[json] colorize string literals based on language information coming from schemas

[voxpelli]

The built in json-language-features extension provides plenty of nice features for JSON Schemas, such as markdownDescription and defaultSnippets.

I would love for it to also enable indicating that the value of a property is an embedded language.

Specifically I would love for the built in npm extension to be able to indicate that values within the "scripts" property of package.json files are shellscript values.

Would make it easier to interpret and author complex npm-scripts setup such as this:

Implementation brainstorming

The current schema that the npm extension uses is json.schemastore.org/package:

vscode/extensions/npm/package.json

Lines 314 to 318 in 444d7a4

	"jsonValidation": [
	{
	"fileMatch": "package.json",
	"url": "https://json.schemastore.org/package"
	},

It actually already contains something similar that's aimed at IntelliJ – "x-intellij-language-injection": "Shell Script", which IntelliJ added in JetBrains/intellij-community@2e490ce

Here's how it looks in the package.json schema:

"scriptsInstallAfter": {
  "description": "Run AFTER the package is installed.",
  "type": "string",
  "x-intellij-language-injection": "Shell Script"
},

Maybe something similar, such as "x-vscode-language-injection": "shellscript" could work? Or, considering that Shell Script is an alias of shellscript, consume x-intellij-language-injection wherever it is available?

Related discussions

Mostly stumbled upon people requesting this for yaml:

• Desire for embedded languages in GitHub Action workflow files: Add support for embedding languages for the runs keyword github/vscode-github-actions#266
• Desire for embedded languages at large in yaml files: Support syntax highlight from different language for multi-line strings with a special comment redhat-developer/vscode-yaml#943

[voxpelli]

Here's the PR that added the "x-vscode-language-injection": "shellscript" to the package.json JSON Schema: SchemaStore/schemastore#2195

[hyperupcall]

This would be a great addition! As someone that co-maintains SchemaStore, I lean towards the shared use of the x-intellij-language-injection property, assuming languages have aliases that generally include the Intellij ids. Otherwise, all schemas that have properties of x-intellij-language-injection will also need to add an extra x-vscode-language-injection, and we'd have to check to ensure the values are in sync.

[aeschli]

Annotating the schema is only a first, small step, the challenge is how to do the coloring.

Syntax highlighting happens on the renderer, through TextMate grammars and grammar injections. It has no access to the JSON schema which is loaded and evaluated from the JSON language extension, running in a process on the extension host. We don't have a way to dynamically inject rules on runtime.

What could be done is to use semantic highlighting, in the JSON language extension, but that would mean implementing a parser for shell script.

Help is welcome...

[vs-code-engineering]

This feature request is now a candidate for our backlog. The community has 60 days to upvote the issue. If it receives 20 upvotes we will move it to our backlog. If not, we will close it. To learn more about how we handle feature requests, please see our documentation.

Happy Coding!

[voxpelli]

@aeschli I'm thinking it would work similarly to how CSS and JS can be embedded in HTML, is that a lot different?

[aeschli]

With HTML, the HTML grammar includes the CSS and TypeScript grammars. when ever it sees <style> is starts using the rules for CSS.

The difference is that the coloring that you want is only for certain files and properties and only at runtime (once we have loaded and evaluated the schema) we know which ones. We have no mechanism to redefine/customize grammars at runtime.
We can do this with semantic highlighting, which goes on top of syntax highlighting.

[vs-code-engineering]

🙂 This feature request received a sufficient number of community upvotes and we moved it to our backlog. To learn more about how we handle feature requests, please see our documentation.

Happy Coding!

[RedCMD]

I could make an extension to do this
but does the JSON language extension provide any api for other extensions to use?
location of strings/objects and the current schema in use
it would nice to not have to re-compute the entire schema and JSON document

I could then provide semantic syntax highlighting using TextMate or tree-sitter

I don't see how it could be possible to provide a JSON TextMate grammar to do it perfectly (built-in TextMate syntax highlighting)
as its impossible for TextMate to lookahead multiple lines (which could be required by some schemas)
however, it would allow for extra language features like bracket highlighting etc

[aeschli]

@RedCMD No the JSON language extension does not have such an API. It's unclear to me how such an API would look like. Also I'd rather avoid dependencies between extensions. Maintaining APIs is a lot of work.

• The semantic highlighting feature could be implemented in the JSON extension itself
• Or the extension that needs schema information can use the vscode-json-languageservice node module. For parsing JSON use the node-jsonc-parser node module.

I would recommend not to worry about schema support first, but just look at e.g. syntax highlighting in package.json script tags. Its already challenging enough.

[RedCMD]

I've created up a mock extension https://marketplace.visualstudio.com/items?itemName=RedCMD.json-embedded-languages
Simply just a TextMate injection to target a root level "scripts" object

"x-intellij-language-injection" lang:json -path:**/globalStorage** -path:**/package.json -path:**/package.schema.json