Huge memory usage in Linux and OSX (Not tested in windows)

I have installed this module in Ubuntu 16.04 and Mac OSX El Capitan. In both OS, when I open a python file and start typing, a python process is spawned and the memory usage of the process keep increasing until it saturates all the RAM. This goes away immediately when I disable the module.

Is it with same project or different ones? Can you enable logging in package settings and look what's going on?

It happens when autocomplete-python start to serialize requests to jedi. I just realize that it only happens when wx is used.

Actually I've meet the same problem in Windows 8.1 Python 2.7.12.
It happens for all my different projects.
When I enable the package autocomplete-python, 20mins later, the process python.exe will use huge memory and CPU usage. When disable the package, all work well.

As long as I cannot reproduce it I'm afraid I cannot do anything.

@sadovnychyi

@sadovnychyi try to disable "Show Descriptions"

@ivasonn

@ivasonn wow, that's useful info, thanks.

I can confirm that with enabled "Show Descriptions" I have ~50mb memory usage after typing import and I'm getting list of all my packages in less than a second, while with disabled "Show Descriptions" it takes tens of seconds and uses 1.3gb of ram.

Don't try to build function signatures for modules

Fixed the huge difference in time while loading list of modules after you type import keyword.

time python lib/completion.py '{"id":"0","prefix":" ","lookup":"completions","source":"import ","line":0,"column":7,"config":{"showDescriptions":false}}'
real    0m1.291s
user    0m1.037s
sys 0m0.172s

time python lib/completion.py '{"id":"0","prefix":" ","lookup":"completions","source":"import ","line":0,"column":7,"config":{"showDescriptions":true}}'
real    0m1.078s
user    0m0.919s
sys 0m0.147s

It took tens of seconds before fix to get a completion with descriptions turned off.

It's still going to take a while for huge modules like tensorflow though. With both config variations.

time python lib/completion.py '{"id":"0","prefix":" ","lookup":"completions","source":"import tensorflow\ntensorflow.","line":1,"column":11,"config":{"showDescriptions":false}}'
real    0m30.721s
user    0m27.276s
sys 0m1.437s

time python lib/completion.py '{"id":"0","prefix":" ","lookup":"completions","source":"import tensorflow\ntensorflow.","line":1,"column":11,"config":{"showDescriptions":true}}'
real    0m23.345s
user    0m21.873s
sys 0m0.776s

Please try it with 1.8.10. Should be better now if you have "Show Descriptions" turned off.

I have been experiencing the same problem on Windows. Anytime I tried to use autocomplete on an object that was subclassed from a wx object, within seconds, all my ram would be gone. But, i realized I had two versions of wxpython installed (2.8 and 3.0). Uninstalled the 2.8 version and have no problems now.

I've been having a similar issue.

I use ubuntu 16.04 and the autocomplete process sometimes ramps up to 100% CPU load and starts eating up all of my RAM at a pretty steep rate.

I've pinned it down to when I'm working with the gnuradio module which is probably quite big.

I've been experiencing this too, my Macbook's fans started to make noise. When I checked Activity Monitor, a python process was consuming almost 100% CPU.

output of : ps aux | grep python

USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
myuser 55563 100.0 0.0 2484232 852 ?? R 5:31PM 183:59.62 /Users/myuser/Projects/env/bin/python /Users/myuser/.atom/packages/autocomplete-python/lib/completion.py

does it have anything to do with python virtual environments?

See related Jedi issue above. It's likely because of a huge package you use.

…

On Tue, Feb 21, 2017, 15:17 Dev Uberoi ***@***.***> wrote: I've been experiencing this too, my Macbook's fans started to make noise. When I checked Activity Monitor, a python process was consuming almost 100% CPU. output of : ps aux | grep python USER PID *%CPU* %MEM VSZ RSS TT STAT STARTED TIME COMMAND myuser 55563 *100.0* 0.0 2484232 852 ?? R 5:31PM 183:59.62 /Users/myuser/Projects/env/bin/python /Users/myuser/.atom/packages/autocomplete-python/lib/completion.py does it have anything to do with python virtual environments? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#233 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAL1SB1eLPYmnQHLuD4_3epeAuiSg3tXks5reo-dgaJpZM4J7EFT> .

@andrasikladi unfortunately right now there's no way to limit the Jedi in number of completions to returns, see that issue. Until it's fixed, there's nothing I can do about it.

Maybe you could modify the Regex to trigger Autocompletions setting to ignore completions from that specific module, but I'm not sure that it will work. (e.g. there's no context available, only current word)

I have the same issue when working on a project that uses wx package.

I am having the same issue- not using jedi (unless this installs it in the background) though. Ubuntu 17.04 and autocomplete-python 1.10.2. Atom 1.19.3. Mine is both memory and 100% CPU for each instance of atom, i.e. 4 instances, 4 cores at 100%.

The first time I tried this, it hard-froze my Ubuntu 16.04 laptop. Subsequent tries worked, but used almost all of my 8GB of RAM until I disabled "Show Descriptions".

Maybe, until this is fixed, "Show Descriptions" should be disabled by default.

To reproduce:

install the package, keep default options
copy code (yes, it uses wx) from http://upgrayd.blogspot.de/2010/11/python-mandelbrot.html
near the top of the file, start a new line and type: self.
watch your RAM usage rapidly climb

Do you have any news about this topic?

Even with "Show Descriptions" disabled, the issue is still present.
The only way to solve it, it's to disable autocomplete-python for projects which use wx.

Cannot do anything until davidhalter/jedi#805 is fixed. Even if you run pure Jedi on wx completion it will eat your memory and CPU.

It's also currently not possible to use triggerCompletionRegex setting to ignore some package completions, since the prefix provided by Atom excludes everything after a dot. E.g. for wx. the prefix is just a dot. So you can either completely disable completions on dot by removing the dot match from that regexp, OR we could try to modify the prefix on the fly so you could be more flexible on what you can ignore with that regexp.

The only solution right now is to disable autocomplete after a dot character by using this regexp:

([ (]|[a-zA-Z_][a-zA-Z0-9_]*)

Put it into Regex To Trigger Autoocmpletions setting. This should limit the number of completions you get. (hopefully! didn't try it by myself)

The Visual Studio Code extension for Python, which also uses Jedi, just dealt with this by restarting the competition process if it uses more RAM than a maximum defaulting to 1 GB (microsoft/vscode-python#744).

Add shebang interpreter discovery, lock down constructor, async and d…

Navigation Menu

Repository navigation

Huge memory usage in Linux and OSX (Not tested in windows) #233

4 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Huge memory usage in Linux and OSX (Not tested in windows) #233

Description

Activity

sadovnychyi commented on Sep 13, 2016

ricwtk commented on Sep 13, 2016

chxxpeng commented on Sep 22, 2016

sadovnychyi commented on Sep 22, 2016

ivasonn commented on Nov 2, 2016

sadovnychyi commented on Nov 2, 2016

sadovnychyi commented on Nov 2, 2016

sadovnychyi commented on Nov 2, 2016

4 remaining items

lunchboxmg commented on Dec 20, 2016

seddarj commented on Feb 17, 2017

devuberoi commented on Feb 21, 2017

sadovnychyi commented on Feb 21, 2017

sadovnychyi commented on May 10, 2017

mkatliar commented on Jul 21, 2017

grantstephens commented on Sep 4, 2017

bitinerant commented on Sep 14, 2017

anxuae commented on Nov 7, 2017

sadovnychyi commented on Nov 7, 2017

elliott-beach commented on Feb 10, 2018

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions