it must crawl the property chain on the object until it eventually reaches 'undefined'
If it were a normal property lookup, yes, but not with `hasOwnProperty' - that's the same as any old hash table lookup. So efficiency here really depends on the implementation, and Google's is probably faster. Regardless, we're really splitting hairs - these optimizations will get a few milliseconds, on average. As much as I am a perfectionist, it's all completely negligible in terms of UX.
I'm talking about the class tracing V8 does, not prototype chains. "Property chain" meaning in V8, it must climb the series of what it calls "hidden classes" in order to find all of the properties. Other engines do various things when you pass a totally dynamic property lookup in. Some slower engines probably use naïve hash tables, I don't know.
Feel free to correct me if I'm wrong on this one, though. I've never really ripped open V8 or any of the other engines to look inside, I just go by what I read from their design docs.
From my reading of the V8 docs, the hidden class climbing is done when properties are added to the class. Each added property causes a transition to the next hidden class. The final hidden class has all the properties as offsets and a failed lookup might chain through prototypes (not sure there) but does not work though the other hidden classes.
If it were a normal property lookup, yes, but not with `hasOwnProperty' - that's the same as any old hash table lookup. So efficiency here really depends on the implementation, and Google's is probably faster. Regardless, we're really splitting hairs - these optimizations will get a few milliseconds, on average. As much as I am a perfectionist, it's all completely negligible in terms of UX.