Uso de gsub vs operadores de regex para limpeza (opcional) de dados.

regex.rb

require 'benchmark/ips'

MATCH     = '123.123.123-99'
NON_MATCH = 'João da Silva'
CPF_REGEX = /\A(\d{3})\.?(\d{3})\.?(\d{3})\-?(\d{2})\z/

def do_gsub(candidate)
  candidate.gsub(CPF_REGEX, '\1\2\3\4')
end

def do_perl(candidate)
  if candidate =~ CPF_REGEX
    "#{$1}#{$2}#{$3}#{$4}"
  else
    candidate
  end
end

Benchmark.ips do |x|
  x.report("match gsub")     { do_gsub(MATCH) }
  x.report("non-match gsub") { do_gsub(NON_MATCH) }
  x.report("match perl")     { do_perl(MATCH) }
  x.report("non-match perl") { do_perl(NON_MATCH) }
  x.compare! # cortesia do andre :)
end

# $ ruby regex.rb 
# Calculating -------------------------------------
#           match gsub    12.489k i/100ms
#       non-match gsub    36.590k i/100ms
#           match perl    20.082k i/100ms
#       non-match perl    56.169k i/100ms
# -------------------------------------------------
#           match gsub    152.499k (±12.8%) i/s -    749.340k
#       non-match gsub    943.747k (±20.3%) i/s -      4.427M
#           match perl    297.897k (±12.4%) i/s -      1.466M
#       non-match perl      1.363M (±19.6%) i/s -      6.459M
# 
# Comparison:
#       non-match perl:  1362968.9 i/s
#       non-match gsub:   943746.8 i/s - 1.44x slower
#           match perl:   297897.2 i/s - 4.58x slower
#           match gsub:   152499.3 i/s - 8.94x slower