ID generation improvement?

[crossbario]

ID generation improvement?

[oberstet]

I can see only 1 potential benefit which might be worth: performance.

However, is .toString(36) working on all browsers?

Have you benchmarked that something like return Math.random().toString(36).substr(2,16); is actually faster than the loop?

[Incognito]

Number.toString(36) has been supported since the early days so we should have it in IE.

I ran some jsPerf tests and the shorter code came out about 10%-20% faster: http://jsperf.com/randomstring-tostring-and-for

The other advantage here is a less complex code base (even if it's marginally so). If the . artifact doesn't bother you, it also improves the speed.

[oberstet]

First off, the jsPerf site is really cool! Didn't know that.

Now, I also did some tests: http://jsperf.com/generate-random-id/2

Results:

• with Firefox the modified ID function is slower than the current one
• all other browsers tested were faster
• the variant with most consistent performance across browsers is variant 4

In any case, we are talking about 300k ops/s even for the slowest (Opera with existing function). In practice, it won't matter at all.

I don't bother about code .. it's trivial in any of the variants.

If you still care, send me a pull with a new function ab._newidFast (leaving the old function there/untouched) with implementation return Math.random().toString(36); and change the single line where it is called to call that new function. Add a link to above perf test in a code comment. I'd merge that ..

Thanks!