See also this benchmark, which seems to have very different results from the scottlogic link: <a href="http://jsperf.com/array-push-vs-array-length23333/2" rel="nofollow">http://jsperf.com/array-push-vs-array-length23333/2</a><p>I just re-ran it on FF24 (nightly), and also saw results where .push beats this kludge.<p>So before everyone runs off and takes 'direct array assignment is fast' as a rule of thumb, please consider that it quite probably isn't.
I prefer readable code to hacks. Unless you absolutely have a hot spot in your code that will benefit, readability should outweigh this. JavaScript engines will catch up.
after reading the underscore source I remember wondering if direct array assignment was really significantly better than calling push. here's a quick link to the perf reference made in the article: <a href="http://www.scottlogic.co.uk/2010/10/javascript-array-performance/" rel="nofollow">http://www.scottlogic.co.uk/2010/10/javascript-array-perform...</a>, oddly interesting is that calling push was faster in chrome than in direct assignment. I wonder what the results would be like in newer browsers...