Introducing require_bench
Ruby Bootstrapping Analysis
Ruby app loading slowly, or never?
Discover bootstrapping issues in Ruby by benchmarking “Kernel.require”
Installation
Add this line to your application’s Gemfile:
gem 'require_bench'
And then execute:
$ bundle
Or install it yourself as:
$ gem install require_bench
Usage
Require the library where it will be loaded prior to any other requires you want to benchmark.
require 'require_bench'
By default this gem does nothing, hacks nothing, and has zero effects.
Turn on benchmarking & output
Add an environment variable, however you normally do such things, so that in Ruby:
ENV['REQUIRE_BENCH'] == 'true'
Any value other than 'true'
means RubyBench is still turned off.
Handy Rake Task for Rails:
Require in Rakefile, as follows:
require 'bundler/setup'
require 'require_bench/tasks' # Near the top, just below require 'bundler/setup'!
This will ensure it will load before other stuff.
When running from command line, you will see output as the Rails app boots.
∴ REQUIRE_BENCH=true bundle exec rake require_bench:hello
[RequireBench] 12.179703 /path/to/my_app/config/application
[RequireBench] 0.001726 resque/tasks
[RequireBench] 0.000917 resque/scheduler/tasks
[RequireBench] 0.000011 rake
[RequireBench] 0.000014 active_record
[RequireBench] 0.008673 sprockets/rails/task
[RequireBench] 0.000012 dynamoid
[RequireBench] 0.000004 dynamoid/tasks/database
[RequireBench] 0.000012 raven/integrations/tasks
[RequireBench] 0.003107 rspec/core/rake_task
[RequireBench] 0.000017 csv
[RequireBench] 0.000012 resque/tasks
[RequireBench] 0.000007 resque/scheduler/tasks
[RequireBench] 0.064950 rails/tasks
[RequireBench] 0.003305 rake/testtask
[RequireBench] 0.001886 rubocop/rake_task
[RequireBench] 0.000012 hubspot-ruby
[RequireBench] 2.291259 /path/to/my_app/config/environment.rb
[RequireBench] Slowest Loads by Library, in order
1. 11.914224 /path/to/my_app/config/application
2. 2.153282 /path/to/my_app/config/environment.rb
3. 0.061008 rails
4. 0.010827 sprockets
5. 0.003179 rspec
6. 0.003144 rake
7. 0.003127 resque
8. 0.001543 rubocop
9. 0.000021 dynamoid
10. 0.000016 csv
11. 0.000016 active_record
12. 0.000010 raven
13. 0.000005 hubspot-ruby
==========
14.150402 TOTAL
Output Options
If the output is too noisy from deep libraries you can add a regex to skip benchmarking of files that match.
If the value is set in the shell, it should be a string. RequireBench will split the string by comma, Regexp escape each value, and join together with pipe (|
) to form the regex pattern.
export REQUIRE_BENCH_SKIP_PATTERN=activesupport,rspec
If the ENV['REQUIRE_BENCH_SKIP_PATTERN']
value is set in Ruby, it can be one of:
- a string, to be split by comma, each Regexp escaped, then joined by pipe (
|
) - an array of strings, each to be Regexp escaped, then joined by pipe (
|
) - a Regexp object, which will be used as is.
ENV['REQUIRE_BENCH_SKIP_PATTERN'] = 'activesupport,rspec'
# or
ENV['REQUIRE_BENCH_SKIP_PATTERN'] = [ 'activesupport', 'rspec' ]
# or
ENV['REQUIRE_BENCH_SKIP_PATTERN'] = Regexp.new('activesupport|rspec')
Any file being required that matches the pattern will use the standard, rather than the benchmarked, require.
Fully qualified paths
Fully qualified paths, or any portion thereof, are fine, because the strings are always Regexp escaped.
comments powered by Disqus