Demonstrates exponential slowdown with string[n] = "x" as n grows.

showseq2.rb

=begin

It seems that Ruby 2.2.4, string[n] = "x" is not O(1).  It was suggested that
this was due to UTF-8 encoding, but I see the same performance with US-ASCII.

Any ideas?

$ ruby --version
ruby 2.2.4p230 (2015-12-16 revision 53155) [x86_64-linux]
$ time ruby -E US-ASCII:US-ASCII storeseq2.rb 1000 1000
["Encodings: ", #<Encoding:US-ASCII>, #<Encoding:US-ASCII>]

real    0m0.636s
user    0m0.584s
sys     0m0.016s

$ time ruby -E US-ASCII:US-ASCII storeseq2.rb 10_000 100
["Encodings: ", #<Encoding:US-ASCII>, #<Encoding:US-ASCII>]

real    0m1.770s
user    0m1.712s
sys     0m0.052s

$ time ruby -E US-ASCII:US-ASCII storeseq2.rb 100_000 10
["Encodings: ", #<Encoding:US-ASCII>, #<Encoding:US-ASCII>]

real    0m12.671s
user    0m12.649s
sys     0m0.012s

$ time ruby -E US-ASCII:US-ASCII storeseq2.rb 1_000_000 1
["Encodings: ", #<Encoding:US-ASCII>, #<Encoding:US-ASCII>]

real    2m4.756s
user    2m3.976s
sys     0m0.660s

$ rvm 1.9.3
$ ruby --version
ruby 1.9.3p484 (2013-11-22 revision 43786) [x86_64-linux]
$ time ruby -E US-ASCII:US-ASCII storeseq2.rb 1_000_000 1
["Encodings: ", #<Encoding:US-ASCII>, #<Encoding:US-ASCII>]

real    0m0.493s
user    0m0.480s
sys     0m0.012s

=end

if ARGV.size != 2
    puts "Usage: STRING-LENGTH REPS"
    exit
end

p ["Encodings: ", Encoding.default_external, Encoding.default_internal]

s = " " * ARGV[0].to_i
ARGV[1].to_i.times do
    ARGV[0].to_i.times do
        |i|
        s[i] = "x"
    end
end

Update: It looks like this is due to default UTF-8 encoding for string literals.

Adding '# encoding: US-ASCII' produces linear time behavior, without needing -E argument.