Java: Understand multiple parse mode flags specified in a regular expression string

[geoffw0]

Understand / correctly parse multiple parse mode flags specified in a regular expression string, for example i and s in:

(?is)abc.*

Previously we were only recognizing the first flag (i).

This is a backport of part of the Swift PR #13715 .

[atorralba]

Changes look plausible to me.

However, there's this test expectation diff:

--- expected
+++ actual
@@ -118,8 +118,9 @@
 | Test.java:16:95:16:101 | t{,15}+ | [RegExpQuantifier] |
 | Test.java:16:102:16:102 | u | [RegExpConstant,RegExpNormalChar] |
 | Test.java:16:102:16:108 | u{16,}+ | [RegExpQuantifier] |
-| Test.java:17:10:17:13 | (?i) | [RegExpZeroWidthMatch] |
+| Test.java:17:10:17:13 | (?i) | [RegExpGroup] |
 | Test.java:17:10:17:36 | (?i)(?=a)(?!b)(?<=c)(?<!d)+ | [RegExpSequence] |
+| Test.java:17:12:17:12 | i | [RegExpConstant,RegExpNormalChar] |
 | Test.java:17:14:17:18 | (?=a) | [RegExpPositiveLookahead] |
 | Test.java:17:17:17:17 | a | [RegExpConstant,RegExpNormalChar] |
 | Test.java:17:19:17:23 | (?!b) | [RegExpNegativeLookahead] |

I get the change from RegExpZeroWidthMatch to RegExpGroup, but the fact that the i gets classified as a RegExpConstant/RegExpNormalChar kind of bugs me. Happy to accept the changes if they look good to you though, just wanted to double check.

[geoffw0]

the fact that the i gets classified as a RegExpConstant/RegExpNormalChar kind of bugs me.

I'm not sure of the right way to fix this. Add flagGroup to the zeroWidthMatch predicate???

[atorralba]

I'm not sure of the right way to fix this. Add flagGroup to the zeroWidthMatch predicate???

🤷 not sure if it's even a problem. @erik-krogh any insight? Can we safely merge this?