v.1.0.1 breaks on multi-line links

[Battleman]

After updating docusaurus-terminology to 1.0.1, our build fails, with error

TypeError: object null is not iterable (cannot read property Symbol(Symbol.iterator))
[ERROR] Client bundle compiled with errors therefore further build is impossible.
error Command failed with exit code 1.

package.json dependencies look like so:

extract of package.json

```json { ... "dependencies": { "@docusaurus/core": "2.4.3", "@docusaurus/preset-classic": "2.4.3", "@docusaurus/theme-mermaid": "2.4.3", "@easyops-cn/docusaurus-search-local": "0.35.0", "@grafana/faro-web-sdk": "1.1.4", "@grnet/docusaurus-glossary-view": "1.0.1", "@grnet/docusaurus-term-preview": "1.0.1", "@grnet/docusaurus-terminology": "1.0.1", "@mdx-js/react": "1.6.22", "clsx": "1.2.1", "docusaurus-plugin-image-zoom": "0.1.2", "hast-util-is-element": "1.1.0", "prism-react-renderer": "1.3.5", "react": "17.0.2", "react-dom": "17.0.2", "rehype-katex": "5", "remark-math": "3" }, "devDependencies": { "@docusaurus/module-type-aliases": "2.4.0" }, ... } ```

Downgrading @grnet/docusaurus-terminology to 1.0.0 fixes the issue.

I'm no node guru, but am willing to provide additional information if necessary, or follow some debugging steps.

[Battleman]

Update: it appears we have a file that contains a link on multiple lines

[a link 
like this](https://example.org)

1.0.0 had support for multi-lines (flags gmi), while 1.0.1 has dropped flags mi. It's not a concerning issue for us, as it was probably introduced long ago by an auto-formatter. Putting back the link on one line fixes the issue, while not having any other side effect for us.

I guess now this issue is up to you: maybe it's worth adding some checks to prevent the obscure issue, or add back the m flag, or just document it somewhere.

As far as we're concerned, this issue is fixed.

[tsironis]

Thanks @Battleman. Glad to see the issue is fixed on your end, and also, for your suggestions. We'll keep this open and see if we can improve the development experience.

[dimisus]

TypeError: object null is not iterable (cannot read property Symbol(Symbol.iterator))

When filtering parsed URLs for one-liners only it builds without errors. Can confirm the issue relates to multi-line URLs as mentioned by @Battleman .

// in ../node_modules/@grnet/webpack-terms-replace-loader/index.js

module.exports = function(source) {
  const urlsRegex = /(?<!!)\[[^\]]+\]\([^)]+\)/g;

  const urlRegex = /\[(.*?)\]\((.*?)\)/;

  const urls = (source.match(urlsRegex) || []).filter(u => {
    regex = /^(\[{1})[\w`].*\]\(.*\)$/g;     // <=== here filtered for only one-liner urls

    return regex.test(u);
  })

Note: this is not a solution, it just throws out all multi line urls

[osterman]

Thank. you @dimisus!! I was tearing my hair out on this one, and this led me to the fix.

Here's my modification, that outputs the offending URLs. Only the ones with a newline needed to be fixed. Others with braces or other characters were redherrings.

const urlsRegex = /(?<!!)\[[^\]]+\]\([^)]+\)/g;

  const urlRegex = /\[(.*?)\]\((.*?)\)/;

  const urls = (source.match(urlsRegex) || []).filter(u => {
    regex = /^(\[{1})[\w`].*\]\(.*\)$/g;     // <=== here filtered for only one-liner urls
    if (!regex.test(u)) {
      console.log(u); // Log if it's not a one-liner URL
      return true; // Include this URL in the filtered results
    }

    return regex.test(u);
  })

[tsironis]

2.0.0-rc.1 was just released and it contains the fix for this bug. Please check it out and feel fee to reopen the issue if the bug still persists.