Add code symbols into outline

[vezwork]

Addresses an issue mentioned by @juliasilge: #167 (comment)

Adds code symbols under code cells in the outline.

[posit-snyk-bot]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[vezwork]

I attempted to add tests until I remembered that our test setup does not have actual LSPs for Python or R in it, so we can't really test this. Well... now that I write that, I realize I could probably mock one out. That seems a little contrived, but maybe still worth it.

[juliasilge]

This PR:

• fixes Feature Request: In qmd file outline view, in source mode, include R functions (that are in R chunks) #880
• fixes Feature Request: Add Outline Support for Named Code Sections in R Code Chunks #918

[juliasilge]

I seem to be getting two of everything, at least in Positron:

[vezwork]

I was able to reproduce duplicated language symbols in code cells in Positron with inline outputs enabled. Looking into what is happening.

[vezwork]

@juliasilge In qmds in Positron with Positron › Quarto › Inline Output: Enabled enabled, the following command (from getCodeCellSymbols in this PR) is returning duplicated symbols:

await commands.executeCommand<DocumentSymbol[] | SymbolInformation[]>(
  "vscode.executeDocumentSymbolProvider",
  uri
);

I checked the surrounding context, including the vdoc content. Everything seems to be in order, nothing strange otherwise. When I disabled Positron › Quarto › Inline Output: Enabled and re-open the document, I no longer experience the duplicated symbols issue. It appears to me that the "vscode.executeDocumentSymbolProvider" command does not function as expected when inline outputs are enabled.

[vezwork]

@seeM mocked out fake LSPs in #980. Maybe I can copy that approach for tests in this PR.

[vezwork]

Cell Diagnostics tests were failing with a message about timing out in CI. The current run of the tests passed, but there wasn't a code change that would've affected the tests or the code tested. The tests seem to be flakey for some reason. Let's keep our eye on that.

[juliasilge]

I looked into the issue with inline output and it is some change we'll need to make over on the Positron side; I'll log that once this is merged.

I do think we need to add at least some basic tests, with a fake in-process symbol provider (not a real LSP). We already do this for the formatting tests. I think the steps would be:

1. Register a temporary provider in symbols.test.ts with vscode.languages.registerDocumentSymbolProvider({ scheme: "file", pattern: "**/.vdoc.*" }, ...) so it only handles Quarto virtual docs.
2. Make that provider return each shape in separate tests:
   • DocumentSymbol[]
   • SymbolInformation[]
   • undefined
3. Run vscode.executeDocumentSymbolProvider on format/basics.qmd or another example file and assert:
   • no throw for undefined
   • code-cell symbols still exist
   • returned embedded symbol names appear under/alongside code-cell outline entries for both result shapes.

That will give us coverage of the middleware logic without having to set up real LSPs.

[vezwork]

Ok I added tests but frustratingly was not able to get them to work when run in sequence, only in isolation. Doubly confusing is that mocking a SymbolInformation provider somehow results in duplicated results in the outline in the test, but idk why. Couldn't figure it out today :|

[juliasilge]

In a0506c9 I made some changes to how we run this new testing. The SymbolInformation[] test was skipped because it failed when run after other tests but passed in isolation, and showed duplicated symbols. The cause was not provider.dispose() actually in the end. What I believe it was instead is that vscode.executeDocumentSymbolProvider caches results per document URI (VS Code's OutlineModel cache). Because every test opened the same format/basics.qmd, a stale cached symbol tree from a previous test (or another suite that opened the same file) was served instead of re-invoking the middleware. That leaked the previous test's symbols, dropped the current test's, and surfaced duplicates.

I added a new helper function to address this in test-utils, which opens a unique on-disk copy of an example file so each test gets a fresh URI and the provider actually runs. I only set it up to be used in the new code cell symbol tests but this likely will be useful for other LSP features that work like this with caching.

[juliasilge]

I opened posit-dev/positron#13907 to fix the duplicated symbols with inline output.