Slowness of parsing a string read from a gzipped file

results.txt

truffleruby 22.3.0, like ruby 3.0.3, GraalVM CE Native [x86_64-darwin]
Base unit is an array of 20 integers

JSON-decode a string encoding an array of 10 of those base units.

Calculating -------------------------------------
       JSON 10 plain      6.817k (±14.0%) i/s -    133.364k in  19.995768s
      JSON 10 gzdata      1.204k (±28.0%) i/s -     15.611k in  20.047997s
   JSON 10 flattened      6.954k (± 9.6%) i/s -    138.006k in  20.090469s

Comparison:
   JSON 10 flattened:     6954.1 i/s
       JSON 10 plain:     6817.4 i/s - same-ish: difference falls within error
      JSON 10 gzdata:     1203.7 i/s - 5.78x  (± 0.00) slower

JSON-decode a string encoding an array of 500 of those base units.

Calculating -------------------------------------
      JSON 500 plain    111.994  (± 8.9%) i/s -      2.224k in  20.040061s
     JSON 500 gzdata      0.643  (± 0.0%) i/s -     13.000  in  20.250865s
  JSON 500 flattened    105.381  (±10.4%) i/s -      2.079k in  20.000481s

Comparison:
      JSON 500 plain:      112.0 i/s
  JSON 500 flattened:      105.4 i/s - same-ish: difference falls within error
     JSON 500 gzdata:        0.6 i/s - 174.20x  (± 0.00) slower

truffleruby_gzipped_data.rb

require 'json'
require 'zlib'
require 'tmpdir'
require 'benchmark'
require 'oj'


LINES = (ARGV.shift || 300).to_i
BMTIMES = (ARGV.shift || 3).to_i

PLAIN = Dir.tmpdir + "/test.json"
GZIP =  Dir.tmpdir + "/test.json.gz"

arr = (1..LINES).each_with_object([]) do |i, a|
  a << (i..(i+20)).to_a
end

File.open(PLAIN, "w:utf-8") {|f| f.puts arr.to_json}
Zlib::GzipWriter.open(GZIP)  {|f| f.puts arr.to_json}

plain_data = File.read(PLAIN)
gzdata = Zlib::GzipReader.open(GZIP).read
flattened =  Truffle::Debug.flatten_string(gzdata)
forced_copy = gzdata + " "

puts "\n" + RUBY_DESCRIPTION
puts "Benchmarking with an array of #{LINES} 20-element arrays (repeat #{BMTIMES} times)"

puts "\nBEGIN STDLIB JSON"

Benchmark.bm do |x|
  BMTIMES.times do
    puts "\n"
    x.report("%-25s" % "Plain" ) do
      3.times do
        json = JSON.parse(plain_data)
      end
    end

    x.report('%-25s' % "Previously gzipped") do
      3.times do
        json = JSON.parse(gzdata)
      end
    end

    x.report("%-25s" % "Flattened" ) do
      3.times do
        json = JSON.parse(flattened)
      end
    end


    x.report('%-25s' % "Gzipped/forced copy") do
      3.times do
        json = JSON.parse(forced_copy)
      end
    end
  end
end



puts "\n\nBEGIN Oj"

Oj.default_options = {:mode => :compat }

BMTIMES.times do
  Benchmark.bm do |x|
    puts "\n"
    x.report('%-25s' % "Plain" ) do
      3.times do
        json = Oj.load(plain_data)
      end
    end

    x.report('%-25s' % "Previously gzipped") do
      3.times do
        json = Oj.load(gzdata)
      end
    end

    x.report('%-25s' % "Gzipped/forced copy") do
      3.times do
        json = Oj.load(forced_copy)
      end
    end
  end
end