optimize __createBinding

[elibarzilay]

When the binding is itself one that was created by __createBinding,
re-use its descriptor, which avoids piling multiple levels of getters in
the case of multiple levels of exports.

Also related to #46744 and to microsoft/tslib#165.

[elibarzilay]

I did the change and added also a bit that does the same optimization of using an existing descriptor if the binding is marked as non-writable and non-configurable -- which means that it will be used as-is for the value too. This makes it possible to hack around the need for a getter if you know that your code does not modify exported bindings (it could be done by tsc in the future). For example:

// @ts-ignore
if (typeof exports === "object") exports = Object.freeze(exports);

(BTW, I removed the redundant parens, but now that the condition is much longer, I added parens around the whole thing which I think makes it more readable, but will drop them if you prefer no redundant parentheses at all...)

[rbuckton]

I'm curious about the conditions here:

• oldDesc.configurable === false could just be !oldDesc.configurable, because configurable is always set by the runtime.
• oldDesc.writable === false means there won't be a "get" in oldDesc, because writable is only added for data properties (and is undefined for accessor properties).
• There's a possibility that oldDesc is undefined (if re-exporting something from a dynamic module), in which case oldDesc.writable could throw

Would the following condition also match the intended behavior here?

Suggested change

	var desc = (oldDesc.writable === false && oldDesc.configurable === false
	|| m.__esModule && "get" in oldDesc)
	? oldDesc : { enumerable: true, get: function() { return m[k]; } };
	var desc = oldDesc && ("get" in oldDesc ? m.__esModule : !oldDesc.writable && !oldDesc.configurable)
	? oldDesc : { enumerable: true, get: function() { return m[k]; } };

By branching on "get" in oldDesc" early, we can just use ! for the writable/configurable tests, since oldDesc.writable will never be false (or true) for an accessor.

[rbuckton]

Alternatively, this could be rewritten into an if and a reassignment (which is a bit more readable):

var desc = Object.getOwnPropertyDescriptor(m, k);
if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
  desc = { enumerable: true, get: function() { return m[k]; } };
}
Object.defineProperty(o, k2, desc);

[elibarzilay]

All of that makes sense, will push that in a minute.

• Re using !, I guess that it makes sense since it's part of the descriptor spec (I was focused on actual node behavior wrt cjs bindings from a frozen exports, which I wanted to make as robust as possible).
• writable vs get -- makes sense.
• And the reassignment is a nice improvement though the resulting negation of the condition makes it more confusing, so I don't care either way.

[rbuckton]

Is there a PR for tslib yet?

[elibarzilay]

Is there a PR for tslib yet?

Made one now: microsoft/tslib#168