Update ActionScript regex heuristic

[BenEddy]

Fixes a performance and correctness issue with the ActionScript heuristic regex. Specifically, the regex didn't account for trailing space between the final function argument type and the closing paren.

test_input = "function testFunc ( argOne : Boolean, argTwo : Boolean, argThree : Boolean )"
# Previous regex hangs 
test_input.match(/^\s*(?:package(?:\s+[\w.]+)?\s+(?:{|$)|import\s+[\w.*]+\s*;|(?=.*?(?:intrinsic|extends))(intrinsic\s+)?class\s+[\w<>.]+(?:\s+extends\s+[\w<>.]+)?|(?:(?:public|protected|private|static)\s+)*(?:(?:var|const|local)\s+\w+\s*:\s*[\w<>.]+(?:\s*=.*)?\s*;|function\s+\w+\s*\((?:\s*\w+(?:\s*:\s*[\w<>.]+)?,?)*\)(?:\s*:\s*[\w<>.]+)?))
# New regex terminates
test_input.match(/^\s*(?:package(?:\s+[\w.]+)?\s+(?:{|$)|import\s+[\w.*]+\s*;|(?=.*?(?:intrinsic|extends))(intrinsic\s+)?class\s+[\w<>.]+(?:\s+extends\s+[\w<>.]+)?|(?:(?:public|protected|private|static)\s+)*(?:(?:var|const|local)\s+\w+\s*:\s*[\w<>.]+(?:\s*=.*)?\s*;|function\s+\w+\s*\((?:\s*\w+(?:\s*:\s*[\w<>.]+)?,?)*\s*\)(?:\s*:\s*[\w<>.]+)?))/)
=> #<MatchData "function testFunc (  argOne : Boolean, argTwo : Boolean, argThree : Boolean )" 1:nil>

Checklist:

• I am adding new or changing current functionality
  • I have added or updated the tests for the new or changed functionality.

[zerowidth]

Suggested change

	pattern: '^\s*(?:package(?:\s+[\w.]+)?\s+(?:{|$)|import\s+[\w.*]+\s*;|(?=.*?(?:intrinsic|extends))(intrinsic\s+)?class\s+[\w<>.]+(?:\s+extends\s+[\w<>.]+)?|(?:(?:public|protected|private|static)\s+)*(?:(?:var|const|local)\s+\w+\s*:\s*[\w<>.]+(?:\s*=.*)?\s*;|function\s+\w+\s*\((?:\s*\w+(?:\s*:\s*[\w<>.]+)?,?)*\s*\)(?:\s*:\s*[\w<>.]+)?))'
	pattern: '^\s*(?:package(?:\s+[\w.]+)?\s+(?:{|$)|import\s+[\w.*]+\s*;|(?=.*?(?:intrinsic|extends))(intrinsic\s+)?class\s+[\w<>.]+(?:\s+extends\s+[\w<>.]+)?|(?:(?:public|protected|private|static)\s+)*(?:(?:var|const|local)\s+\w+\s*:\s*[\w<>.]+(?:\s*=.*)?\s*;|function\s+\w+\s*\((?:\s*\w+\s*:\s*[\w<>.]+\s*(,\s*\w+\s*:\s*[\w<>.]+\s*)*\)?\)(?:\s*:\s*[\w<>.]+)?))'

I think a head(, tail)* form for the function arguments would be more resilient and less prone to backtracking:

• (?:\s*\w+(?:\s*:\s*[\w<>.]+)?,?)*\s* has a lot of starred sub-patterns and because the comma separator itself is optional, arg:type could be matched as [arg:type], or [a, rg:type] or [a, r, g:type], etc.
• (?:\s*\w+\s*:\s*[\w<>.]+\s*(,\s*\w+\s*:\s*[\w<>.]+\s*)*\)? changes this to match an argument, which now requires a :type, and optionally matches additional comma-arg pairs.

I'm curious about the [\w<>.]+ pattern allowed to the right of the :. Looking at the AS3 docs I'm not sure if <> ever shows up there. The . could be valid if we want to match ...args, and if we did, we could add that as a trailing pattern: (?:\s*\w+\s*:\s*[\w<>.]+\s*(,\s*\w+\s*:\s*[\w<>.]+\s*)(,\s*\.{3}\s*\w+)?*\)?

[BenEddy]

@zerowidth Looks like the suggested regex is missing a closing paren and doesn't match functions with an arity of 0.

I'm not sure if <> ever shows up there

Heh, no idea 😅 . My main goal with this PR is to fix the acute perf problem and I didn't consider auditing the regex fidelity to AS3 syntax.

[lildude]

LGTM. Thanks @BenEddy 🙇