Update: Upgrade remark-parse to v7 (fixes #77, fixes #78)

[btmills]

remark-parse@7.0.0 includes remarkjs/remark#423, which fixes handling of leading and trailing newlines in fenced code blocks.

For reviewers:

1. remark-parse v6, v7, and v8 were semver-major fixes. v9 was a full rewrite. Are we comfortable with a parser major version bump in a semver-minor release? This is currently tagged as Update: so it's semver-minor rather than semver-patch.
2. If so, how far? This PR bumps to v7 because it fixes two known issues, First empty line in code fragment breaks line numeration in error messages #77 and First empty line in code fragment silents rules about first empty line in code #78. I'm not aware of any eslint-plugin-markdown issues that would be fixed by v8 or v9.
3. If we're fine upgrading past v7 for speculative parsing accuracy improvements that don't knowingly affect eslint-plugin-markdown, should we stop at v8 or go all the way to the re-written parser in v9? If the latter, I'll close this in favor of Chore: upgrade remark-parse to 9.0.0 #171.

Changelogs:

• remark-parse@6.0.0
• remark-parse@7.0.0
• remark-parse@8.0.0
• remark-parse@9.9.0

[nzakas]

LGTM

[btmills]

Thanks for the review @nzakas. What are your thoughts on those three follow-up questions in the description? I at least want v7 to fix these bugs, but I could use some input on v8 or v9.

[nzakas]

Ah sorry, totally breezed over those. My vote would be to just do the minimum to fix the issues we were having right now (stop the bleeding). We can always do further upgrades later but I’d rather get a safe fix out first.

[remcohaszing]

All remark-parse does is make mdast-util-from-markdown work with unified.

Since eslint-plugin-markdown doesn’t really use unified features, the following changes will yield the same result with fewer dependencies and a properly typed ast.

- const unified = require("unified");
- const remarkParse = require("remark-parse");
+ const fromMarkdown = require("mdast-util-from-markdown");

- const markdown = unified().use(remarkParse);
- 
  /**

  function preprocess(text, filename) {
-     const ast = markdown.parse(text);
+     const ast = fromMarkdown(text);
      const blocks = [];

[btmills]

Thanks for the tip @remcohaszing. It'd be nice to reduce the size of the dependency tree, so I gave that a shot, but it has the same change to indents that I pointed out in #171 (comment). That makes using remark-parse v9 or the underling mdast-util-from-markdown parser a non-trivial change, but I'd welcome the research if someone wanted to figure out what would be required.