#1216 Markdown lexer emphasis spans should be paragraph-oriented

[Eric Promislow]

According to http://github.github.com/github-flavored-markdown/preview.html
text like these bits:

_starts with underscore

_do_not_emphasize

should *not* be emphasized, because the paragraph ends with an unclosed "_".
Note that in this text:

one two _three
four five_ six

"three four five" only will be emphasized. Leaving the "_" off
after "five" leaves all text in both lines unemphasized.

I'm relying on http://github.github.com/github-flavored-markdown/preview.html
to determine the actual Markdown syntax, since the spec is minimal, even with
github saying the preview is deprecated.

[Neil Hodgson]

OK. I'll leave this to anhy Markdown user.

[Bewied]

Over five years later, same problem :(

Ideally, the parser would recognize that emphasis cannot span paragraphs. I guess checking for [\r?\n][\r?\n] should be both minimal and sufficient.

[rdipardo]

Patched by implementing the above suggestion of Bewied.[1]

Note that the corrected parsing rules don't conform to any standard, whether GitHub's [2] or otherwise.

In fact, GitHub's Markdown editor will only apply the emphatic style when a closing token is seen on the same line as the first:

Evidently Github's parser scans backward from the end of the style region, while LexMarkdown does the opposite. Emulating the former's behaviour would involve a susbtantial re-write, and the original author's comments make clear that broad feature support was never part of the lexer's design.

---

[1]: Two changed type specifiers and one cast were also added to weed out some more signed/unsigned mismatches; new tests are enclosed separately

[2]: This link was the best substitute I could find for the obsolete original one

[rdipardo]

On a second attempt, it wasn't that hard to get LexMarkdown to behave in the GitHub way:

It was just a matter of scanning forward for a closing emphasis marker.

Which one looks better is a matter of taste; but the original issue was really about compliance with GitHub's interpretation of Markdown syntax, so I guess this newer patch set is a better solution.

Note: the earlier test files had misspellings that have now been corrected.

[Neil Hodgson]

The second approach, disallowing emphasis that is not closed, appears more correct to me so that has been committed as
https://github.com/ScintillaOrg/lexilla/commit/ca976aaa2d505e5d7fe5f11a2d7be9b610e2fac0

The type changes were separated and applied first as
https://github.com/ScintillaOrg/lexilla/commit/1ba3e848ab924c54856878880e8ed57cab609510