I've often heard Ruby's inject method criticized as being "slow." As I rather like the function, and see equivalents in other languages, I'm curious if it's merely Ruby's implementation of the method that's slow, or if it is inherently a slow way to do things (e.g. should be avoided for non-small collections)?
+2
A:
inject
is like fold
, and can be very efficient in other languages, fold_left
specifically, since it's tail-recursive.
nlucaroni
2008-11-25 21:13:08
+2
A:
It's mostly an implementation issue, but this gives you a good idea of the comparison:
$ ruby -v
ruby 1.8.7 (2008-08-11 patchlevel 72) [i486-linux]
$ ruby exp/each_v_inject.rb
Rehearsal -----------------------------------------------------
loop 0.000000 0.000000 0.000000 ( 0.000178)
fixnums each 0.790000 0.280000 1.070000 ( 1.078589)
fixnums each add 1.010000 0.290000 1.300000 ( 1.297733)
Enumerable#inject 1.900000 0.430000 2.330000 ( 2.330083)
-------------------------------------------- total: 4.700000sec
user system total real
loop 0.000000 0.000000 0.000000 ( 0.000178)
fixnums each 0.760000 0.300000 1.060000 ( 1.079252)
fixnums each add 1.030000 0.280000 1.310000 ( 1.305888)
Enumerable#inject 1.850000 0.490000 2.340000 ( 2.340341)
exp/each_v_inject.rb
require 'benchmark'
total = (ENV['TOTAL'] || 1_000).to_i
fixnums = Array.new(total) {|x| x}
Benchmark.bmbm do |x|
x.report("loop") do
total.times { }
end
x.report("fixnums each") do
total.times do |i|
fixnums.each {|x| x}
end
end
x.report("fixnums each add") do
total.times do |i|
v = 0
fixnums.each {|x| v += x}
end
end
x.report("Enumerable#inject") do
total.times do |i|
fixnums.inject(0) {|a,x| a + x }
end
end
end
So yes it is slow, but as improvements occur in the implementation it should become a non-issue. There is nothing inherent about WHAT it is doing that requires it to be slower.
dgtized
2008-11-26 21:43:29