bug: magnet links are treated as relative links

Steps To Reproduce

[some magnet link](magnet:?xt=urn:btih:asdf)

The current behavior

relative-links Relative links should be valid [Link "magnet:?xt=urn:btih:asdf" is dead]

currently, some external protcols are hard-coded

const EXTERNAL_PROTOCOLS = new Set([
  'http:',
  'https:',
  'mailto:',
  'tel:',
  'ftp:'
])

        if (hrefSrc != null) {
          const url = new URL(hrefSrc, pathToFileURL(params.name))
          url.hash = ''
          const isRelative =
            hrefSrc.startsWith('./') ||
            hrefSrc.startsWith('../') ||
            !EXTERNAL_PROTOCOLS.has(url.protocol)
          if (isRelative && !fs.existsSync(url.pathname)) {
            const detail = `Link "${hrefSrc}" is dead`
            addError(onError, lineNumber, detail)
          }
        }

The expected behavior

it should ignore magnet links

more generally, it should ignore all links with protocol except the file: protocol

$ node
> new URL("magnet:?xt=urn:btih:asdf").protocol
'magnet:'
> new URL("file:asdf").protocol
'file:'
> new URL("file:///asdf").protocol
'file:'
> new URL("asdf").protocol
Uncaught TypeError [ERR_INVALID_URL]: Invalid URL
> new URL("./asdf").protocol
Uncaught TypeError [ERR_INVALID_URL]: Invalid URL

Description

GitHub markdown supports using line length as the anchor, for example:

[file](index.md#L7)

I don't want the linter to throw error on this.

Describe the solution you'd like

I don't know if there is a way to optionally turn on the anchor validation for a specific line, it seems a bit complex to implement.

Instead maybe just hardcode in the plugin to ignore that?

Describe alternatives you've considered

Description

It would be very useful to validate that anchors exist:

some.md:

[Link](other.md#existing)
[Link](other.md#not-existing)

other.md:

# Existing

This would give an error that other.md#not-existing is not a valid link.

This would be useful, because I moved a section from one file to another, and now I don't know what links I've broken. Similarly, I would like to know when renaming a header invalidates a link.

Describe the solution you'd like

A piece of config to specify whether to opt-out of this behavior, with it being on by default.

Describe alternatives you've considered

I considered switching to markdownlint-link-check, but they don't do it either.

The other alternative would be the current behavior, not linting for dead link fragments.

Description

Currently, the rule doesn't report an error for the following:

├── file1.md
└── file2.md

With file1.md content:

## Existing Heading

With file2.md content:

[Link](./file1.md#ExistIng-Heading)

Whereas, we should report an error, to enforce consistency with the built-in markdownlint rule - MD051.
More explanations there: DavidAnson/markdownlint#1117 (comment)

Describe the solution you'd like

The rule should reports an error, saying that it's an invalid link fragment, as with the above example, the link should have been: [Link](./file1.md#existing-heading).

There is a test fixture to ensures that we don't consider as errors, the links fragments with mixed casing: https://github.com/theoludwig/markdownlint-rule-relative-links/tree/develop/test/fixtures/valid/existing-heading-case-insensitive, we should instead move it to "invalid", and make the test pass. However, this change should be considered as a BREAKING CHANGE.

Describe alternatives you've considered

Continue to consider heading links fragments valid even if the link fragment is not only in lowercase, with mixed case for example.

Type of Improvement

Refactoring code

Proposal

We should use ESM imports instead of CommonJS.
See: https://gist.github.com/sindresorhus/a39789f98801d908bbc7ff3ecc99d99c

Related issues:

• markdownlint-cli2 commit: Remove references to ESM/.mjs because VS Code only supports it during development
• markdownlint-cli2 issue: Attempt to load .mjs configuration file fails with "Invalid host defined options"
• VSCode: Enable consuming of ESM modules in extensions

Description

It seems that currently only links to markdown headers are processed, unlike https://github.com/DavidAnson/markdownlint/blob/main/doc/md051.md which also supports links to html anchors.

Describe the solution you'd like

Would be great to have the same processing logic as md051 so that this plugin is usable with anchors.

Describe alternatives you've considered

Current workaround is using https://github.com/tcort/markdown-link-check with the following config:

{
    "ignorePatterns": [
      {
        "pattern": "^#"
      }
    ]
}

This utility also does not work well with anchors so we can ignore all the links to them while links to files are being verified just fine.

I am trying to use markdownlint-cli2 as part of my CI GitHub workflow. To install this package it runs:

npm install -g [email protected]

However when I run the markdownlint-cli2 action it gives me

Run DavidAnson/markdownlint-cli2-action@ed4dec634fd2ef689c7061d5647371d8248064f1
  with:
    globs: markdown-test/index.md
    config: .markdownlint-cli2.jsonc
    separator: ,
markdownlint-cli2 v0.10.0 (markdownlint v0.31.1)
Error: Failed due to error: AggregateError: Unable to require or import module 'markdownlint-rule-relative-links'.

Steps To Reproduce

1. Add the above command as a step in GitHub workflow
2. Run the workflow

The current behavior

The markdownlint-cli cannot find the installed package.

The expected behavior

The action should run without error.

Type of Improvement

Refactoring code

Proposal

Rewrite the code in TypeScript.
The project wasn't started as TypeScript, as markdownlint types was light and often incorrect, but this changed, and greatly improved.
As the types improved, and are now "good", we use TypeScript to lint the JSDOC since this commit: 1ddcdc7.

As we're already using TypeScript, but to lint JavaScript, it makes sense, to instead directly use TypeScript syntax.

Description

Since the following PR: DavidAnson/markdownlint#1117, the built-in markdownlint rule - MD051 supports links fragments with columns, e.g: [Link](#L30C11-L31C9).

Describe the solution you'd like

At least ignore those links fragments to not consider them as errors.

Or similarly as explained in #6 (comment), we could "do better than ignoring" and verify that the lines range is correct, and columns are correct too, for example #L30C11 means that at the line 30, there should be at least 11 characters (.length of the string greater or equal than 11).

Describe alternatives you've considered

Ignoring the rule with <!-- markdownlint-disable-next-line relative-links --> when necessary (usage of columns numbers in links fragments).