Performance: compile default slots to functions

[KaelWD]

What problem does this feature solve?

The original v-slot implementation was not available in this.$slots, but that was added in 2.6.4. Scoped slots reduce unnecessary re-renders with nested slots, so this change would hopefully provide a performance improvement for component libraries.

What does the proposed API look like?

<foo>default slot</foo>

should output

_c('foo',{scopedSlots:_u([{key:"default",fn:function(){return [_v("default slot")]},proxy:true}])})

Equivalent to <foo v-slot>default slot</foo>

---

Note #9580 could be a problem if lots of people rely on that behaviour.

[lbennett-stacki]

I've got a working version of this locally but I'd like to test it further over the weekend, it's my first look here so have probably naturally oversimplified. I also auto-prettified codegen/index which is a bit of a pain.

WIP: lbennett-stacki@677ac91

[lbennett-stacki]

After running the tests and seeing many failures, my suspicions were confirmed 😛 I corrected for a lot of them and I seem to have the general compilation of the function together but am struggling to handle cases like the following...

  <foo>
    default-slot-content
    <span slot="test">
      test-slot-content
    </span>
  </foo>

With my current changes, foo is the same as <foo v-slot> and this breaks the test slot as no default slot placeholder is declared. Without v-slot, default-slot-content is ignored. I hope to put some more time into this as its beneficial to look around but let me know if you wanted to take it off me for a release.

[KaelWD]

Hmm that could be a problem, maybe why it wasn't done originally. I'm not familiar with the vue compiler but my first thought is to just filter anything without slotTarget

[lbennett-stacki]

Hmm that could be a problem, maybe why it wasn't done originally. I'm not familiar with the vue compiler but my first thought is to just filter anything without slotTarget

@KaelWD I thought the same! It seems this isn't (always?) available but it has to be at some point so I'll keep digging, the behaviour might not even be in the same ground I've already covered 🔍 Maybe its impossible or leads to some tough patterns and we leave it :)

Thought: Anyone know if there is a significantly rewritten compiler for 3.0?

Update: No progress thus far. May end up in the wind.