Repository navigation
Whitespace check inconsistency #9973
Description
GWT version: 2.11
Browser (with version): Chrome
Operating System: N/A
Description
GWT emulation of isWhitespace is listing characters explicitly to avoid discrepancies between Java and JS notion of "whitespace", but there is one inconsistency with Java 8 and one inconsistency with Java 11.
For comparison the table below also shows which characters are recognized as whitespace using chr.match(/\s/) in JS (Chrome).
| range | java 8 | java 11 | GWT | JS |
|---|---|---|---|---|
| \u0009-\u000D, \u0020, \u1680, \u2000-\u2006, \u2008-\u200A, \u2028, \u2029, \u205F, \u3000 | yes | yes | yes | yes |
| \u001C-\u001F | yes | yes | yes | no |
| \u00A0,\u2007,\u202F | no | no | no | yes |
| \u180E | yes | no | yes | no |
| \uFEFF | no | no | yes | yes |
Steps to reproduce
- Call
Character.isWhitespace('\uFEFF')in GWT and JVM (version 8 or 11), observe different results. - Call
Character.isWhitespace('\u180E')in GWT and JVM (version 11, Temurin), observe different results.
Known workarounds
Implement own utility method for whitespace detection.
Links
- https://www.compart.com/en/unicode/U+FEFF zero width no break space
- https://www.compart.com/en/unicode/U+180E Mongolian vowel separator
- https://stackoverflow.com/a/4731164/2270026 explicit regex for Java 8
- j2objc emulation https://github.com/google/j2objc/blob/add4718cf098c3565f49576a5f642173cac0c884/jre_emul/android/platform/libcore/ojluni/src/main/java/java/lang/Character.java#L7350
Activity
niloc132 commentedon Jul 8, 2024
With regards to
\u180E, do we consider Java 8 to be broken here? Or do we think that GWT should be aware if it is running "as Java8" or "as Java 11"?tbroyer commentedon Jul 8, 2024
AFAICT, U+180E was a space (Unicode Zs category) priori to Unicode 6.3.0, and Java 8 was based on Unicode 6.2.0 (Java 9 used Unicode 8)
I'd go with the more recent Unicode stance and not consider U+180E a space, even if it means GWT with
-sourceLevel 8behaves differently from a JDK 8 (but that would likely be the same as running Java 8 bytecode on a JDK 8 vs a more recent JDK)String method emulation for Java 11, align Character.isWhitespace wit…