feat($compile): add one-way collection bindings

[jbedard]

Implements #14039

[gkalpak]

One minor comment. LGTM otherwise.

[gkalpak]

Missing a test to verify the $watchCollection part. (AFAICT, these tests would have passed even if * was ignored.)

[jbedard]

I think the first test verifies it, and this test is just wrong/not-applicable (you can tell I copied them both from the =* tests...).

[petebacondarwin]

Is it worth testing the $watchCollection is being used rather than a deep $watch?

[jbedard]

Updated the tests. Added the one @petebacondarwin suggested, removed the one @gkalpak pointed out was pointless, and added 2 which cause infdig with the non-collection version.

[jbedard]

I wonder if =* should be deprecated since <* is exactly the same? Right now =* has a slightly more complicated action, but I'm pretty sure the logic/perf is the same, and #16550 actually makes it the exact same.

[gkalpak]

They shouldn't be the same. =* should update the parent if the child gets re-assigned, while <* shouldn't, right?

[jbedard]

So you'd think... but it only does $watchCollection on the outer value. So when the inner value changes it won't do anything until the outer value changes, but at that time the outer value should be copied to the inner.

[petebacondarwin]

No point in deprecating anything now @jbedard :-)

[gkalpak]

So when the inner value changes it won't do anything until the outer value changes

😱 😱 😱 Sounds like a bug in =*. So, essentially, we did support <* all along 😛

[jbedard]

@petebacondarwin I was thinking it would be to encourage using <* instead, especially since =* is misleading

@gkalpak yeah... unless I'm missing something?

[jbedard]

Updated the commit to also close #15874 in addition to #14039

[Narretz]

There's an eslint error in compile.js (doublequotes)

[jbedard]

Fixed