Repository navigation
[Feature Request] More performant server #17
Description
Hi. Python language server is well written, thank you for your contribution.
However, after using I find some improvements can be done.
-
logger
Tracked by [Suggestion] Use File based logger #15. Logging to stderr will block stdout writing. -
threaded reading
Now PLS in stdio mode is single threaded.
So if client sends too many requests than stdin can buffer, PLS will also block because it cannot write to stdout. (It seems stdin/stdout/stderr share one common buffer) -
cached compeletion
PLS generally is slower than deoplete-jedi, another jedi based completion plugin. The main reason is deoplete-jedi uses a cache. https://github.com/zchee/deoplete-jedi/blob/master/rplugin/python3/deoplete/sources/deoplete_jedi/cache.py If PLS can also support cache it should be faster.
Thanks!
Activity
gatesn commentedon Mar 7, 2017
+1 to all three here!
HerringtonDarkholme commentedon Mar 7, 2017
Hi @gatesn! Thanks for quick reply.
Currently I use PLS with neovim. But I would like to try it out with VSCode like that in README. (Fairly, I cannot tell the behavior I see is due to client or server, so I need another client to compare.)
However I cannot find extension in VSCode market. Would you like provide a installation manual?
Or do I need create a wrapper extension for PLS? Thanks!
gatesn commentedon Mar 7, 2017
@HerringtonDarkholme nice idea. I've noticed a few language servers do come with wrapper extensions. Would be happy to look into it some time, could you create a separate issue for it? Thanks
gatesn commentedon Jun 29, 2017
@HerringtonDarkholme thanks for raising these points. I want to close this ticket and instead track individually any ideas for performance improvements. Here are some to get us started:
performance