Meta for tracking Language Server issues

[MikhailArkhipov]

List of issues we need to resolve in PTVS GitHub

• Implement module information tooltip (Implement information on modules PTVS#3849)
• Document formatting does not remove double spaces in imports (Document formatting does not remove double spaces in imports PTVS#3850)
• No completion for abc in editor (works in REPL) No completion for abc in editor (works in REPL) PTVS#3857
• Migrate special completions to server Migrate special completions to server PTVS#3798
• Hover tooltip does not show documentation Hover tootip for 'capitalize' does not show documentation PTVS#3868
• Signature display inconsistent Signature display inconsistent with range PTVS#3869
• Information is duplicated in completion item tooltips Information is duplicated in completion item tooltips PTVS#3882
• Support PEP style type hints Support comment type hints PTVS#3917
• Implement go to definition Migrate find definition to server PTVS#3803
• Support for import * can't PTVS complete function names imported by from import *? PTVS#3967
• Option not to trim tooltip documentation Option not to trim tooltip documentation PTVS#3977
• No return types for pandas/requests modules IntelliSense returning None types for pandas/requests library PTVS#4001
• No error for wrong number of arguments No error when calling function with missing arguments PTVS#4003
• First time completion list is almost always wrong b/c of analysis delay First time completion list is almost always wrong b/c of analysis delay PTVS#4032
• Relative imports Intellisense is broken in the presence of relative imports PTVS#949
• AF 'Accessing Mro before SetBases' AF 'Accessing Mro before SetBases has been called' PTVS#4126
• No completions for 'open' No completions for 'open' PTVS#4124
• No completion for list function argument No completion for list function argument PTVS#4130
• No documentation on imported class No documentation on imported class PTVS#4137
• No completion in namespace packages No completion in namespace packages PTVS#4138
• NamedTuple completions only show fields with default values NamedTuple completions only show fields with default values PTVS#3608
• .NET Core tests for analysis engine .NET Core tests for analysis engine PTVS#4185
• Better handle 're' module Better handle 're' module PTVS#4312
• Better handle time module Better handle time module PTVS#4318
• Better handle unittest module Better handle unittest module PTVS#4319

[brettcannon]

Why don't you want to track this in the project?

[MikhailArkhipov]

I need a single place to find all related PTVS links

[zooba]

FYI, I renamed "Support PEP style type hints" to "Support comment type hints". We already support PEP 484 style hints, but only in annotations.

[davidhalter]

Hi guys

I was wondering if you could explain a bit why you are pushing forward to use PTVS instead of Jedi (correct me if I'm wrong). IMO Jedi is still the better solution for a couple of reasons:

• Jedi is written in Python and has therefore better community support.
• Jedi has a long history (6+ years) with a lot of learnings and years of refactorings.
• Jedi understands dynamic features very well (list.append(), searching for function executions...)
• Jedi understands many of the weird import practices (old/new school namespace packages, eggs, zip archives, etc)
• Jedi can complete f-strings, async/await things and has generally a very good understanding of modern Python features.
• Since the latest Jedi version, Virtual Environments and different Python versions are perfectly handled in Jedi.
• Parso is generally a very fast parser, because it only parses the changes people made. This is unfortunately still the source of some errors, but it also makes parsing very fast in big files (3 kLOC+)
• Jedi is generally very well tested (~ 2k tests, 850 in parso).
• There are currently 196 bugs open in PTVS, 26 bugs being related to Intellisense; Jedi has currently 12 open bugs, 6 being related to "Intellisense" (6 being linter bug, which is not used by anyone). parso (Jedi's parser) has currently one open bug.

So what I'm asking is, is there a way we could make Jedi better so you would reconsider? Or is this decision driven by the idea that you only have one stack and only C# dependencies?

There are many features that have not been implemented in Jedi, but would easily be possible given time or appropriate resources: Refactoring, Projects, Better Usages, Better Django support...

BTW: There are a lot of features on Jedi's roadmap this year anyway (among them typeshed).

~ Dave

[brettcannon]

There are basically two reasons we are working on bringing the analysis engine from the Python workload in Visual Studio over to VS Code:

1. Consolidation to a single analysis engine (as you accurately guessed): since Visual Studio also needs an analysis engine we had to make a decision between Visual Studio or Jedi, and it's easier for us to change than it is for Visual Studio to switch to Jedi
2. Doing more than completions/IntelliSense: we want this new language server to do more than just IntelliSense to remove other third-party dependencies while also benefiting Visual Studio

IOW this has nothing to do with the job you have done with Jedi or its feature set -- now or in the future -- and everything to do with us and our specific needs.

[davidhalter]

Thanks for your answer. This is of course an understandable decision - even if it is a bit disappointing for me :)

Doing more than completions/IntelliSense: we want this new language server to do more than just IntelliSense to remove other third-party dependencies while also benefiting Visual Studio

If you're looking for a bigger feature set within Jedi, please also let me know. I can obviously not help you with your first point, but there are quite a few features planned that are not just InteliSense, like refactoring or auto-imports. It's just a question of what to prioritize.

[sbstp]

Does the PTVS work on Mac and Linux without having to install Mono? If not, it would be nice if the jedi provider was still maintained.

[brettcannon]

@sbstp we assume the user has no .NET runtime installed (even on Windows).

[brettcannon]

@davidhalter I emailed you privately about that (and I'm always happy to share what feedback we're getting to help you prioritize your work).

[MikhailArkhipov]

Closing this one as we are tracking issues directly