TIL that `empty?` is still significantly faster than `any?` despite `any?` being lazy.

any_size_benchmark.rb

require 'benchmark'

array = 10000000.times.to_a

puts "Any"
puts Benchmark.measure { 1000.times { array.any? } }

puts "Empty"
puts Benchmark.measure { 1000.times { array.empty? } }

puts "Iterating over the entire collection"
puts Benchmark.measure { 1000.times { array.each {} } }

output.txt

Any
  0.000000   0.000000   0.000000 (  0.000303)
Empty
  0.000000   0.000000   0.000000 (  0.000076)
Iterating over the entire collection
467.250000   0.150000 467.400000 (468.296172)

I thought this might be because the you're comparing C code to ruby interpreted code. But you you can also supply a block to any?

require 'benchmark'

array = 10000000.times.to_a

puts "Any"
puts Benchmark.measure { 1000.times { array.any? } }

puts "Any with a block"
puts Benchmark.measure { 1000.times { array.any?{|obj|obj} } }

puts "Empty"
puts Benchmark.measure { 1000.times { array.empty? } }

puts "Iterating over the entire collection"
puts Benchmark.measure { 10.times { array.each {} } }

Note I reduced the number of iterations on each because it's so damn slow.

Output:

Any
  0.000000   0.000000   0.000000 (  0.000272)
Any with a block
  0.000000   0.000000   0.000000 (  0.000329)
Empty
  0.000000   0.000000   0.000000 (  0.000050)
Iterating over the entire collection
  4.050000   0.000000   4.050000 (  4.053021)

I was a little surprised by this, but perhaps it optimises for the default implementation??

require 'benchmark'

array = 10000000.times.to_a

puts "Any"
puts Benchmark.measure { 1000.times { array.any? } }

puts "Any with a block"
puts Benchmark.measure { 1000.times { array.any?{|obj|obj} } }

puts "Any with a more complex block"
puts Benchmark.measure { 10.times { array.any?{|obj| obj * 2 == 5} } }

puts "Empty"
puts Benchmark.measure { 1000.times { array.empty? } }

puts "Iterating over the entire collection"
puts Benchmark.measure { 10.times { array.each {} } }

Note I have to drop the any? complex block benchmark down to 10 iterations because it's so slow.

Output:

Any
  0.000000   0.000000   0.000000 (  0.000266)
Any with a block
  0.000000   0.000000   0.000000 (  0.000323)
Any with a more complex block
  7.530000   0.000000   7.530000 (  7.528715)
Empty
  0.000000   0.000000   0.000000 (  0.000051)
Iterating over the entire collection
  4.010000   0.000000   4.010000 (  4.013209)

The complex example you provide is the worst case scenario, it never returns true so iterates over the entire collection.

Is it the extra logic cruft around detecting no block was passed, and deciding what to do about it... ?