Benchmarking the #matches method from Gary Bernhardt's 'selecta' (https://github.com/garybernhardt/selecta). Here I offer an 'unfair' (in that I'm asking the method to check through ~235,000 choices) run, but it highlights to me that this method is not as slow as I first thought.

gistfile1.txt

$ ruby tmp.rb

Rehearsal -------------------------------------------------
file read       0.000000   0.000000   0.000000 (  0.003113)
file to array   0.040000   0.010000   0.050000 (  0.039986)
#matches        0.230000   0.000000   0.230000 (  0.230347)
---------------------------------------- total: 0.280000sec

                    user     system      total        real
file read       0.000000   0.000000   0.000000 (  0.001798)
file to array   0.040000   0.000000   0.040000 (  0.040375)
#matches        0.240000   0.000000   0.240000 (  0.240455)
matches: 2181
choices: 235886

-- ruby tmp.rb --------------------------------------------------------

require "benchmark"

search_string = "term"
choices = File.read("/usr/share/dict/words").split("\n")

def matches(search_string, choices)
  re = search_string.split(//).map(&Regexp.method(:escape)).join('.*')
  re = /#{re}/i
  choices.select { |s| s =~ re }.sort_by(&:length)
end

Benchmark.bmbm do |bm|
  bm.report("file read") do
    File.read("/usr/share/dict/words")
  end

  bm.report("file to array") do
    File.read("/usr/share/dict/words").split("\n")
  end

  bm.report("#matches") do
    matches(search_string, choices)
  end
end

puts "matches: #{matches(search_string, choices).size}"
puts "choices: #{choices.size}"