bug: Syntax highlighting broken by LaTeX string

[dmca-glasgow]

Did you check existing issues?

• I have read all the tree-sitter docs if it relates to using the parser
• I have searched the existing issues of tree-sitter-typescript

Tree-Sitter CLI Version, if relevant (output of tree-sitter --version)

No response

Describe the bug

Hello,

I originally created an issue at zed-latex about this syntax highlighting bug but have been directed to the TypeScript language, which I believe is here?

This is an example test with broken syntax-highlighting, where I create a LaTeX string, in TypeScript:

LaTeX makes heavy use of the backslash character, and I've found String.raw helps to ignore character escape sequences.

I noticed removing the line \usepackage{graphicx} fixes the issue. If I remove it the colours are now back to normal:

Steps To Reproduce/Bad Parse Tree

The code from the screenshot:

test('image', async () => {
  const latex = String.raw`
    \documentclass{article}
    \usepackage{graphicx}
    \begin{document}
    \includegraphics[]{image.png}
    \end{document}
  `;

  const markdown = await testProcessor.latex(latex);
  // console.log(markdown);

  const expectedMarkdown = unindentStringAndTrim(`
    ![](image.png)
  `);

  expect(markdown).toBe(expectedMarkdown);

  const html = await testProcessor.md(markdown);
  // console.log(html);

  const expectedHtml = unindentStringAndTrim(`
    <p><img src="image.png" alt="Image" /></p>
  `);

  expect(html).toBe(expectedHtml);

  // const quartoHtml = await markdownToQuartoHtml(markdown);
  // console.log(quartoHtml);
});

(Just looking at GitHub's syntax highlighting here, it also seems to struggle with the line \usepackage{graphicx}. Perhaps \u is problematic as it can represent the unicode escape sequence).

Expected Behavior/Parse Tree

Syntax highlighting works correctly for TypeScript files which include strings of LaTeX code using String.raw.

Repro

[lnay]

Sorry, I meant the typescript extension for Zed (which is still inbuilt into the main zed project it seems: https://github.com/zed-industries/zed/blob/main/crates%2Flanguages%2Fsrc%2Ftypescript%2Finjections.scm).
On a second glance it looks like it's not even trying to do a LaTeX injection (there's nothing in the context to make it clear that the string contains LaTeX), probably some other language instead.

[dmca-glasgow]

My guess is it's not to do with code injections, but tree-sitter or tree-sitter-typescript incorrectly parses strings inside String.raw template functions and is not ignoring escape sequences.

[lnay]

Ok yeah you're probably right