Muse: do not allow newlines after opening and before closing "*"

[link2xt]

This should not be highlighted as emphasized:

*
foo bar*

This is unclosed emphasis:

*foo bar
*

[Alhadis]

This is actually impossible to fix due to the limitations of the TextMate grammar system. It processes documents one line at a time, which means something like this:

foo *bar
*

is impossible to differentiate from this:

foo *bar
baz*

I could prevent a lone * from starting an emphasised text run, however:

/^\*$/

[link2xt]

I have it fixed in language-muse, this spec passes:
https://github.com/link2xt/atom-muse/blob/ea85da13a26cd89876ef2bd1ba4bfd3fd7f68f50/spec/muse-spec.coffee#L72-L80

My emphasis parsing is here:
https://github.com/link2xt/atom-muse/blob/ea85da13a26cd89876ef2bd1ba4bfd3fd7f68f50/grammars/muse.cson#L69-L85

[Alhadis]

Yes, that's because Atom recently changed their highlighting-engine to support newlines. However, other TextMate-grammar implementations don't support this (notably GitHub's).

You can see for yourself:

• *\nFoo Bar*
• *Foo bar\n* baz*

[link2xt]

You can see for yourself:
...

First example works correctly, the second I just fixed:
https://github.com/link2xt/atom-muse/commit/0a728c9e918f0ba90406cf5c2c1d635d27e2cc98

Now both examples work correctly on https://github-lightshow.herokuapp.com/

• First
• Second

[Alhadis]

What about this?

*foo bar
* baz

Lightshow is tokenising it as

<pre class="code"><span class="pl-mi">*foo bar</span>
<span class="pl-mi">* baz</span>
</pre>

... but Pandoc gives me:

<p>*foo bar</p>
<h1 id="baz">baz</h1>

[link2xt]

This markup is invalid because * is not closed, so I am not sure here. Pandoc also gives this:

$ pandoc -f muse -t native
*foo bar
----
baz*
[Para [Emph [Str "foo",Space,Str "bar",SoftBreak,Str "----",SoftBreak,Str "baz"]]]

Generally speaking, any block markup should break paragraphs, and it is better to parse it as

[Para [Str "*foo",Space,Str "bar"]
,HorizontalRule
,Para [Str "baz*"]]

because otherwise we can't determine if the second line is a block or not without parsing the third line. Pandoc implementation does backtracking which may be quite expensive in extreme cases, like lots of nested unclosed tags.

In muse.cson emphasis consumes more than it should because I do not parse paragraphs beforehand, it is a bug.

[link2xt]

Alternatively, it can be treated as a part of inline markup, even if there is no closing *. In other workds, unclosed inline markup is undefined behavior.

[Alhadis]

How should this be treated?

Foo **bold**bar baz qux** qul

The Text::Amuse wiki says:

The opening one must be preceded by something which is not an alphanumerical character (or at the beginning of the line) and followed by something which is not a space.

This means that the code above should be rendered like this:

<p>Foo *<em>boldbar baz qux</em>* qul</p>

But, wouldn't it be more natural to render it like this...?

<p>Foo <strong><em>boldbar baz qux</em></strong> qul</p>

[link2xt]

I think it should be

<p>Foo <strong>bold**bar baz qux</strong> qul</p>

Text::Amuse produced this:

<p>
Foo <strong>bold**bar baz qux</strong> qul
</p>

[link2xt]

Pandoc is fixed, now it produces

<p>Foo <strong>bold**bar baz qux</strong> qul</p>

[Alhadis]

Is there a painless way to run an up-to-date Text::Amuse from command-line? muse-quick.pl is deprecated and installing Text::Amuse::Compile is proving to be a pain-in-the-arse due to this shitty MacBook's crippled build system (I don't know what the previous owner did to it, but anything relying on a C compiler is broken...).

All I need is the raw HTML output, I don't give a crap about EPUB or LaTeX, so there's a million CPAN dependencies I don't need which are probably adding to the installation grief.

[link2xt]

You can use this script prepared for GitHub markup repository:
https://github.com/link2xt/markup/blob/muse/lib/github/commands/muse2html

[Alhadis]

Lol, too late.

Thanks anyway though. 😅

[Alhadis]

Okay, I think I've fixed the emphasis matching. Could I get you to update your branch locally and tell me if there's anything I've missed?

Bear in mind there's nothing we can do about this:

Foo *italic*unclosed italic

Which, given TextMate's limitations, is for all intents and purposes the same as:

Foo *italic*unclosed italic
bar* baz