Option to minify HTML output

[alexgleason]

We're using Earmark in Pleroma, a self-hosted social media platform, to allow users to write posts in Markdown.

Pleroma is federated, so it has to handle posts from other ActivityPub implementations. Sometimes these implementations will use newlines instead of <br> tags to indicate line breaks, so we apply white-space: pre-wrap; on the text content to preserve this formatting. There's more info in this issue.

This is a problem when using Earmark, since it will always output prettified HTML. Eg for this input: hello\n\nworld! we get this output: <p>\nhello</p>\n<p>\nworld!</p>\n. With white-space: pre-wrap; applied it treats these \n very literally and creates a lot of unwanted whitespace.

I'd like the ability to pass an option, maybe like %{minify: true}, to format the output like so: <p>hello</p><p>world!</p>

For now I've created a function that seems to handle most cases:

def minify(text, "text/html") do
  text
  |> String.replace(">\n", ">")
  |> String.replace(">  ", ">")
  |> String.replace("  <", "<")
end

(There also seems to be issues with extraneous spaces, which the last two lines try to resolve, eg:)

iex(4)> Earmark.as_html! "<ul><li>one</li><li>two</li><li>three</li><li>four</li></ul>"
"<ul>\n  <li>one</li><li>two</li><li>three</li><li>four</li></ul>\n"

[RobertDober]

Sorry for this inconvenience, this will be fixed but I had foreseen to do that later because so far nobody was really impacted (ex_doc does not use the transformer anymore).

I cannot promise to get this into 1.4, it really depends on my upcoming schedule.

A PR would be great the goal is to have quite compact, but reasonable HTML, something like not indented with line breaks before and after block tags (but not all, e.g. not <code>)

[RobertDober]

Oh BTW I was not thinking about an option, this should be the default behavior IMHO

[RobertDober]

However I was talking about something completely different, 😓

I was talking about rendering Markdown to HTML, not HTML, please note that, that kind of input is not even supported (it kinda works by chance)

Earmark.as_html! "<ul><li>one</li><li>two</li><li>three</li><li>four</li></ul>"

Please see #390 and https://github.com/pragdave/earmark#html-blocks

As mentioned this shall be handled RobertDober/earmark_parser#7

I will keep this issue open and change the rendering to a more compact (and possibly minified) output.

But before the related EarmarkParser issue, nothing can be done

[alexgleason]

Solving it with Markdown to HTML is what we're trying to do. My example above was just to demonstrate a case with extraneous spaces, but the issue is mainly about newlines added from Markdown to HTML output. Thank you!

[RobertDober]

IIUC we have two issues

• Rendering is bogus because of two too (cheating 😉) much whitespace (which is a bug)
• HTML is not totally parsed (which is specified) and might lead to unexpected results, will go away with EarmarkParser's issue Parse HTML recursively RobertDober/earmark_parser#7

Therefor in theory I can fix the first bug if I have time 😨 (or could accept a bug request)

Can you confirm please?

[alexgleason]

Rendering is bogus because of two too (cheating 😉) much whitespace (which is a bug)

Yes, you're correct. The issue is extraneous newlines. Thank you!

[feld]

Hello, checking in on this

[rinpatch]

I have a draft commit for this rinpatch@2e8b4dc, it works perfectly in my testing, however causes 110 tests to fail due to expected fixtures containing newlines. Unfortunately my attempts to automate the fixture rework ran into some edge cases and modifying it all manually is not my idea of fun to say the least. I might get back to it at some point, but would appreciate if someone who has a smarter idea could attempt it.

[RobertDober]

@rinpatch can you please make a PR against branch i391_minify_html, I will gladly have a look