Add more tests for the nullish coalescing operator

[azz]

Q	A
Bug fix?	no
Breaking change?	no (breaks beta 28)
New feature?	no
Deprecations?	no
Spec compliancy?	yes
Tests added/pass?	yes
Fixed tickets	#761 (comment)
License	MIT

Added some more tests for associativity, multiline parsing, and precedence.

Due to this operator being similar to ||, and it being in the LogicalORExpression non-terminal in the grammar, should it move from BinaryExpression to the LogicalExpression AST node?

Update:

?? has been moved to LogicalExpression

[hzoo]

Oh yeah that may be better actually

[azz]

Has been changed from Binary- to LogicalExpression in this PR.

[jridgewell]

Needs tests for || especially:

a || b ?? c;
a ?? b || c;
a && b ?? c;
a ?? b && c;

[jridgewell]

So:

a || b ?? c === (a || b) ?? c
a ?? b || c === (a ?? b) || c

But:

a && b ?? c === (a && b) ?? c
a ?? b && c === a ?? (b && c)

Do we have any data for why this was chosen?

[littledan]

@jridgewell Is that what this code does? In the spec I wrote for ??, it parses exactly like ||, and in particular && groups more tightly.

[jridgewell]

Is that what this code does?

Yes.

In the spec I wrote for ??, it parses exactly like ||, and in particular && groups more tightly.

That appears to be what's happening? I was just curious why it's not tighter than && or looser than ||.

[littledan]

@jridgewell I thought it would make sense to be exactly the same because people are likely to be upgrading their || to ??, and why change anything? If I were to change it, it would be to be looser than ||. Seems like this deserves a bug to discuss further.

[azz]

My intuition is it should be completely hot-swappable with ||. Great for codemods/fixers.

[littledan]

Opened tc39/proposal-nullish-coalescing#15 to discuss this question; for now, the spec is hot-swappable like that.

[jridgewell]

LGTM. We can discuss the precedence in tc39/proposal-nullish-coalescing#15.