compare inject block vs manual entries

[kbrock]

Someone had asked me about using the block short form vs manually entering the code.

Turns out, in 1.8.7, using the block was 1.7x slower, but since then, the symbol form is a little quicker.

Thanks for this repo

[Arcovion]

Unused variable
How about ARRAY.inject(:+) too?

[kbrock]

lol. must have been a copy paste from another enumerable one. oops

[nateberkopec]

Already covered by Symbol#to_proc IMO

[nateberkopec]

Whoops, didnt mean to unilaterally close it. My bad!

[kbrock]

Made fixes.

Think @nateberkopec is right. that is covered.
Sorry I didn't see it.

agreed to close?

[Arcovion]

Just tested, seems like even when 0 was passed the block ones were still exactly equal.
The symbol one is much faster though:

require "benchmark/ips"

ARRAY = (1..10_000_000).to_a

def fastest
  ARRAY.inject(:+)
end

def fast
  ARRAY.inject(&:+)
end

def slow
  ARRAY.inject { |a, i| a + i }
end

Benchmark.ips do |x|
  x.report('inject symbol') { fastest }
  x.report('inject block') { fast }
  x.report('inject sum')   { slow }

  x.compare!
end

Warming up --------------------------------------
       inject symbol     1.000  i/100ms
        inject block     1.000  i/100ms
          inject sum     1.000  i/100ms
Calculating -------------------------------------
       inject symbol      2.372  (± 0.0%) i/s -     12.000 
        inject block      1.668  (± 0.0%) i/s -      9.000  in   5.398692s
          inject sum      1.667  (± 0.0%) i/s -      9.000  in   5.408312s

Comparison:
       inject symbol:        2.4 i/s
        inject block:        1.7 i/s - 1.42x slower
          inject sum:        1.7 i/s - 1.42x slower

So I think it's worth having a benchmark to show this—even though it's very specific and only applies to #inject/#reduce.
You need to add the benchmark results to the README (I think it was there before, but not in your last commit?)

[kbrock]

aah - symbol. forgot that one. Yea. there was a reason why I was visiting this code :)

[kbrock]

@Arcovion what version of ruby are you running?
I've tried with small arrays and larger arrays, and they never show sum and block with the same values

for that matter, what version of ruby should I be using for the readme?

[Arcovion]

ruby 2.3.0p0 (2015-12-25 revision 53290) [x86_64-linux]
Doesn't matter what Ruby you use for the benchmark, no problem if the results differ.

[kbrock]

@Arcovion @nateberkopec Is this good for you? Any other changes?

[Arcovion]

Thanks, LGTM.

[nirvdrum]

I think naming the methods after their relative speed is a bit presumptuous. They're subject to change across runtimes and versions. I think it'd be better to name them after the type of operation being performed.

[kbrock]

+1. I was trying to follow the style guide but I agree

[nirvdrum]

Yeah, sorry. I see now that that could be read as an attack. I just meant the convention in general should change, IMHO.

[kbrock]

lol. we're good.