Cannot inject grammar to some scopes in Markdown

[KermanX]

Does this issue occur when all extensions are disabled?: Yes. Except the current extension

• VS Code Version: 1.19.1
• OS Version: WinNT x64 10.0.22631

Steps to Reproduce:

1. Clone https://github.com/KermanX/grammar-injection-repro
2. Start debugging the extension
3. See this:

The "Success Injection" is injected correctly, while the "Failed Injection" is not.
5. Open ./synatxes/failed.json and ./synatxes/success.json, they are all the same except the injectionSelector.

After some testing, I think most of the scopes can be injected, while the following can't:

• fenced_code.block.language.attributes.markdown
• fenced_code.block.language.markdown
• punctuation.definition.markdown

[vscodenpa]

Thanks for creating this issue! It looks like you may be using an old version of VS Code, the latest stable release is 1.91.1. Please try upgrading to the latest version and checking whether this issue remains.

Happy Coding!

[RedCMD]

you can only inject in-between begin/end rules or capture groups that have a recapturing patterns array

entity.name.section.markdown is recapturing

fenced_code.block.language.attributes.markdown is not

https://github.com/mjbvz/vscode-fenced-code-block-grammar-injection-example

[KermanX]

you can only inject into begin/end rules or capture groups that have a recapturing patterns array

Thanks for your reply. Then how can I inject grammars into the fenced_code.block.language.attributes.markdown part without re-defining all the code block grammars for every embedded language?

For example, the following grammar is a modified version of Markdown grammar which works injected to text.html.markdown:

    "fenced_code_block_ts": {
      "begin": "(^|\\G)(\\s*)(`{3,}|~{3,})\\s*(?i:(typescript|ts)((\\s+|:|,|\\{|\\?)[^`]*)?$)",
      "beginCaptures": {
        "3": {
          "name": "punctuation.definition.markdown"
        },
        "4": {
          "name": "fenced_code.block.language.markdown"
        },
        "5": {
          "name": "fenced_code.block.language.attributes.markdown",
          "patterns": [
            {
              "include": "#codeblock-attributes" // <--- My grammar
            }
          ]
        }
      },
      "end": "(^|\\G)(\\2|\\s{0,3})(\\3)\\s*$",
      "endCaptures": {
        "3": {
          "name": "punctuation.definition.markdown"
        }
      },
      "name": "markup.fenced_code.block.markdown",
      "patterns": [
        {
          "begin": "(^|\\G)(\\s*)(.*)",
          "contentName": "meta.embedded.block.typescript",
          "patterns": [
            {
              "include": "source.ts"
            }
          ],
          "while": "(^|\\G)(?!\\s*([`~]{3,})\\s*$)"
        }
      ]
    },

However, this is only for TypeScript, and in this way all embedded languages need a new pattern.

[RedCMD]

as far as I'm aware
you can't

you could maybe ask @mjbvz kindly to add an empty patterns array under capture 5
but this would also require every single other extension author to do the same (if you wish to support their injections),
wouldn't work on any old version of VSCode/extensions
and might cause other injections that target markup.fenced_code.block.markdown to also inject into it

[KermanX]

Thanks for your advice. In that case I may generate all the code blocks on my grammar file to simplify the task. However, I am wondering why the patterns array is required to be injected? Is it possible to change this behavior?

Thanks a lot!

[mjbvz]

What's the goal with these injections?

[KermanX]

What's the goal with these injections?

Slidev.js has some special attributes for code blocks. For example, line highlighting. And the expected highlighting is:

which is exactly the fenced_code.block.language.attributes.markdown scope.

And it is common for Markdown-based CMS to add its syntax into this scope, for example, VitePress

Sorry for the confusion, I've fixed my last comment which used to say capture groups instead of patterns

[mjbvz]

Thanks. Can you test if adding the empty patterns element to fenced_code.block.language.attributes.markdown would allow you to do the injection?

We can pretty easily add this for all of our built-in languages

[KermanX]

Thanks. Can you test if adding the empty patterns element to fenced_code.block.language.attributes.markdown would allow you to do the injection?

Yes, it works. In fact, I've used some code to generate new grammars with patterns: [] for each built-in languages and everything works as expected. So if you feel this grammar scope is too specific, it's completely fine to not update the official grammar file.

May I ask if is there any way to to support injection to any scope without patterns defined? I am not sure why currently this is not allowed.

[mjbvz]

Thanks. Can you please file an issue against https://github.com/microsoft/vscode-textmate for the injection issue

Also if you want to add the empty patterns element, you can do so over in this repo: https://github.com/microsoft/vscode-markdown-tm-grammar

Should just require changing markdown.tmLanguage.base.yaml. See the contributing instructions for more details