wtf-tools

wtf-tools

Some tools for debugging and profiling Ruby on Rails projects. Included:

• data dumper
• code timing tool
• method tracker - a kind of middle ground between simple timing and full profiling
• SQL tracker - detect where exactly given SQL query originated from

User is responsible for requiring gems necessary for rendering output. For example, when :yaml option is used with WTF?, we expect that YAML is already loaded. Library requirements for specific options are documented below.

Usage examples

Data dumping

WTF? my_var                                   # basic
WTF? my_var, other_var, { some: data }, :pp   # more data, option: pretty-print
WTF? :label, records, :bare, :time            # multiple options given
data.wtf(:time).some_method                   # inline call

Supported options

Prefix
  (default)  WTF (my_file_name/method_name:227): 
  :np        no default prefix
  :nl        new line before the record
  :time      with timestamp: [2014-10-28 12:33:11 +0200]

Formatting
  (default)  simple Ruby inspect
  :pp        pretty-print, require 'pp'
  :yaml      YAML format,  require 'yaml'
  :json      JSON format,  require 'json'
  :csv       CSV format,   require 'csv'
  :text      simple Ruby to_s
  :line      modifier, each object in separate line
  :bare      modifier, ActiveRecord with just id attributes: #<MyClass id: 1234>
  :name      like :line, but with names taken from source file

Output control
  :puts      to STDOUT (default)
  :file      to a separate file in configured location
  :error     raise the string containing data as exception

---

Code timing

class MyClass
  def run_something
    WTF.time {
      # your code
    }
  end
end

Output:

WTF (my_class/run_something:3): 0.001

---

Method tracking

class MyClass
  def run_something
    WTF.track(self)
    # lots of code
    WTF.track_finish
  end
end

Additionally, extra classes/modules can be given:

WTF.track(self, OtherClass, Helpers)

This will create a CSV file in configured location, containing profiling info. Profiling happens only in the methods of the calling class, and any other given class.

How it works: every method in MyClass and OtherClass is overridden, adding resource measuring code. All calls to those methods from the code between track and track_finish are measured (time and memory). Sum amounts are written as CSV file, sorted by total time used.

Example output:

class,method,count,time,heap_mb
,top,0,0.948,0.0
Overview,some_helper,16408,0.351,0.0
Overview,my_records,28,0.172,0.0
Overview,load_data,1,0.166,0.0 
...

Warning: tracking method calls adds time overhead (somewhere from 10% to 3x, depending on number of times the methods were called).

---

SQL tracking

class MyClass
  def run_something
    WTF.sql %q{SELECT * FROM `my_records` WHERE `attribute`='value' AND `etc`}
    WTF.sql /^UPDATE `my_records` SET .*?`some_values`=/, size: 10
    # lots of code
  end
end

This will add a WTF?-style dump in the default location, containing stacktrace where given SQL statement was generated. SQL must match exactly as strings.

---

Configuration

Configure WTF before using the above-mentioned facilities. Rails initializers directory is a good place to put it. Subkeys of output must be lambdas taking one string argument. They are merged into default output options.

WTF.options = {
  files:   "#{Rails.root}/log/wtf",
  output: {
    default: ->(data) { Rails.logger.info(data) },
    redis:   ->(data) { Redis.new.rpush(:wtf, data) },
  }
}

require 'yaml' # to use :yaml option, etc

Requirement: Ruby 2.0, because the technique used involves module prepend-ing, which is not available in Ruby 1.9.

---

License

This is released under the MIT License.

Sponsors

This gem is sponsored and used by SameSystem