Fix destructuring with holes in assign pattern

[magic-akari]

• fix const assign pattern
• optimize the output of let/var assign pattern

Q	A
Fixed Issues?	Fixes #14239
Patch: Bug Fix?	👍
Major: Breaking Change?
Minor: New Feature?
Tests Added + Pass?	Yes
Documentation PR Link
Any Dependency Changes?
License	MIT

[JLHwung]

I think we should fix the issue in pushUnpackedArrayPattern, so we can handle other patterns:

const [...x] = [,];
const [...{0: x}] = [,];

[magic-akari]

It seems Babel has handled it well.

REPL

const [...x] = [,];
const [...{ 0: y }] = [,];

const x = [,];
const y = [,][0];

[JLHwung]

Could you check if const [x] = [,]; is now transformed?

[magic-akari]

Could you check if const [x] = [,]; is now transformed?

I tested it locally. It behaves as same as the REPL.

const x;

It is indeed a bug related to this Pull Request.

But I don't have a clear idea about how to fix it.

For the test case like following

const [ x = 1 ] = [,];

I am sure that const x = 1; is the best transformed output.

But for const [x] = [,], if we give it a value like

const x = void 0;

Meanwhile we will get void 0 when transformed to es5

var x = void 0;

The void 0 seems to be redundant for the "var" declaration.

What do you think about it? @JLHwung

[JLHwung]

@magic-akari

const x = void 0;

looks good to me. As a transpiler we should support valid language usage. The example above could be further transpiled into var x, but that should be addressed by the transform-block-scoping.

[JLHwung]

Thanks!

[magic-akari]

Thanks!

Sorry, not finished.

[magic-akari]

I think it's ready now.

[nicolo-ribaudo]

Thanks!