I am wondering why the second map declaration (using the diamond operator) does not compile when the first one does. Compilation error:
error: cannot
Diamond inference doesn't work for instantiating anonymous classes, which is what you're doing here.
Try this:
Map<String, String> map1 = new HashMap<>();
{
map1.put("abc", "abc");
}
Note that you could also omit the diamond altogether. However, while this compiles, it's only because it's ignoring the Java generics and leans on the fact that Java is backwards compatible with previous versions.
Map<String, String> map1 = new HashMap() { //compiles fine
{
put("abc", "abc");
}
};
You don't have a static initializer here (the keyword static is missing altogether).
Basically you create a new anonymous subclass of HashMap
and define the instance intializer block here. Btw, this only works since HashMap
is not final.
Since you'll get an anonymous subclass of HashMap
the diamond operator doesn't work here, since the subclass would then be compiled as if you wrote ... extends HashMap<Object, Object>
and this clearly isn't compatible to Map<String, String>
.
This feature is part of Project Coin 2 and will be avalaible in Java 9 coming September 22 of 2016.
It's call Allowing the diamond syntax with some anonymous class
constructors
.
Link.