I also noticed that if we use Array.prototype.some, the loop runs until it finds a "falsy" value and then it stops, having less "looping", an even shorter source and minified code:

export default (obj, path, def, undef) => {
	path = path.split ? path.split(".") : path;
	path.some(key => !(obj = obj ? obj[key] : undef));
	return obj === undef ? def : obj;
};

export default((a,b,c,d)=>(b=b.split?b.split("."):b,b.some(b=>!(a=a?a[b]:d)),a===d?c:a));

Then again, maybe is a performance thing.

Using an arrow function instead of a function the source is smaller, and also the minified code:

Hmm, did you use npm run build? It should also transpile code to ES5, while in your example it stays ES6 (hence the size difference).

@RReverser maybe making a PR to microbundle so the "module" build preserves the ES6 syntax instead of turning it into ES5? Also, even using microbundle, the other option I mentioned in #32 (comment) has a smaller size:

export default function(t,n,e,r){return(n=n.split?n.split("."):n).some(function(n){return!(t=t?t[n]:r)}),t===r?e:t}

And if you have and object like { foo: "bar" } and you do dlv(obj, 'foo.bar.foobar.0.whatever') it will run twice instead of 5 times (thanks to the some method).

This library targets IE9+, so we can't rely on .some(). Modern output is enabled by default in Microbundle 0.13 (in beta), which should automate the compression optimizations from this PR.

Sorry @developit, I never take IE into consideration. My bad! .... Closing then!