What's the rationale behind switch/yield in Java?

Question

After upgrading to Java 21, I realized that the promised switch expression has several surprising properties. The most surprising part is the yield keyword which must be inserted if a block is used in a branch. Like in

int x = switch (y) {
   ...
   case 7 -> {
     assert ...;
     8; // this is not allowed; a yield must be added in front
   }
};

While digging for a reason, I found JEP 354 which apparently introduced the syntax. Now, the obscure part is, that the very last paragraph of that documents indicates that such a move would be a bad idea. Are there any insights, why this contradiction was ignored and the yield syntax was introduced? Are there any reasons that would prevent using either a break expr syntax or having expressions at the end of a block in that context as a result?

Answer 1

It's more consistent to require the yield keyword than for it to be auto-inserted into the final expression. Consider the need to differentiate between the following two cases:

//Version 1
default -> {
    if(cond) {
        z=0;
    }
    yield 15;
}

//Version 2
default -> {
    if(cond) {
        yield z=0;
    }
    yield 15;
}