Add check for valid HTML translation

[yishus]

Fixes #59

I'm running into issues validating HTML with Nokogiri (img, br, svg tags are flagged as errors), I've switched it out to Nokogumbo after checking this issue. Getting good results with that so far.

en.default.json

{
    "test_html": "<img src='something.jpg'>",
    "hello_html": "<h1>wutwuti<p><img><h2",
    "images": {
        "assets": {
            "svg": "<svg viewBox='0 0 10 10' x='200' width='100'><circle cx='5' cy='5' r='4' /></svg>"
        }
    },
    "accounts": {
        "one": "<h3>Your account</h3>",
        "other": "<h3>Your accounts</h3>"
    }
}

dev check path/to/my/theme

suggestion: ValidHTMLTranslation: 'hello_html' contains invalid HTML:
1:23: ERROR: End of input in tag.
<h1>wutwuti<p><img><h2
                      ^
1:20: ERROR: Premature end of file  Currently open tags: html, h1, p.
<h1>wutwuti<p><img><h2
                   ^.

4 files inspected, 1 offenses detected

[macournoyer]

Looking good!

[yishus]

@macournoyer okay ready for review!

[macournoyer]

Awesome work! We can now replace the old theme-lint 🎉

[macournoyer]

Oh! One more thing. Can you test it will not cause errors w/ self-closing tags: <img src="...> or <br>.

[yishus]

I've updated to use Nokogumbo, details in the PR description above ☝️

[macournoyer]

Nice find! Didn't know about nokogumbo. Looks perfect.

[jbourassa]

I spent some time looking for a parser that would be easy to use and did not come across Nokogumbo, but looks great. Well done!

Left some minor comments but from my perspective this is a great addition.

[jbourassa]

The check for "is this a pluralization set of keys?" is a bit different than what we had in ThemeCheck::MatchingTranslations but I think that's a good enough heuristic, and we can tweak those later if need be.

theme-check/lib/theme_check/checks/matching_translations.rb

Lines 84 to 88 in f1956ef

	def pluralization?(hash)
	hash.all? do |key, value|
	PLURALIZATION_KEYS.include?(key) && !value.is_a?(Hash)
	end
	end

[jbourassa]

The error messages are a bit odd when we're parsing single lines, wonder if we can tweak them? Here for instance

1:17:

Line 1 most likely isn't line 1

Premature end of file

From a consumer's perspective, a string is being parsed, not a file, right?

Currently open tags: html, h1.

Looks like html is automatically injected here.

Do you think there's a way for us to improve these?

[yishus]

Agree with all the points above, but sadly most of these information are only available through the to_s method. We are able to obtain the line and column values, and also str1 which returns the HTML standard error code. (source)

If we are printing only the error code, it will look like

suggestion: ValidHTMLTranslation: 'hello_html' contains invalid HTML: eof-in-tag, generic-parser.

Which is still okay, it doesn't point to the exact location of the error, but can be helpful

[macournoyer]

Merging this. We can tweak the error message in another PR. I've used dirty regexes in other places to do it :/