Port inlay hints

[gabritto]

Port of inlay hints, mostly the same as Strada, plus a bug fix in node builder when setting raw text of template literal types.
Notable differences:

• Removed interactiveInlayHints. This used to control whether we used strings or inlay hint label parts, and was hardcoded as true in vscode. If we need to optimize the performance of this in the future, we could resolve a label's location lazily, which is the thing that could most significantly affect performance, depending on how it's implemented.
• We don't compute label locations yet in Corsa, because the way we used to do it in Strada relied on symbols stored in identifiers, and we don't have that in Corsa anymore, and there's no trivial way to get it. Leaving this for a future PR, and the non-submodule accepted diffs are all because of that.

[jakebailey]

This is fun, maybe we should just quick pull this into main?

[gabritto]

The effect of this is that you can't hover over an identifier in an inlay hint in the editor. But I'm leaving this to be fixed later, because it seems more important to have inlay hints working in the editor now, and it could take some non-trivial work to fix this, e.g. constructing a map of identifiers to symbols in the node builder.

[gabritto]

This diff is because I fixed a bug from Strada where we didn't fully detect if a comment had the parameter name.

[gabritto]

These two are due to a good change in the checker when we check JavaScript expandos: they used to be checked in a special way, but are now checked the same as TS expressions.

[gabritto]

Diffs from here on are due to removing the interactiveInlayHints option, which is now always true in Corsa.

[Copilot]

Pull Request Overview

This PR adds support for inlay hints functionality in the TypeScript language server port. The main change introduces a DiffFixupNew function alongside the existing DiffFixupOld to allow transformations on baseline output during test comparisons.

Key Changes

• Added DiffFixupNew field to baseline test options for transforming actual output before comparison
• Updated baseline diff generation to apply both old and new fixup functions
• Accepted 23 new fourslash inlay hints test baselines showing convergence with TypeScript's output format (labels now use structured arrays instead of plain strings)
• Fixed template literal escape sequence handling in types (e.g., \t and \r\n now properly preserved instead of being expanded)

Reviewed Changes

Copilot reviewed 165 out of 165 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
testdata/submoduleAccepted.txt	Added 23 inlay hints test files to accepted baseline differences
testdata/baselines/reference/submoduleAccepted/fourslash/inlayHints/*.baseline.diff	Accepted baseline diffs showing inlay hint label format changes from strings to structured arrays
testdata/baselines/reference/submodule/fourslash/inlayHints/*.baseline	New baseline files for inlay hints tests with correct structured label format
testdata/baselines/reference/submodule/compiler/templateLiteralsInTypes.*	Fixed template literal escape sequences in type representations
internal/testutil/baseline/baseline.go	Added DiffFixupNew option for transforming actual output in baseline comparisons

[jakebailey]

For baselining purposes, should we convert these to plain strings?

[gabritto]

For the baselines or for diffing the baselines?

[jakebailey]

Location is empty in baselines; are we missing some code that can baseline these?

[jakebailey]

(Given we didn't baseline them before, I do wonder if we should just not bother look at this in baselines to have fewer diffs)

[gabritto]

We did baseline them before, the thing is in Strada we use position + file, and in Corsa it's LSP location, and I didn't want to write code to convert one into the other, I figured for now the important part is to notice if we do have a location or not, so I'm just emptying locations in the diff fixup functions.

[jakebailey]

Hm, but there's only a string here, not something with a location? Maybe I'm missing something

[gabritto]

This label stands for the identifier a, so its value is "a", and the location should point to the original parameter identifier, but location fields are being omitted when computing the diffs. They do show up in the .baseline file. If there's something I can do to make this all less confusing, let me know, I haven't put a ton of effort into the diff fixing code because we're eventually getting rid of it.

[jakebailey]

I think the thing I was missing was that the Strada baseline says:

{
  "text": "a:",
  "position": 88,
  "kind": "Parameter",
  "whitespaceAfter": true
}

Whereas the new baseline says:

{
  "position": {
    "line": 6,
    "character": 6
  },
  "label": [
    {
      "value": "a",
      "location": {
        "uri": "file:///a.js",
        "range": {
          "start": {
            "line": 3,
            "character": 17
          },
          "end": {
            "line": 3,
            "character": 18
          }
        }
      }
    },
    {
      "value": ":"
    }
  ],
  "kind": 2,
  "paddingRight": true
}

So with that context, it makes sense what is happening here.

[gabritto]

Ah yes, that's because "interactive inlay hints" is always true in Corsa now, so we always return inlay hint label parts.

[jakebailey]

Skimming, I think this seems good. All of the baseline diffs I read look good, and even though VS Code enables it by default, the defaults are such that they won't actually turn on until the user turns on one of the typescript.inlayHints. settings, so for checker.ts this costs nothing.

For the people who turn this on, we're going to use increased resources, but I think this is not a bad situation given that VS code doesn't enable these.