ActiveRecord vs Sequel (require time)

01-requirements.md

$ brew install cloc
$ bundle install

02-loc.rb

require "yaml"

def lib_dir(gem_name)
  gem = Gem::Specification.find_by_name(gem_name)
  File.join(gem.full_gem_path, gem.require_path)
end

def loc(gem_name)
  output = `cloc #{lib_dir(gem_name)} --yaml`
  yaml = output.lines[6..-1].join("\n")
  YAML.load(yaml)["Ruby"]["code"]
end

sequel_loc = loc("sequel")

activerecord_loc = loc("activerecord") +
                   loc("arel") +
                   loc("activemodel") +
                   loc("squeel")

puts "Sequel LOC: #{sequel_loc}"
puts "ActiveRecord LOC: #{activerecord_loc}"

# Sequel LOC:       31277
# ActiveRecord LOC: 27384

03-require-speed.rb

require "benchmark"
require "sqlite3"

def require_sequel
  require "sequel"
  Sequel.sqlite
end

def require_activerecord
  require "active_record"
  ActiveRecord::Base.establish_connection(adapter: "sqlite3", database: ":memory:")
  files("activerecord").grep(/association/).each &method(:require)
  require "squeel"
end

def files(gem_name)
  gem = Gem::Specification.find_by_name(gem_name)
  lib_dir = File.join(gem.full_gem_path, gem.require_path)
  files = Dir["#{lib_dir}/**/*.rb"]
end

puts "Sequel load time:       #{Benchmark.realtime{require_sequel}}"
puts "ActiveRecord load time: #{Benchmark.realtime{require_activerecord}}"

# Sequel load time:       0.08
# ActiveRecord load time: 0.4 (5 times slower)

Gemfile

source "https://rubygems.org"

gem "sqlite3"

gem "sequel", "4.24"

gem "activerecord", "4.2"
gem "squeel"

Why do you require squeel inside the require_activerecord method? If you're benchmarking the require time of both gems I'd assume you only require them.

@dskecse I'm requiring squeel because this is the functionality Sequel offers in addition to Active Record, and it's loaded by default.

Not only that, but you also comparing the loading of full suite of AR functionality vs basic bare Sequel.

@alfuken Yeah, loading bare Sequel is not representative. However, loading Sequel plugins like column_encryption or single_table_inheritance wouldn't be fair either, because with Sequel you don't have to load code for features you're not using, unlike with Active Record where you have to load everything.

Second, comparing things by simply requiring is half the deal, you should be comparing full code initialisation time, since that is an integral part as well.

I was testing require time plus connecting to the database as well. I'm not sure what else you meant.

And, lastly, advocating a library based on a difference in 1/3 of a second boot time, is like saying "you should choose Audi over BMW because it's engine starts in 0.3s faster". This says more about the person comparing them, rather than about compared objects, honestly.

I wasn't claiming that Sequel is superior because you start with lower boot time, it's just one of multiple factors. For me boot time plays a role in developer happiness, so it's not just a theoretical benchmark.