Prefer %w(...) to a literal array?

ruby

Solution

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.

Problem

While entering a simple literal array of strings in to RubyMine: ``` cols = [ "Col1", "Col2" ] ``` RubyMine complained that I should prefer to use `%w` over a literal array: ``` cols = %w(Col1 Col2) ``` It presented a link to a Ruby Style Guide (here), which said: Prefer %w to the literal array syntax when you need an array of strings. ``` # bad STATES = ['draft', 'open', 'closed'] # good STATES = %w(draft open closed) ``` I can see how the `%w` provides potentially more concise code. Aside from brevity, is there any reason to prefer one method over the other?

Original source