Add R Markdown support (part deux)

[burodepeper]

• Add better/wider support for using functions as attribute values (ie, {r code=capture.output(dump('fivenum', ''))})
• Use value of engine attribute to parse the contents of the code-block
• Bug: trailing space shouldn't be needed, ie: {r }

[ZigZagT]

The tailing space issue, I found it's more serious than it looks. See the two screenshots:

without tailing space:

with tailing space at line 2:

[burodepeper]

I can't tell for sure from your minimap screenshots, but my first impression is (because that has been the case more often) that the fenced-code-block doesn't close properly until it finds a next ``` instead of the one you intended. If you feel comfortable with it, I invite you to try your hand at a fix. The code that's causing this can be found here.

[ZigZagT]

This should be more clear.
It seems it not recognizes the whole {js}</code> pattern, until a tailing space is added. However, the <code>{r} pattern don't met this issue.

I'll try to fix it

[burodepeper]

Yeah, that's what I meant. In your second example it detects the start of the code-block on line 10, but it doesn't detect the closing element until line 18. Let me know if you figure it out, and feel free to PR any changes into this dev-r-markdown branch.

[ZigZagT]

#139 fix it

[kylebarron]

The trailing space is trivial to fix.

With the current compiled grammar, i.e. for js:

language-markdown/grammars/language-markdown.json

Line 3106 in de31dac

"begin": "^\\s*([`~]{3,})\\s*(\\{?)((?:\\.?)(?:javascript|js|jsx))(?=( |$|{))\\s*(\\{?)([^`\\{\\}]*)(\\}?)$",

The regex includes the group (?=( |$|{)) after the language name, which forces a space (or $ or { after the language name. Changing this to e.g.:

(?=(}| |$|{))

fixes the trailing space problem.

[kylebarron]

r without the trailing space always worked because the grammars/repositories/flavors/rmarkdown.cson file didn't require a space after {r} and that file was included before the general code.