Contextually type the rightmost operand of comma operators and logical 'AND's

[DanielRosenwasser]

See #6196 (comment)

[weswigham]

&& is probably important since we use it's type in narrowing. Also i think its required to support the "short circuiting as ternary" pattern - if ("foo" === a && "foo" || "bar"). If the && doesn't get contextually typed here, I think we get string | "bar" for the rhs type, rather than "foo" | "bar".

[DanielRosenwasser]

You'd just get string actually. Did you mean if ("foo" === a && a || "bar") with your type narrowing change applied?

[weswigham]

Maybe? I'm just pretty sure the rhs operand of && should get contextually typed, mostly so you get consistent results with narrowing within/involving compound binary logic. In my example above I'm pretty sure I want the type of the RHS expression to be contextually typed as "foo"|"bar" rather than as string because this matches how a functionally equivalent ternary would have worked. I think.

[weswigham]

So my point is that if in the statement if ("foo" === a ? "foo" : "bar") the RHS has type "foo" | "bar", the RHS of the statement if ("foo" === a && "foo" || "bar") should be the same. Mostly since if we plan to disallow if ("foo" === a ? "baz" : "bar") (as neither "baz" nor "bar" are assignable to type "foo"), it follows that if ("foo" === a && "baz" || "bar") should be the same type error by extension, since the expression is equivalent.

[DanielRosenwasser]

@weswigham Just to clarify, did you mean to parenthesize the conditional? In other words, by "foo" === a ? "foo" : "bar" you meant "foo" === (a ? "foo" : "bar").

[weswigham]

Aye. Parenthesis.

On Sun, Jan 3, 2016, 12:30 PM Daniel Rosenwasser notifications@github.com
wrote:

@weswigham https://github.com/weswigham Just to clarify, did you mean
to parenthesize the conditional? In other words, by "foo" === a ? "foo" :
"bar" you meant "foo" === (a ? "foo" : "bar").

—
Reply to this email directly or view it on GitHub
#6199 (comment)
.

[DanielRosenwasser]

I put out a fix for this with #6355.