisman]

How do we apply this fix? Thanks very much!

[jakebailey]

Reload VS Code, and if that doesn't update it, you can delete the downloaded language server and have the extension download it again. See Microsoft/vscode-python#3977.

[AlexanderSher]

@zhen8838 , @BeanNan , @marconfus - can you please check if the issue is fixed for you?

[zhen8838]

Yes, Now analysis successful. The analysis speed is now much faster. but I don't know why something can't autocomplete:

I have to import tensorflow.python as tf

Anyway, I am still very grateful for your work. thanks so much.

[andrew-weisman]

For me I'm still stuck at "Analyzing in the background, X items left..." where X is unchanging for a couple hours, but it appears that less CPU is being used. Thanks for the fix so far!

[AlexanderSher]

@zhen8838 , this is probably #867

@andrew-weisman , can you enable full logging ("python.analysis.logLevel": "Trace"), copy full log before hanging and attach it to the bug as a file?

[jakebailey]

I don't think that'd be #867; the new code should still bail when it sees the complicated comprehension that defines __all__ in tensorflow. I'll check tomorrow to be sure.

[zhen8838]

Thanks for your efforts, now I working with python-language-server so good.

[lonelygo]

@AlexanderSher @jakebailey Thanks for your efforts.
For me, after upgrade languageServer to 0.2.43, the analysis ends in about 5~7 minutes, but I got some problems like @zhen8838 https://github.com/Microsoft/python-language-server/issues/869#issuecomment-479729081 .

1 - Launch VS Code with shell command in iTerm2.
_1.1 - languageServer:_

Analysis stopped after 5 to 7 minutes, but CPU and RAM usage doesn't look good.

_1.2 - Intelligent Code Completion_
IntelliSense does not work properly,I try:

import tensorflow as tf
import pandas as pd

It seems that IntelliSense is not working for tensorflow, but working for pandas.

My python virtual environment is correct and the code works.

BUT parameter info for tf looks working :(

I don't know why, that's it.

2 - Launch VS Code.app directly.
It seems that code can't find the python virtual environment, and analysis seems not working.

I tried debugging and the code worked fine.

[marconfus]

@zhen8838 , @BeanNan , @marconfus - can you please check if the issue is fixed for you?

Yes, the bug is gone now. Thanks!

[AlexanderSher]

@jakebailey:

[jakebailey]

Dang, okay. I'll file a separate issue for that and look into it. The tensorflow __all__ should have been marked as unusable.

[jakebailey]

Opened as #881 to keep this thread about the analysis getting stuck rather than unrelated issues.

[zhen8838]

I don't know why it can't Intellisense Class init parameter :

[andrew-weisman]

Just FYI it works for me now, i.e., the background analysis doesn't hang forever and finishes in about 30min or so, and the CPU usage is way down overall. Thanks for all your work!

[MrTango]

It still stays with hight CPU load even on small project like this Pyramid example project: https://github.com/Pylons/pyramid_blogr.
Just check it out and install it in a Python 3 virtualenv with pip install -e .
After that when ever i open one of the Python files, this process keeps working on 100% for ever: .vscode/extensions/ms-python.python-2019.3.6352/languageServer.0.2.43/Microsoft.Python.LanguageServer
I often work on bigger projects based on the CMS Plone, there VS Code is currently kind of useless.

[jakebailey]

@zhen8838 please file a separate issue, as that's unrelated to the analysis getting stuck.

@MrTango The analyzer still needs to inspect dependencies; even if the project is small the imports still need to be inspected. We're still working on more ways to improve performance, so things should continue to improve.

[MrTango]

@jakebailey thx for the info. I'm aware of the fact that VS Code needs to analyse the libraries as Pyramid too, but that is the case for pretty much for every project. And Pyramid is not a big library not talking about frameworks like Django & co. Currently as much as i like VS Code in general, I must say for Python development it's not very much useful. Even the basic goto definition of Sumlime test works better and with the anaconda plugin the autocompletion just works. VS Code is heating the room for a long while and still eats up more than 10GB memory after a while. So i gues until the caching and performance issues are really solved, I'll switch back to sublime.

[DonJayamanne]

@MrTango

. So i gues until the caching and performance issues are really solved, I'll switch back to sublime.

You can still use VSCode with a work around, just change the following setting in your settings.json to turn off the language server and use Jedi:

"python.jediEnabled": true

[MrTango]

Not really, if I do that, jedi is driving the CPU the same way and worse, the autocompletion is blocking even normal snippets. :(

[MikhailArkhipov]

@MrTango - yes, Jedi sometimes has CPU and RAM issues as well. Is your code uses just Pyramid or anything else as well? Is it Plain Python or Anaconda? What Python version is it?

[MrTango]

Hi, it's plain Python 3 in this case, the example project i was talking about has just a view dependencies https://github.com/Pylons/pyramid_blogr/blob/master/setup.py
But I'm developing a lot with bigger stack like the CMS Plone or the ERP Odoo.
The Pyramid project I'm currently working on is bit bigger than the example, but even the smal project is slow in VS Code that's what i wanted to show. It might be also simpler to figgure out why on that code base, than on bigger ones. I gues i really need to wait for the caching feature, you talked about for the future.
but thx a lot for your effort, I'll keep an eye on vs code ;)

[MikhailArkhipov]

Since original issues appear to be resolved, closing this one. Opened https://github.com/Microsoft/python-language-server/issues/936 on @MrTango case. Please open separate issues with specific configurations if you are still experiencing problems. Thanks.

[raaaaaymond]

The original issue of hanging during "analyzing in background". Same symptoms as originally described, and I'm on server version 0.2.66.0. I think I've pin-pointed which package import is causing it, but unfortunately I can't reveal which one it is. It's a very small package though, so I'm surprised it's causing this problem.

I'm on MacOS Mojave with Pyenv Python 3.7.2.

[AlexanderSher]

@raaaaaymond , can you try LS 0.2.74 from beta channel?

[dzpt]

using LS 0.2.74 and today i got it.

[MikhailArkhipov]

@aaveidt, @raaaaaymond - when you say 'stuck' do you mean it is actually running (i.e. CPU consumption is visible) or CPU is 0 and it is just the stale message?

[raaaaaymond]

TLDR: SOLVED for now. VSCode 1.34.0, languageServer 0.2.82.0.

@MikhailArkhipov I mean the CPU was running, but nothing happens for hours and hours.

Update @AlexanderSher I never tried languageServer 0.2.74 because I just switched to Jedi for a while (which has its own problems), but recently I upgraded to VSCode 1.34.0, with languageServer 0.2.82.0, and it's working fine for me for now.

[petrosschilling]

The issue is still here.
Any workarounds?

I don't understand why is there 2 processes running.

[jakebailey]

This issue is old, and the langauge server has had many changes and a version bump since then. Your issue may be different.

Please follow the instructions at this section and file a new issue with logs and some more info, thanks: https://github.com/microsoft/python-language-server/blob/master/TROUBLESHOOTING.md#filing-an-issue

[kaushall]

This issue occured again. Language Server always stuck at analysis stage. Please check and provide possible solution as soon as possible.

Environment data

• VS Code version: 1.36.0 (Stable Channel)
• Extension version (ms-python.python): 2019.6.22090
• OS and version: Ubuntu 18.04 LTS
• Python version (& distribution if applicable, e.g. Anaconda): 2.7.15
• Type of virtual environment used (N/A | venv | virtualenv | conda | ...): virtualenv

@jakebailey @DonJayamanne @MikhailArkhipov

[jakebailey]

See my previous comment; we need some code sample, list of requirements, etc, in order to reproduce this and figure out if something's wrong.

[kaushall]

@jakebailey As per your instruction I've created a new issue with details, Please help.

Here is the url: https://github.com/microsoft/python-language-server/issues/1291

[petrosschilling]

Hey @kaushall a temporary solution for you:
uninstall IntelliCode and enable jedi on settings.json
"python.jediEnabled": true,

[kaushall]

@petrosschilling Thanks, Yes I'm using jedi right now.

[raaaaaymond]

I confirm I'm having the same problem again immediately after updating to VSCode 1.36.0. I've posted on @kaushall thread #1291 with more details.

For the moment I can use Jedi but it doesn't work as well with VSCode as does the official Python extension.

[jingyulong]

1.Run the Extenesions: Open Extensions Folder command
2.Quit VS Code
3.Delete the folder containing the Python extension (name should have a prefix of ms-python.python)
4.Reinstall Python Extension

This step is work for me!

[nnay13]

What worked for me :

1. Change from virtualenv to default python 3 interpreter .
2. Close and restart VSCODE
3. Change back to virtualenv.

[AlJohri]

I'm running into this with the remote extension. I'm using Remote SSH with the Python plugin installed on the remote instance. It's just constantly stuck at "Analyzing in background, 3888 items left..." and PyLint works super slowly which I think is related. It will keep underlining code that is already changed with stale messages.

[MikhailArkhipov]

@AlJohri - please don't reuse closed issues. Per @jakebailey comment above open separate issue as each case is different and we need to track which change/PR relates to a particular problem. Thanks