pry-auto_benching.rb

pry-auto_benching.rb

• Introduction
• Configuration (optional)
• Examples
• Accuracy
• Install
• License

Introduction

Automatic benchmarking inside the Pry REPL.

Configuration (optional)

Configuration can be skipped unless you want to change the defaults.
To change the defaults, open ~/.pryrc in your editor of choice and change
one or more of the following:

Pry.configure do |config|
  #
  # Start benchmarking when Pry starts?
  # The default is true.
  #
  config.auto_benching.enabled = false

  #
  # The max number of previous benchmark results to store in memory,
  # for use by `auto-benching --past`. The default is 70.
  #
  config.auto_benching.max_history_size = 120

  #
  # The type of clock to use, default is Process::CLOCK_MONOTONIC.
  # See https://www.rubydoc.info/stdlib/core/Process:clock_gettime
  #
  config.auto_benching.clock_type = Process::CLOCK_REALTIME

  #
  # Configure where benchmark results will be displayed.
  # Accepted options are:
  #
  # * prompt
  #   The default, results are written to the prompt.
  #
  # * none
  #   Results are not displayed but are still recorded and can be
  #   viewed on demand by running 'auto-benching --past'.
  #
  config.auto_benching.target_display =  ':prompt | :none'

  #
  # Define a Proc that decides whether or not to record or display benchmark
  # results. The default condition is 'duration >= 0.01'.
  #
  config.auto_benching.display_duration_if = ->(pry, duration) { duration >= 0.02 }

  #
  # The color to use for benchmark results when `target_display` is :prompt.
  # The default is :green.
  #
  config.auto_benching.prompt_color = :red
end

Examples

1.

Get started by reading the help menu:

[1] pry(main)> auto-benching -h
Usage: auto-benching [options] [enable|disable]

The auto-benching command can enable, disable, and
provide other features related to pry-auto_benching.rb

    -v, --version             Display version.
    -p, --past                Display past benchmark results.
    -t, --target-display      Choose the target display for results. Valid options are: prompt, none
    -h, --help                Show this message.

2.

The prompt is updated with benchmark results, this happens to be the default:

[1] pry(main)> sleep 0.2
 => 0
[2] 0.2s (main)> Net::HTTP.get_response URI.parse('https://github.com')
 => #<Net::HTTPOK 200 OK readbody=true>
[3] 1.34s (main)> Net::HTTP.get_response URI.parse('https://www.ruby-lang.org')
 => #<Net::HTTPOK 200 OK readbody=true>
[4] 1.53s (main)>

3.

auto-benching --past shows past benchmark results:

[1] 1.30s (main)> auto-benching -past
 1 0.01 sleep 0.01
 2 0.02 sleep 0.02
 3 0.03 sleep 0.03
 4 1.30 sleep 1.3
 5 1.40 Net::HTTP.get_response(URI.parse('https://github.com'))
[2] pry(main)>

4.

Benchmarking can be temporarily disabled then enabled again:

[1] pry(main)> auto-benching disable
pry-auto_benching.rb: stopped benchmarking.
[2] pry(main)> auto-benching enable
pry-auto_benching.rb: started benchmarking.

Accuracy

pry-auto_benching.rb can measure down to 10 milliseconds (0.01 = 10ms).
When I benchmarked sleep 0.01 I saw no overhead in the results:

[1] pry(main)> sleep 0.01
=> 0
[2] 0.01s (main)>

But that's not to say results can't be reported inaccurately since the benchmark
includes a very small number of method calls to Pry internals, and it could
include time spent by other Pry plugins who run before/after eval hooks as well.
I have seen inaccuracies of 10ms (0.01) but not all the time.

Install

gem install pry-auto_benching.rb

License

This project uses the MIT license, see LICENSE.txt for details.