Experimental project structure #773
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Closed
scohen
force-pushed
the
scohen/references-request
branch
2 times, most recently
from
9c7e44e
to
5322d1d
Compare
scohen
force-pushed
the
scohen/references-request
branch
2 times, most recently
from
a5f79da
to
1a3692f
Compare
scohen
force-pushed
the
scohen/references-request
branch
2 times, most recently
from
48fd12f
to
443b959
Compare
This was referenced Jan 1, 2023
This commit represents a new structure for the experimental project, and
a path forward. With these changes the project now has:
* A build option enabling the experimental server
* Per-message routing for the experimental server. If enabled, it can
either share messages with the existing server or take them over.
Presently, the find references and formatting providers are
implemented and "exclusive", meaning that they're handled solely by
the experimental server
* A consistent interface for building providers.
* A consistent way to convert lsp messages into data structures and
back again. This conversion is handled automatically for providers.
* A genserver-like interface for providers to implement
* Data structures representing LSP messages that are simple to define
and build.
* Fast and efficient conversion between utf8 and utf16.
* A separation between what a provider does and how it responds to
messages. This allows the work that underpins providers to be tested
independently from the language server.
Created a code action that prepends an underscore to unused variable names.
First attempt at a standard interface for code modification. Code mod modules take the original text, the ast of the original text and arguments that they specify. They return a list of code edits or an error.
The problem was that the character positions are _before_ the reported unit, so the 0th code unit is before the start of the line and the 1st code unit is the first character. The prior code added one to character counts to smooth this out, but you can't do that, because you could end up indexing into the middle of a multibyte character.
Code mods deal with snippets of code that need to have their line numbers fixed up by the code actions.
The AST type is very complicated, and dialyzer was telling us I got it wrong.
While working on the automatic protocol generators, it became clear that type aliases needed to be their own thing, as they operate quite differently from the other defined things in the jsonrpc protocol. Since they're just aliases, it makes sense to keep their definitions on hand and then spit them out when other things make use of them during encode and decode. This did require going back to encoding and ensuring all the encode functions return OK tuples.
When patches are unapplied, getting the beam file returned an empty path charlist, which dialyzer assumed was a real file name due to a weak assumption, which caused unit tests to fail. This was remedied by checking for a non-empty charlist, which allows tests to succeed. Also made patch a test only dependency for .formatter.exs, as this was causing formatters to fail.
scohen
force-pushed
the
scohen/references-request
branch
from
443b959
to
7fde975
Compare
Under 1.12, Macro.to_string proudces wonky output, making `def` calls look like function calls by adding needless parenthesis. These parenthesis throw off the diff algorithm, and caused an off-by-one error in the code mod. Sourceror has backported the newer code generation so that it's compatible all the way back to 1.10, and produces the correct output.
Patch's assertions will fail in CI due to `mix format --check-formatted` running in dev. Importing patch's deps in test will fix this.
scohen
force-pushed
the
scohen/references-request
branch
from
5531018
to
6c8e9d8
Compare
|
@lukaszsamson This is ready to go. |
Shall we bump I've started to develop a Shall I develop the |
Yes, feel free to bump it
Go for it. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Experimental project structure
There is only one commit in this pr, d8b9694
This commit represents a new structure for the experimental project, and a path forward. With these changes the project now has:
either share messages with the existing server or take them over.
Presently, the find references and formatting providers are
implemented and "exclusive", meaning that they're handled solely by
the experimental server if it's enabled.
back again. This conversion is handled automatically for providers.
and build.
messages. This allows the work that underpins providers to be tested
independently from the language server.
This branch is designed to be able to be merged without affecting the rest of the project, but if this path looks good, I recommend the following course of action.
We should pause feature development and focus on porting existing features to the experimental server. There aren't that many existing providers, and that should happen fairly quickly. New feature development should take place in the
experimentalmodule from then on.We should also standardize on some type of source code modification, likely https://github.com/doorgan/sourceror. It's an excellent little library that has features that are missing from the elixir macro modules and will allow us to transform source code more readily than before.
We should also focus on feature quality over quantity. The goals for this stage of the project would be completely eliminating crash bugs, increasing performance and correctness and usability of the existing features. If i may pick on two recently added features, Code actions are completely broken and automatic imports have severe usability issues that make them fairly worthless. We should put together and publish a product roadmap for this.
Once the experimental server has gained feature parity with the classic server, we should do a cutover release, at which point the project can start adding features again, with more rigor and confidence than was possible in the past.
FYI, I can work on this pretty much full time until March.