[WIP/RFC] Only use Jedi, without worker/client/server/cache by blueyed · Pull Request #184 · deoplete-plugins/deoplete-jedi

blueyed · 2018-07-08T19:47:03Z

I have tried to implement restarting in case
g:deoplete#sources#jedi#python_path gets changed, and figured that it
is way easier when using only Jedi.

There are likely previous features missing, and it needs cleaning up /
deleting still.

It uses the Jedi branch to improve handling of Environments:
davidhalter/jedi#1108

TODO:

jedi.settings.additional_dynamic_modules
revisit existing features, e.g. the boilerplate handling
consider keeping the cache itself, only remove Worker/Client/Server
allow to control environment using some variable, possibly wrapped into a command to select it (ref: [WIP/RFC] Improve selection of environments davidhalter/jedi-vim#836)

Works quite well already, tested with:

import numpy

numpy.

2018-07-08 23:25:51,485 DEBUG    [14486] (deoplete.jedi) Line: 3, Col: 6, Filename: '…/vim/plugged/jedi-vim/pythonx/jedi/t.py'
2018-07-08 23:25:52,442 INFO     [14486] (deoplete.jedi) get_completions t = 956.456334ms, µ = 0, σ = 0)
2018-07-08 23:25:53,194 INFO     [14486] (deoplete.jedi) massage_completions t = 752.618779ms, µ = 0, σ = 0)
2018-07-08 23:25:53,195 INFO     [14486] (deoplete.jedi) gather_candidates t = 1712.006350ms, µ = 0, σ = 0)
2018-07-08 23:26:01,394 DEBUG    [14485] (deoplete.core) on_event: InsertEnter
2018-07-08 23:26:03,061 DEBUG    [14485] (deoplete.core) completion_begin: 'numpy.' (TextChangedI)
2018-07-08 23:26:12,425 DEBUG    [14485] (deoplete.core) on_event: InsertEnter
2018-07-08 23:26:12,944 DEBUG    [14485] (deoplete.core) completion_begin: '' (TextChangedI)
2018-07-08 23:26:13,003 DEBUG    [14485] (deoplete.core) completion_begin: 'n' (Async)
2018-07-08 23:26:14,475 DEBUG    [14485] (deoplete.core) completion_begin: 'numpy.' (TextChangedI)
2018-07-08 23:26:14,483 DEBUG    [14486] (deoplete.jedi) Line: 5, Col: 6, Filename: '…/vim/plugged/jedi-vim/pythonx/jedi/t.py'
2018-07-08 23:26:14,534 DEBUG    [14485] (deoplete.core) completion_begin: 'numpy.' (Async)
2018-07-08 23:26:14,702 INFO     [14486] (deoplete.jedi) get_completions t = 219.013144ms, µ = 0, σ = 0)
2018-07-08 23:26:15,270 INFO     [14486] (deoplete.jedi) massage_completions t = 567.887416ms, µ = 0, σ = 0)
2018-07-08 23:26:15,270 INFO     [14486] (deoplete.jedi) gather_candidates t = 791.059430ms, µ = 0, σ = 0)

So massage_completions takes a lot of the time by itself.

It might make certainly sense to add some caching on top, but it shouldn't be so complicated anymore.

/cc @tweekmonster

blueyed · 2018-07-08T21:47:23Z

-            'deoplete#sources#jedi#python_path', '')
+        # TODO
        self.extra_path = vars.get(
            'deoplete#sources#jedi#extra_path', [])


Jedi has jedi.settings.additional_dynamic_modules (used in jedi-vim).

Shougo · 2018-07-08T23:03:10Z

I think it is better version.
Because, deoplete is already parallel.

blueyed · 2018-07-08T23:29:57Z

Yep, I like it also so far.

Shougo · 2018-08-13T23:20:30Z

@blueyed
I want to test the branch.
Please rebase it.

Shougo · 2018-08-13T23:21:50Z

+            if comp.module_path not in cache and comp.line and comp.line > 1 \
+                    and os.path.exists(comp.module_path):
+                with open(comp.module_path, 'r') as fp:
+                    cache[comp.module_path] = fp.readlines()


I get the slow reason.
It is very slow. Because, deoplete-jedi loads all module files.
I think it should be removed.

It is faster and simpler.

Yes, likely only needed for some old and hopefully fixed misbehavior with regard to @property decorators.

Shougo · 2018-08-13T23:22:55Z

So massage_completions takes a lot of the time by itself.

I get the reason. Please see the comment.
I think the cache feature is not needed.

I have tried to implement restarting in case `g:deoplete#sources#jedi#python_path` gets changed, and figured that it is way easier when using only Jedi. There are likely previous features missing, and it needs cleaning up / deleting still. It uses the Jedi branch to improve handling of Environments: davidhalter/jedi#1108

[ci skip]

…06a0 * rplugin/python3/deoplete/vendored/jedi f6bc166...d6306a0 (48): > With the recent changes one performance optimization got lost > FunctionExecutionContext should use the parent if possible > Use anonymous instance arguments in a different way > Move some anonymous instance function execution stuff > Get rid of InstanceFunctionExecution, because it's really not needed > Use the InstanceArguments for super as well > Use InstanceArguments directly and not via InstanceFunctionExecution > Remove old garbage code > Don't use arguments that are not needed > Also move the remaining get_params to get_executed_params > get_params -> get_executed_params where possible > Subprocess error reporting improvements > Fix a recursion issue about compiled objects > Use a CompiledInstanceNameFilter that wraps the class name as well > Prefer Python 3 import over 2 > Now it's actually possible to specify a pytest environment for the same Python version > Note that Python 3.3 support was dropped in Changelog > Drop support for EOL Python 3.3 (#1019) > Rewrite the pyc test > Fix an issue with stderr debugging of subprocesses > stderr of the child processes should be printed in debug output > Use close_fds for posix. > Remove some redundant code > Use names of classes to infer names of instances > Don't have execute and execute_evaluated on name > Fix an issue where __ prefixed variables where not hidden when accessed from a class > Bound methods are now working correctly in all Python versions. Therefore a test was wrong. > Remove a print in tests > BoundMethods now have access to the function that they are using > Remove another usage of is_class where it's not needed > FunctionContext should be created from a unified interface > Don't create a FunctionExecutionContext if it's not used. > Use TreeContext in a good way > Fix broken link in documentation > The implicit namespace package test from 4b276bae87a3170672f7ddb3e00f5851fe24d562 can only be used for Python 3.4+ > The import resolution for namespace packages was wrong > Add a way to use the interpreter environment for tests > MergedExecutedParams -> DynamicExecutedParams > Fix a recursion error, fixes #1173 > Remove two recursion tests again that will belong into a commit at a point where it is not failing anymore > Don't merge params if it's just one param > Add a repr for AnonymousArguments > Fix a doctest > Some minor flake8 fixes > Script.__repr__: include environment > Add params to CallSignature.__repr__ > Environments are now always created on request > Improve Environment

…f1582 * rplugin/python3/deoplete/vendored/parso 7064ecf...96f1582 (1): > Update usage.rst

blueyed · 2018-08-14T02:13:04Z

@Shougo
You can pick it up, and either rebase it only locally first, or also push here.

But "Allow edits from maintainers." might not be enough.. :/

Ping @zchee.

Rebased it now, updated the submodules, but no other changes.

blueyed · 2018-08-14T02:13:27Z

@Shougo
Maybe you want to create a new PR then - I won't have time for it currently.

Shougo · 2018-08-16T23:44:23Z

Maybe you want to create a new PR then - I won't have time for it currently.

OK. I will do it.

Shougo · 2018-08-18T06:02:20Z

I have created #192.

blueyed · 2018-09-13T05:01:49Z

Closing in favor of #192.

blueyed mentioned this pull request Jul 8, 2018

[RFC] Upgrade Jedi #179

Merged

blueyed added the enhancement label Jul 8, 2018

blueyed commented Jul 8, 2018

View reviewed changes

blueyed force-pushed the jedi12-light branch from edd94db to f98e854 Compare July 8, 2018 23:49

blueyed mentioned this pull request Jul 8, 2018

Environment: improve/fix getting sys.prefix davidhalter/jedi#1108

Closed

blueyed force-pushed the jedi12-light branch 2 times, most recently from d95b57c to ce768aa Compare July 10, 2018 10:58

blueyed mentioned this pull request Aug 10, 2018

deoplete-jedi does show proper module completions on import statements like "from x.y" #191

Closed

Shougo reviewed Aug 13, 2018

View reviewed changes

blueyed added 10 commits August 14, 2018 04:11

Refactor for profiling information

615356b

[ci skip]

fixup! Refactor for profiling information

3eed398

fixup! fixup! Refactor for profiling information

e7f1ff7

fixup! fixup! fixup! Refactor for profiling information

b9b591f

fixup! fixup! fixup! fixup! Refactor for profiling information

fa86ca5

fixup! [WIP/RFC] Only use Jedi, without worker/client/server/cache

935894f

fixup! [WIP/RFC] Only use Jedi, without worker/client/server/cache

99cb101

Update submodule rplugin/python3/deoplete/vendored/parso 7064ecf...96…

331abab

…f1582 * rplugin/python3/deoplete/vendored/parso 7064ecf...96f1582 (1): > Update usage.rst

blueyed force-pushed the jedi12-light branch from ce768aa to 331abab Compare August 14, 2018 02:12

This was referenced Aug 16, 2018

Multi-line import with parenthesises not completing after line break #163

Closed

deoplete-jedi can not display properly with pandas library. #155

Closed

PyQt5 completion problem #151

Closed

Shougo mentioned this pull request Aug 18, 2018

Remove worker,server,cache #192

Merged

blueyed closed this Sep 13, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[WIP/RFC] Only use Jedi, without worker/client/server/cache#184

[WIP/RFC] Only use Jedi, without worker/client/server/cache#184
blueyed wants to merge 10 commits intodeoplete-plugins:masterfrom
blueyed:jedi12-light

blueyed commented Jul 8, 2018 •

edited

Loading

Uh oh!

blueyed Jul 8, 2018

Uh oh!

Shougo commented Jul 8, 2018

Uh oh!

blueyed commented Jul 8, 2018

Uh oh!

Shougo commented Aug 13, 2018

Uh oh!

Shougo Aug 13, 2018

Uh oh!

blueyed Aug 14, 2018

Uh oh!

Shougo commented Aug 13, 2018

Uh oh!

blueyed commented Aug 14, 2018

Uh oh!

blueyed commented Aug 14, 2018

Uh oh!

Shougo commented Aug 16, 2018

Uh oh!

Shougo commented Aug 18, 2018

Uh oh!

blueyed commented Sep 13, 2018

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

blueyed commented Jul 8, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blueyed Jul 8, 2018

Choose a reason for hiding this comment

Uh oh!

Shougo commented Jul 8, 2018

Uh oh!

blueyed commented Jul 8, 2018

Uh oh!

Shougo commented Aug 13, 2018

Uh oh!

Shougo Aug 13, 2018

Choose a reason for hiding this comment

Uh oh!

blueyed Aug 14, 2018

Choose a reason for hiding this comment

Uh oh!

Shougo commented Aug 13, 2018

Uh oh!

blueyed commented Aug 14, 2018

Uh oh!

blueyed commented Aug 14, 2018

Uh oh!

Shougo commented Aug 16, 2018

Uh oh!

Shougo commented Aug 18, 2018

Uh oh!

blueyed commented Sep 13, 2018

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

blueyed commented Jul 8, 2018 •

edited

Loading