Voila!, one benchmark:
require 'benchmark'
n = 1_000_000
Benchmark.bm(11) do |b|
b.report('%w') { n.times { %w[a b c d e f g h i j k l m n o p q r s t u v w x y z] } }
b.report('explicit') { n.times { ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z'] } }
b.report('numerics') { n.times { [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26] } }
end
user system total real
%w 2.590000 0.000000 2.590000 ( 2.591225)
explicit 2.590000 0.000000 2.590000 ( 2.584781)
numerics 0.300000 0.000000 0.300000 ( 0.309161)
user system total real
%w 2.590000 0.000000 2.590000 ( 2.591516)
explicit 2.590000 0.000000 2.590000 ( 2.584155)
numerics 0.300000 0.000000 0.300000 ( 0.308896)
user system total real
%w 2.590000 0.000000 2.590000 ( 2.592848)
explicit 2.590000 0.000000 2.590000 ( 2.585558)
numerics 0.300000 0.000000 0.300000 ( 0.308570)
I added the "numerics" array test because I suspected that %w is faster than using explicit strings due to testing for stringiness. %w doesn't need to do that, because it assumes everything is a string. After running it three times it's a wash as far as dealing with strings. Numbers rule, strings drool, and all that.
The previous benchmarks were run using Ruby 1.9.3-p286 on my system at work. I tested again using my old MacBook Pro at home, using Ruby 1.8.7-p358, so the following numbers are slower due to the differences in the hosting hardware, plus running an older Ruby:
user system total real
%w 3.070000 0.000000 3.070000 ( 3.080983)
explicit 3.100000 0.000000 3.100000 ( 3.093083)
numerics 0.950000 0.040000 0.990000 ( 0.990535)
user system total real
%w 3.080000 0.010000 3.090000 ( 3.076787)
explicit 3.090000 0.000000 3.090000 ( 3.089246)
numerics 0.950000 0.030000 0.980000 ( 0.989579)
user system total real
%w 3.080000 0.000000 3.080000 ( 3.073877)
explicit 3.090000 0.000000 3.090000 ( 3.091576)
numerics 0.950000 0.030000 0.980000 ( 0.989132)
On 1.8.7, %w was a tiny bit faster consistently, which probably gave rise to the speed-rumors.
