You can use objects instead of switches most of the time.
function getInnerText(o){
return o === null? null : {
string: o,
array: o.map(getInnerText).join(""),
object:getInnerText(o["childNodes"])
}[typeis(o)];
}
Update: if you're concerned about the cases evaluating in advance being inefficient (why are you worried about efficiency this early on in the design of the program??) then you can do something like this:
function getInnerText(o){
return o === null? null : {
string: function() { return o;},
array: function() { return o.map(getInnerText).join(""); },
object: function () { return getInnerText(o["childNodes"]; ) }
}[typeis(o)]();
}
This is more onerous to type (or read) than either a switch or an object, but it preserves the benefits of using an object instead of a switch, detailed in the comments section below. This style also makes it more straightforward to spin this out into a proper "class" once it grows up enough.
update2: with proposed syntax extensions for ES.next, this becomes
let getInnerText = o -> ({
string: o -> o,
array: o -> o.map(getInnerText).join(""),
object: o -> getInnerText(o["childNodes"])
}[ typeis o ] || (->null) )(o);