Remove unnecessary check

[zwhitchcox]

This seems faster, because you are not assigning a variable, then doing a check, then possibly assigning another variable. You're just assigning one variable.

[zpao]

It seems faster in code but it adds N-1 DOM operations. Speculatively, that will be more expensive (though will be browser dependent, but even setting a property to the same value likely needs to cross the bridge from JS to C++). I'm going to say that we should just leave this as is. Thanks though!

[zwhitchcox]

Cool, thanks for the info

[jimfb]

@zpao I'm not sure I buy that argument. Any browser that had the value available in JS (ie. available for read) would probably check that value for a change before passing it into C++. I think most browsers do have the value available in JS and do check before passing the write to C++ (the exception being moderately old versions of IE). Anyway, I give it a 51% chance that @zwhitchcox's code is actually better/faster for all the cases we care about.

@zwhitchcox I'd be interested in seeing a benchmark. If it is faster, I think it would be reasonable to take this change.

[jimfb]

cc @sebmarkbage (he has a dream of always doing writes)

[sebmarkbage]

Write all the things!

The only case this would not be faster is when we can compare to a JS model and that is local. This is comparing to a DOM object. So yea, I agree with @jimfb and OP.

Sometimes, a setter will cause other side-effects though so we have to be careful with that. This seems fine though.

[zwhitchcox]

Well, if we're really looking for optimization...

function updateOptions(inst, multiple, propValue) {
  propValue = [].concat(propValue) // makes propValue an array even if it's just a single value
  ReactDOMComponentTree
    .getNodeFromInstance(inst)
    .options
    .forEach(function(opt) {
      (propValue.indexOf(opt.value) < 0 ? opt.selected : !opt.selected) &&
        opt.selected = !opt.selected
    })
}

[zwhitchcox]

Nevermind...I broke things