I don't really want to open the @lezer namespace to 3rd-party code. Just lezer-printer or @lezer-unofficial/printer is perfectly fine.

(You are aware of the very basic pretty-printing that Tree.toString provides, no?)

No. That's not something I'm going to work on.

Recently, I started making a new parser and discovered there's still no official pretty printer. I found one on the CodeMirror forum (link) and saw the original post has gotten 1.8K views since 2020.

Currently, I packaged that pretty printer (link) to use for my new package. However, it'd be nice to have an official package, with at least some pretty-printing format (not necessarily JSON).

@marijnh, could you please consider adding such a printer to the @lezer family?

The constructors aren't really part of the public interface. In what situation are you calling them?

I hope this message finds you well. I am currently enrolled in the Programming for Everybody (Getting Started with Python) - University of Michigan on Coursera and have encountered persistent issues with Graded Assignment 3.1 and Assignment 2.3. Despite multiple attempts and careful review of the provided materials, I continue to encounter errors even when I believe my code is correct.

I have been working on these assignments for several weeks, and I am eager to resolve the issues to progress in the course. I have included the links to the assignments below for your reference:

Graded Assignment 3.1: [ https://www.py4e.com/tools/pythonauto/?PHPSESSID=8b640d8629795cfa3d157127560c1179 ]

Assignment 2.3: [ https://www.py4e.com/tools/pythonauto/?PHPSESSID=f1014755d5607c2bae9dd9a5b4f02059 ]

I kindly request your assistance in investigating and resolving these issues. I believe there may be technical errors on the platform, as I have followed the instructions closely and reviewed my code thoroughly.

Your prompt attention to this matter would be greatly appreciated. If possible, please provide guidance or solutions to the issues I am facing, or direct me to the appropriate channels for assistance.

Thank you for your support, and I look forward to your response.

Sorry, this is not what this bug tracker is for, and this kind of assistance isn't something you'll be able to get here.

To be on the safe side, you could add a wildcard export like this

  "exports": {
    ".": {
      "import": "./dist/esm/markdown-it.mjs",
      "require": "./index.js"
    },
    "./*": {
      "require": "./*",
      "import": "./*"
    }
  },

That way, imports of the root will properly use the ESM build, but direct subpaths are also still supported.

The parser-generator algorithm used by this tool definitely has limits, and you'll want to keep your (expanded) grammar size within reasonable bounds. A rule with 600 or-ed tokens sounds like a bad idea. Do these tokens actually serve a different purpose in the grammar, or could you collapse them to a single more generic token type? Optionals and or-expressions are compiled by expanding them into multiple rules (i.e. (a | b) c? becomes a, b, a c, b c), and if you combine a bunch of them you can get a serious blowup in the amount of rules.

Describe the issue

This may be a platform difference, but on the Mac, the CMD-Backspace (implemented here in deleteToLineStart should delete to the beginning of the visual line, but instead delete's the the start of the logical line.

With the link below, add enough text to wrap in the window and then use CMD-Backspace. The text will be deleted to the beginning of the logical line, not the visual line.

Browser and platform

Safari, Macintosh

Reproduction link

https://codemirror.net/try/?c=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

Could you check result of this PR https://github.com/markdown-it/markdown-it/pull/863/files ? Are ESM exports as expected or some tweaks needed?

I've added a comment to #863

In 2023, we'll also want an exports field in package.json, something like ...

  "exports": {
    "import": "./index.js",
    "require": "./dist/esm/markdown-it.mjs"
  },

That makes sure resolution that doesn't use the non-standard "module" field (such as node's own) can still find the ESM code.

Any further feedback on this? My feeling is that this would be mostly unnecessary complexity, due to the availability of a stack trace.

That seems reasonable. Done in attached patch.

Hi @marijnh,

This is not a bug report but a small change request in https://github.com/codemirror/view/blob/main/src/layer.ts#L29.

Would it be possible to make these public?

private left;
private top;
private width;
private height;

Asking this because I am leveraging RectangleMarker.forRange to calculate rects, but I am not directly using layers in my case. For example, sometimes I would like to extend the rect to make it longer.

It seems some calculation inside it cannot be used without accessing private properties.

Thanks!

The checkbox example does not work properly when checked is not set in the constructor.

Thanks for the reproduction. Attached patch fixes this.

Describe the issue

I'm not sure how to intentionally trigger this (I do have a repro though), but sometimes when updating decorations with inline widgets the old widgets won't be removed from the DOM. I'm pretty sure that if you scroll away from these broken widgets, they will always get removed from the DOM due to CodeMirror's content virtualization.

Here is a video of what happens:

Screencast from 09-28-2023 02:11:06 PM.webm

My reproduction is based off an old repro I had about inline widgets before/after block widgets. To reproduce:

• use the repro linked below
• move your selection around (way you do it shouldn't really matter?)
• expect the blue inline widgets to get duplicated

Ideally, you should just see a inline -> block -> inline widget following your selection.

Browser and platform

Firefox, Chrome - Linux

Reproduction link

https://codemirror.net/try/?c=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

The readonly checked syntax in TypeScript means precisely this—initialize that property to the value of the argument. What makes you think it does not work?

<!-- Do not use this bug tracker for questions. We have a forum (https://discuss.codemirror.net) for those.

In order to get us to look at something, you need to make it easy for us to understand what you are doing and what went wrong. When possible, a code example that demonstrates the problem helps a lot. If there's any chance at all that a problem is related to the browser, include information about which browser(s) you tested with. -->

Codemirror will show error when using elements get from getElementsByClassName, however, if I change the code to getElementById, it will work.

Error: codemirror_6.65.7_codemirror.min.js:1 Uncaught TypeError: Cannot read properties of undefined (reading 'body') at N (codemirror_6.65.7_codemirror.min.js:1:1800) at p.fromTextArea (codemirror_6.65.7_codemirror.min.js:1:168905) at loadHighlight (highlight.js:30:20) at load (highlight.js:23:5) at window.onload (model.html?u=/lessons/ch/1.html:12:17) N @ codemirror_6.65.7_codemirror.min.js:1 p.fromTextArea @ codemirror_6.65.7_codemirror.min.js:1 loadCodeHighlight @ highlight.js:30 load @ highlight.js:23 window.onload @ model.html?u=/lessons/ch/1.html:12 load(async) (anonymous) @ model.html?u=/lessons/ch/1.html:11

Example code: var elements = document.getElementsByClassName("codeTextarea"); for(var a in elements){ CodeMirror.fromTextArea(elements[a], { lineNumbers: true, mode: "text/x-csharp" }); }

This is definitely going to be related to the actual value you're passing in, not whether it is coming from getElementById or getElementsByClassName. It is not a CodeMirror problem.

This could be a number of things—the selection somehow containing invalid positions, the content tree being broken, or some logic mistake in way it is being scanned. If someone manages to catch it, the values of view.docView.toString() and view.state.selection.toJSON() would be useful for narrowing it down. But of course, a set of reproduction instructions would be even better.