IdentifierName: support JDK 22 unnamed variables #4848
Conversation
commonquail
force-pushed
the
identifiername-accept-unnamed
branch
2 times, most recently
from
afc0184 to
cc90fe1
Compare
|
|
||
| @Test | ||
| public void unnamedVariables() { | ||
| assume().that(Runtime.version().feature()).isAtLeast(22); |
There was a problem hiding this comment.
Or 21 and --enable-preview --release n
References: google#4847 References: https://openjdk.org/jeps/456
| @@ -278,6 +278,9 @@ private static boolean isConformant(Symbol symbol, String name) { | |||
| if (isStaticVariable(symbol) && isConformantStaticVariableName(name)) { | |||
| return true; | |||
| } | |||
| if (name.isEmpty()) { | |||
There was a problem hiding this comment.
isConformant is also called for methods. That seems ok; javac does not accept an absent method name and rejects the meta-name _ with
/Test.java:5: error: underscore not allowed here
void _() {
and if, improbably, that ever did change, presumably an unnamed method should be treated no differently from an unnamed variable.
A more difficult question is whether something without a name can "conform" to a naming scheme. It seems more like asking the question in the first place is nonsensical. So perhaps this check should be at a higher level inside the checker, and perhaps it should be more explicitly isUnnamed(name).
There was a problem hiding this comment.
A more difficult question is whether something without a name can "conform" to a naming scheme.
I think most definitions you'd give of, say, lowerCamelCase would pass for the empty string, so I think the answer here is just "yes".
| Function<String, String> f = _ -> "bar"; | ||
| String _ = f.apply("foo"); | ||
| } | ||
| catch (Exception _) {} |
There was a problem hiding this comment.
Possibly all these are excessive compared to just
int _ = 1;
|
Thanks for the fix! |
This PR nominally fixes #4847 but it is perhaps more illustrative than usable.