Add clarification about commands arguments

[lpinca]

This tweaks a comment to make clear that when the number of arguments is big, using an array results in better performances.

Correct me if I'm wrong.

[luin]

Actually they should have similar performance since ioredis passes arguments to the flatten method of Lodash without checking whether they are arrays: https://github.com/luin/ioredis/blob/master/lib/command.js#L48

[lpinca]

Yes, but using an array of arguments will "remove" this loop.

[luin]

Sounds interesting. I wrote a tests for it: https://gist.github.com/luin/d06770c736866e0571d0. The result shows the former form is little faster. It looks like _.flatten would be faster if the arguments are all plain strings/numbers: https://github.com/lodash/lodash/blob/3.10.0/lodash.src.js#L2122

[lpinca]

That's indeed interesting, I would have never expected that.
Feel free to close this.

[lpinca]

Here is my own test case which confirms my original assumption.
The difference is quite noticeable:

macbook:bench luigi$ node index.js 
without copy loop x 50,886 ops/sec ±0.64% (96 runs sampled)
with copy loop x 33,855 ops/sec ±0.38% (95 runs sampled)
Fastest is without copy loop

[luin]

I ran your test case and got the similar result as yours. I though there may be too few arguments in my previous test case, so I updated it but the results between the two forms are still slight. Don't know why.

[lpinca]

I don't know either, but it doesn't matter, I doubt that someone will ever use a number of arguments big enough to make the difference.

[AVVS]

It could be that array wasnt optimized (diff types of values, not all strings for instance) or some other similar case

On 03 Aug 2015, at 09:56, Luigi Pinca notifications@github.com wrote:

I don't know either, but it doesn't matter, I doubt that someone will ever use an amount of arguments that will make a difference.

—
Reply to this email directly or view it on GitHub.