BigBinary Blog: Ruby 2.5

Ruby 2.5 added lazy proc allocation for block parameters

2018-05-22T00:00:00.000Z

irb> def greet
irb>   yield
irb> end
  => :greet
irb>
irb> def greet_with_welcome(&block)
irb>   puts 'Welcome'
irb>   greet(&block)
irb> end
  => :greet_with_welcome
irb> greet_with_welcome { p 'BigBinary' }
Welcome
"BigBinary"
  => "BigBinary"

In Ruby 2.4 when we pass a block to a method, which further passes to another method, Ruby creates a new Proc object by the given block before passing this proc to the another method.

This creates unnecessary objects even when the block parameter is not accessed. It also creates a chain of Proc objects when the block parameter is passed through various methods.

Proc creation is one a heavyweight operation because we need to store all local variables (represented by Env objects in MRI internal) in the heap.

Ruby 2.5 introduced a lazy proc allocation. Ruby 2.5 will not create a Proc object when passing a block to another method. Instead, it will pass the block information. If the block is accessed somewhere else, then it creates a Proc object by the given block.

This results in lesser memory allocation and faster execution.

Ruby 2.4

irb> require 'benchmark'
  => true
irb> def greet
irb>   yield
irb> end
  => :greet
irb>
irb> def greet_with_welcome(&block)
irb>   puts 'Welcome'
irb>   greet(&block)
irb> end
  => :greet_with_welcome
irb>
irb> Benchmark.measure { 1000.times { greet_with_welcome { 'BigBinary' } } }
Welcome
Welcome
...
...
...
  => #

Ruby 2.5

irb> require 'benchmark'
  => true
irb> def greet
irb>   yield
irb> end
  => :greet
irb>
irb> def greet_with_welcome(&block)
irb>   puts 'Welcome'
irb>   greet(&block)
irb> end
  => :greet_with_welcome
irb>
  irb> Benchmark.measure { 1000.times { greet_with_welcome { 'BigBinary' } } }
Welcome
Welcome
...
...
...
  => #

As we can see, there is considerable improvement in execution time when a block param is passed in Ruby 2.5.

Here is the relevant commit and discussion.

Ruby 2.5 adds Thread.report_on_exception by default

2018-04-18T00:00:00.000Z

Let's see what happens when an exception is raised inside a thread.

division_thread = Thread.new do
  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

sleep 1

puts "In the main thread"

Execution of it looks like this.

$ RBENV_VERSION=2.4.0 ruby thread_example_1.rb

Calculating 4/0 in division_thread

In the main thread

Note that the last two lines from the block were not printed. Also notice that after failing in the thread the program continued to run in main thread. That's why we got the message "In the main thread".

This is because the default behavior of Ruby is to silently ignore exceptions in threads and then to continue to execute in the main thread.

Enabling abort_on_exception to stop on failure

If we want an exception in a thread to stop further processing both in the thread and in the main thread then we can enable Thread[.#]abort_on_exception on that thread to achieve that.

Notice that in the below code we are using Thread.current.

division_thread = Thread.new do
  Thread.current.abort_on_exception = true

  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

sleep 1

puts "In the main thread"

$ RBENV_VERSION=2.4.0 ruby thread_example_2.rb

Calculating 4/0 in division_thread

thread_example_2.rb:5:in `/': divided by 0 (ZeroDivisionError)
  from thread_example_2.rb:5:in `block in '

As we can see once an exception was encountered in the thread then processing stopped on both in the thread and in the main thread.

Note that Thread.current.abort_on_exception = true activates this behavior only for the current thread.

If we want this behavior globally for all the threads then we need to use Thread.abort_on_exception = true.

Running program with debug flag to stop on failure

Let's run the original code with --debug option.

$ RBENV_VERSION=2.4.0 ruby --debug thread_example_1.rb

thread_example_1.rb:1: warning: assigned but unused variable - division_thread

Calculating 4/0 in division_thread

Exception `ZeroDivisionError' at thread_example_1.rb:3 - divided by 0
Exception `ZeroDivisionError' at thread_example_1.rb:7 - divided by 0
thread_example_1.rb:3:in `/': divided by 0 (ZeroDivisionError)
  from thread_example_1.rb:3:in `block in '

In this case the exception is printed in detail and the code in main thread was not executed.

Usually when we execute a program with --debug option then the behavior of the program does not change. We expect the program to print more stuff but we do not expect behavior to change. However in this case the --debug option changes the behavior of the program.

Running program with join on thread to stop on failure

If a thread raises an exception and abort_on_exception and $DEBUG flags are not set then that exception will be processed at the time of joining of the thread.

division_thread = Thread.new do
  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

division_thread.join

puts "In the main thread"

$ RBENV_VERSION=2.4.0 ruby thread_example_3.rb

Calculating 4/0 in division_thread

thread_example_3.rb:3:in `/': divided by 0 (ZeroDivisionError)
  from thread_example_3.rb:3:in `block in '

Both Thread#join and Thread#value will stop processing in the thread and in the main thread once an exception is encountered.

Introduction of report_on_exception in Ruby 2.4

Almost 6 years ago, Charles Nutter (headius) had proposed that the exceptions raised in threads should be automatically logged and reported, by default. To make his point, he explained issues similar to what we discussed above about the Ruby's behavior of silently ignoring exceptions in threads. Here is the relevant discussion on his proposal.

Following are some of the notable points discussed.

Enabling Thread[.#]abort_on_exception, by default, is not always a good idea.
There should be a flag which, if enabled, would print the thread-killing exception info.
In many cases, people spawn one-off threads which are not hard-referenced using Thread#join or Thread#value. Such threads gets garbage collected. Should it report the thread-killing exception at the time of garbage collection if such a flag is enabled?
Should it warn using Warning#warn or redirect to STDERR device while reporting?

Charles Nutter suggested that a configurable global flag Thread.report_on_exception and instance-level flag Thread#report_on_exception should be implemented having its default value as true. When set to true, it should report print exception information.

Matz and other core members approved that Thread[.#]report_on_exception can be implemented having its default value set to false.

Charles Nutter, Benoit Daloze and other people demanded that it should be true by default so that programmers can be aware of the silently disappearing threads because of exceptions.

Shyouhei Urabe advised that due to some technical challenges, the default value should be set to false so as this feature could land in Ruby. Once this feature is in then the default value can be changed in a later release.

Nobuyoshi Nakada (nobu) pushed an implementation for Thread[.#]report_on_exception with a default value set to false. It was released in Ruby 2.4.0.

Let's try enabling report_on_exception globally using Thread.report_on_exception.

Thread.report_on_exception = true

division_thread = Thread.new do
  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

addition_thread = Thread.new do
  puts "Calculating nil+4 in addition_thread"
  puts "Result is: #{nil+4}"
  puts "Exiting from addition_thread"
end

sleep 1

puts "In the main thread"

$ RBENV_VERSION=2.4.0 ruby thread_example_4.rb

Calculating 4/0 in division_thread

# terminated with exception:
thread_example_4.rb:5:in `/': divided by 0 (ZeroDivisionError)
  from thread_example_4.rb:5:in `block in '

Calculating nil+4 in addition_thread

# terminated with exception:
thread_example_4.rb:11:in `block in ': undefined method `+' for nil:NilClass (NoMethodError)

In the main thread

It now reports the exceptions in all threads. It prints that the Thread:0x007fb10f018200 was terminated with exception: divided by 0 (ZeroDivisionError). Similarly, another thread Thread:0x007fb10f01aca8 was terminated with exception: undefined method '+' for nil:NilClass (NoMethodError).

Instead of enabling it globally for all threads, we can enable it for a particular thread using instance-level Thread#report_on_exception.

division_thread = Thread.new do
  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

addition_thread = Thread.new do
  Thread.current.report_on_exception = true

  puts "Calculating nil+4 in addition_thread"
  puts "Result is: #{nil+4}"
  puts "Exiting from addition_thread"
end

sleep 1

puts "In the main thread"

In the above case we have enabled report_on_exception flag just for addition_thread.

Let's execute it.

$ RBENV_VERSION=2.4.0 ruby thread_example_5.rb

Calculating 4/0 in division_thread

Calculating nil+4 in addition_thread

# terminated with exception:
thread_example_5.rb:11:in `block in ': undefined method `+' for nil:NilClass (NoMethodError)

In the main thread

Notice how it didn't report the exception which killed thread division_thread. As expected, it reported the exception that killed thread addition_thread.

With the above changes ruby reports the exception as soon as it encounters. However if these threads are joined then they will still raise exception.

division_thread = Thread.new do
  Thread.current.report_on_exception = true

  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

begin
  division_thread.join
rescue => exception
  puts "Explicitly caught - #{exception.class}: #{exception.message}"
end

puts "In the main thread"

$ RBENV_VERSION=2.4.0 ruby thread_example_6.rb

Calculating 4/0 in division_thread

# terminated with exception:
thread_example_6.rb:5:in `/': divided by 0 (ZeroDivisionError)
  from thread_example_6.rb:5:in `block in '

Explicitly caught - ZeroDivisionError: divided by 0

In the main thread

See how we were still be able to handle the exception raised in division_thread above after joining it despite it reported it before due to Thread#report_on_exception flag.

report_on_exception defaults to true in Ruby 2.5

Benoit Daloze (eregon) strongly advocated that both the Thread.report_on_exception and Thread#report_on_exception should have default value as true. Here is the relevant feature request.

After approval from Matz, Benoit Daloze pushed the implementation by fixing the failing tests and silencing the unnecessary verbose warnings.

It was released as part of Ruby 2.5.

Now in ruby 2.5 we can simply write like this.

division_thread = Thread.new do
  puts "Calculating 4/0 in division_thread"
  puts "Result is: #{4/0}"
  puts "Exiting from division_thread"
end

addition_thread = Thread.new do
  puts "Calculating nil+4 in addition_thread"
  puts "Result is: #{nil+4}"
  puts "Exiting from addition_thread"
end

sleep 1

puts "In the main thread"

Let's execute it with Ruby 2.5.

$ RBENV_VERSION=2.5.0 ruby thread_example_7.rb

Calculating 4/0 in division_thread

# terminated with exception (report_on_exception is true):
Traceback (most recent call last):
  1: from thread_example_7.rb:3:in `block in '
thread_example_7.rb:3:in `/': divided by 0 (ZeroDivisionError)

Calculating nil+4 in addition_thread

# terminated with exception (report_on_exception is true):
Traceback (most recent call last):
thread_example_7.rb:9:in `block in ': undefined method `+' for nil:NilClass (NoMethodError)

In the main thread

We can disable the thread exception reporting globally using Thread.report_on_exception = false or for a particular thread using Thread.current.report_on_exception = false.

Future Possibilities

In addition to this feature, Charles Nutter also suggested that it will be good if there exists a callback handler which can accept a block to be executed when a thread dies due to an exception. The callback handler can be at global level or it can be for a specific thread.

Thread.on_exception do
  # some stuff
end

In the absence of such handler libraries need to resort to custom code to handle exceptions. Here is how Sidekiq handles exceptions raised in threads.

Important thing to note is that report_on_exception does not change behavior of the code. It does more reporting when a thread dies and when it comes to thread dies more reporting is a good thing.

Ruby 2.5 supports measuring branch and method coverages

2018-04-11T00:00:00.000Z

Ruby comes with Coverage, a simple standard library for test coverage measurement for a long time.

Before Ruby 2.5

Before Ruby 2.5, we could measure just the line coverage using Coverage.

Line coverage tells us whether a line is executed or not. If executed, then how many times that line was executed.

We have a file called score.rb.

score = 33

if score >= 40
  p :PASSED
else
  p :FAILED
end

Now create another file score_coverage.rb.

require "coverage"

Coverage.start
load "score.rb"
p Coverage.result

We used Coverage#start method to measure the coverage of score.rb file. Coverage#result returns the coverage result.

Let's run it with Ruby 2.4.

$ RBENV_VERSION=2.4.0 ruby score_coverage.rb
:FAILED
{ "score.rb"=> [1, nil, 1, 0, nil, 1, nil] }

Let's look at the output. Each value in the array [1, nil, 1, 0, nil, 1, nil] denotes the count of line executions by the interpreter for each line in score.rb file.

This array is also called the "line coverage" of score.rb file.

A nil value in line coverage array means coverage is disabled for that particular line number or it is not a relevant line. Lines like else, end and blank lines have line coverage disabled.

Here's how we can read above line coverage result.

Line number 1 (i.e. 0th index in the above result array) was executed once.
Coverage was disabled for line number 2 (i.e. index 1) as it is blank.
Line number 3 (i.e. index 2) was executed once.
Line number 4 did not execute.
Coverage was disabled for line number 5 as it contains only else clause.
Line number 6 was executed once.
Coverage was disabled for line number 7 as it contains just end keyword.

After Ruby 2.5

There was a pull request opened in 2014 to add method coverage and decision coverage metrics in Ruby. It was rejected by Yusuke Endoh as he saw some issues with it and mentioned that he was also working on a similar implementation.

In Ruby 2.5, Yusuke Endoh added branch coverage and method coverage feature to the Coverage library.

Let's see what's changed in Coverage library in Ruby 2.5.

Line Coverage

If we execute above example using Ruby 2.5, we will see no change in the result.

$ RBENV_VERSION=2.5.0 ruby score_coverage.rb
:FAILED
{ "score.rb" => [1, nil, 1, 0, nil, 1, nil] }

This behavior is maintained to ensure that the Coverage#start API stays 100% backward compatible.

If we explicitly enable lines option on Coverage#start method in the above score_coverage.rb file, the coverage result will be different now.

require "coverage"

Coverage.start(lines: true)
load "score.rb"
p Coverage.result

$ RBENV_VERSION=2.5.0 ruby score_coverage.rb
:FAILED
{ "score.rb" => {
    :lines => [1, nil, 1, 0, nil, 1, nil]
  }
}

We can see that the coverage result is now a hash which reads that the score.rb file has lines coverage as [1, nil, 1, 0, nil, 1, nil].

Branch Coverage

Branch coverage helps us identify which branches are executed and which ones are not executed.

Let's see how to get branch coverage.

We will update the score_coverage.rb by enabling branches option.

require "coverage"

Coverage.start(branches: true)
load "score.rb"
p Coverage.result

$ RBENV_VERSION=2.5.0 ruby score_coverage.rb
:FAILED
{ "score.rb" =>
  { :branches => {
      [:if, 0, 3, 0, 7, 3] => {
        [:then, 1, 4, 2, 4, 15] => 0,
        [:else, 2, 6, 2, 6, 15] => 1
      }
    }
  }
}

Here is how to read the data in array.

[
  BRANCH_TYPE,
  UNIQUE_ID,
  START_LINE_NUMBER,
  START_COLUMN_NUMBER,
  END_LINE_NUMBER,
  END_COLUMN_NUMBER
]

Please note that column numbers start from 0 and line numbers start from 1.

Let's try to read above printed branch coverage result.

[:if, 0, 3, 0, 7, 3] reads that if statement starts at line 3 & column 0 and ends at line 7 & column 3.

[:then, 1, 4, 2, 4, 15] reads that then clause starts at line 4 & column 2 and ends at line 4 & column 15.

Similarly, [:else, 2, 6, 2, 6, 15] reads that else clause starts at line 6 & column 2 and ends at line 6 & column 15.

Most importantly as per the branch coverage format, we can see that the branch from if to then was never executed since COUNTER is 0. The another branch from if to else was executed once since COUNTER is 1.

Method Coverage

Measuring method coverage helps us identify which methods were invoked and which were not.

We have a file grade_calculator.rb.

students_scores = { "Sam" => [53, 91, 72],
                    "Anna" => [91, 97, 95],
                    "Bob" => [33, 69, 63] }

def average(scores)
  scores.reduce(&:+)/scores.size
end

def grade(average_score)
  case average_score
  when 90.0..100.0 then :A
  when 80.0..90.0 then :B
  when 70.0..80.0 then :C
  when 60.0..70.0 then :D
  else :F
  end
end

def greet
  puts "Congratulations!"
end

def warn
  puts "Try hard next time!"
end


students_scores.each do |student_name, scores|
  achieved_grade = grade(average(scores))

  puts "#{student_name}, you've got '#{achieved_grade}' grade."

  if achieved_grade == :A
    greet
  elsif achieved_grade == :F
    warn
  end

  puts
end

To measure method coverage of above file, let's create grade_calculator_coverage.rb by enabling methods option on Coverage#start method.

require "coverage"

Coverage.start(methods: true)
load "grade_calculator.rb"
p Coverage.result

Let's run it using Ruby 2.5.

$ RBENV_VERSION=2.5.0 ruby grade_calculator_coverage.rb
Sam, you've got 'C' grade.

Anna, you've got 'A' grade.
Congratulations!

Bob, you've got 'F' grade.
Try hard next time!

{ "grade_calculator.rb" => {
    :methods => {
      [Object, :warn, 23, 0, 25, 3] => 1,
      [Object, :greet, 19, 0, 21, 3] => 1,
      [Object, :grade, 9, 0, 17, 3] => 3,
      [Object, :average, 5, 0, 7, 3] => 3
    }
  }
}

The format of method coverage result is defined as shown below.

[ CLASS_NAME,
  METHOD_NAME,
  START_LINE_NUMBER,
  START_COLUMN_NUMBER,
  END_LINE_NUMBER,
  END_COLUMN_NUMBER ]

Therefore, [Object, :grade, 9, 0, 17, 3] => 3 reads that the Object#grade method which starts from line 9 & column 0 to line 17 & column 3 was invoked 3 times.

Conclusion

We can measure all coverages at once also.

Coverage.start(lines: true, branches: true, methods: true)

What's the use of these different types of coverages anyway?

Well, one use case is to integrate this in a test suite and to determine which lines, branches and methods are executed and which ones are not executed by the test. Further, we can sum up these and evaluate total coverage of a test suite.

Author of this feature, Yusuke Endoh has released coverage-helpers gem which allows further advanced manipulation and processing of coverage results obtained using Coverage#result.

Ruby 2.5 adds Exception#full_message method

2018-03-13T00:00:00.000Z

Before Ruby 2.5, if we want to log a caught exception, we would need to format it ourselves.

class AverageService
  attr_reader :numbers, :coerced_numbers

  def initialize(numbers)
    @numbers = numbers
    @coerced_numbers = coerce_numbers
  end

  def average
    sum / count
  end

  private

  def coerce_numbers
    numbers.map do |number|
      begin
        Float(number)
      rescue Exception => exception
        puts "#{exception.message} (#{exception.class})\n\t#{exception.backtrace.join("\n\t")}"
        puts "Coercing '#{number}' as 0.0\n\n"

        0.0
      end
    end
  end

  def sum
    coerced_numbers.map(&:to_f).sum
  end

  def count
    coerced_numbers.size.to_f
  end
end

average = AverageService.new(ARGV).average
puts "Average is: #{average}"

$ RBENV_VERSION=2.4.0 ruby average_service.rb 5 4f 7 1s0
invalid value for Float(): "4f" (ArgumentError)
	average_service.rb:18:in `Float'
	average_service.rb:18:in `block in coerce_numbers'
	average_service.rb:16:in `map'
	average_service.rb:16:in `coerce_numbers'
	average_service.rb:6:in `initialize'
	average_service.rb:37:in `new'
	average_service.rb:37:in `'

Coercing '4f' as 0.0

invalid value for Float(): "1s0" (ArgumentError)
	average_service.rb:18:in `Float'
	average_service.rb:18:in `block in coerce_numbers'
	average_service.rb:16:in `map'
	average_service.rb:16:in `coerce_numbers'
	average_service.rb:6:in `initialize'
	average_service.rb:37:in `new'
	average_service.rb:37:in `'

Coercing '1s0' as 0.0

Average of [5.0, 0.0, 7.0, 0.0] is: 3.0

It was proposed that there should be a simple method to print the caught exception using the same format that ruby uses while printing an uncaught exception.

Some of the proposed method names were display, formatted, to_formatted_s, long_message, and full_message.

Matz approved the Exception#full_message method name.

In Ruby 2.5, we can re-write above example as follows.

class AverageService
  attr_reader :numbers, :coerced_numbers

  def initialize(numbers)
    @numbers = numbers
    @coerced_numbers = coerce_numbers
  end

  def average
    sum / count
  end

  private

  def coerce_numbers
    numbers.map do |number|
      begin
        Float(number)
      rescue Exception => exception
        puts exception.full_message
        puts "Coercing '#{number}' as 0.0\n\n"

        0.0
      end
    end
  end

  def sum
    coerced_numbers.map(&:to_f).sum
  end

  def count
    coerced_numbers.size.to_f
  end
end

average = AverageService.new(ARGV).average
puts "Average is: #{average}"

$ RBENV_VERSION=2.5.0 ruby average_service.rb 5 4f 7 1s0
Traceback (most recent call last):
	6: from average_service.rb:37:in `'
	5: from average_service.rb:37:in `new'
	4: from average_service.rb:6:in `initialize'
	3: from average_service.rb:16:in `coerce_numbers'
	2: from average_service.rb:16:in `map'
	1: from average_service.rb:18:in `block in coerce_numbers'
average_service.rb:18:in `Float': invalid value for Float(): "4f" (ArgumentError)

Coercing '4f' as 0.0

Traceback (most recent call last):
	6: from average_service.rb:37:in `'
	5: from average_service.rb:37:in `new'
	4: from average_service.rb:6:in `initialize'
	3: from average_service.rb:16:in `coerce_numbers'
	2: from average_service.rb:16:in `map'
	1: from average_service.rb:18:in `block in coerce_numbers'
average_service.rb:18:in `Float': invalid value for Float(): "1s0" (ArgumentError)

Coercing '1s0' as 0.0

Average of [5.0, 0.0, 7.0, 0.0] is: 3.0

Note that, Ruby 2.5 prints exception backtrace in reverse order if STDERR is unchanged and is a TTY as discussed in our previous blog post.

Ruby 2.5 prints backtrace & error message in reverse

2018-03-07T00:00:00.000Z

Stack trace or backtrace is a sequential representation of the stack of method calls in a program which gets printed when an exception is raised. It is often used to find out the exact location in a program from where the exception was raised.

Before Ruby 2.5

Before Ruby 2.5, the printed backtrace contained the exception class and the error message at the top. Next line contained where in the program the exception was raised. Next we got more lines which contained cascaded method calls.

Consider a simple Ruby program.

class DivisionService
  attr_reader :a, :b

  def initialize(a, b)
    @a, @b = a.to_i, b.to_i
  end

  def divide
    puts a / b
  end
end

DivisionService.new(ARGV[0], ARGV[1]).divide

Let's execute it using Ruby 2.4.

$ RBENV_VERSION=2.4.0 ruby division_service.rb 5 0

division_service.rb:9:in `/': divided by 0 (ZeroDivisionError)
	from division_service.rb:9:in `divide'
	from division_service.rb:13:in `'

In the printed backtrace above, the first line shows the location, error message and the exception class name; whereas the subsequent lines shows the caller method names and their locations. Each line in the backtrace above is often considered as a stack frame placed on the call stack.

Most of the times, a backtrace has so many lines that it makes it very difficult to fit the whole backtrace in the visible viewport of the terminal.

Since the backtrace is printed in top to bottom order the meaningful information like error message, exception class and the exact location where the exception was raised is displayed at top of the backtrace. It means developers often need to scroll to the top in the terminal window to find out what went wrong.

After Ruby 2.5

Over 4 years ago an issue was created to make printing of backtrace in reverse order configurable.

After much discussion Nobuyoshi Nakada made the commit to print backtrace and error message in reverse order only when the error output device (STDERR) is a TTY (i.e. a terminal). Message will not be printed in reverse order if the original STDERR is attached to something like a File object.

Look at the code here where the check happens if STDERR is a TTY and is unchanged.

Let's execute the same program using Ruby 2.5.

$ RBENV_VERSION=2.5.0 ruby division_service.rb 5 0

Traceback (most recent call last):
	2: from division_service.rb:13:in `'
	1: from division_service.rb:9:in `divide'
division_service.rb:9:in `/': divided by 0 (ZeroDivisionError)

$

We can notice two new changes in the above backtrace.

The error message and exception class is printed last (i.e. at the bottom).
The stack also adds frame number when printing in reverse order.

This feature makes the debugging convenient when the backtrace size is a quite big and cannot fit in the terminal window. We can easily see the error message without scrolling up now.

Note that, the Exception#backtrace attribute still holds an array of stack frames like before in the top to bottom order.

So if we rescue the caught exception and print the backtrace manually

class DivisionService
  attr_reader :a, :b

  def initialize(a, b)
    @a, @b = a.to_i, b.to_i
  end

  def divide
    puts a / b
  end
end

begin
  DivisionService.new(ARGV[0], ARGV[1]).divide
rescue Exception => e
  puts "#{e.class}: #{e.message}"
  puts e.backtrace.join("\n")
end

we will get the old behavior.

$ RBENV_VERSION=2.5.0 ruby division_service.rb 5 0

ZeroDivisionError: divided by 0
division_service.rb:9:in `/'
division_service.rb:9:in `divide'
division_service.rb:16:in `'

$

Also, note that if we assign STDERR with a File object, thus making it a non-TTY

puts "STDERR is a TTY? [before]: #{$stderr.tty?}"
$stderr = File.new("stderr.log", "w")
$stderr.sync = true
puts "STDERR is a TTY? [after]: #{$stderr.tty?}"

class DivisionService
  attr_reader :a, :b

  def initialize(a, b)
    @a, @b = a.to_i, b.to_i
  end

  def divide
    puts a / b
  end
end

DivisionService.new(ARGV[0], ARGV[1]).divide

we can get the old behavior but the backtrace would be written to the specified file and not to STDERR.

$ RBENV_VERSION=2.5.0 ruby division_service.rb 5 0

STDERR is a TTY? [before]: true
STDERR is a TTY? [after]: false

$ cat stderr.log

division_service.rb:14:in `/': divided by 0 (ZeroDivisionError)
	from division_service.rb:14:in `divide'
	from division_service.rb:18:in `'

$

This feature has been tagged as experimental feature. What it means is that Ruby team is gathering feedback on this feature.

Ruby 2.5 added Hash#slice method

2018-02-06T00:00:00.000Z

Ruby 2.4

Let's say we have a hash { id: 1, name: 'Ruby 2.5', description: 'BigBinary Blog' } and we want to select key value pairs having the keys name and description.

We can use the Hash#select method.

irb> blog = { id: 1, name: 'Ruby 2.5', description: 'BigBinary Blog' }
  => {:id=>1, :name=>"Ruby 2.5", :description=>"BigBinary Blog"}

irb> blog.select { |key, value| [:name, :description].include?(key) }
  => {:name=>"Ruby 2.5", :description=>"BigBinary Blog"}

Matzbara Masanao proposed a simple method to take care of this problem.

Some of the names proposed were choice and pick.

Matz suggested the name slice since this method is ActiveSupport compatible.

Ruby 2.5.0

irb> blog = { id: 1, name: 'Ruby 2.5', description: 'BigBinary Blog' }
  => {:id=>1, :name=>"Ruby 2.5", :description=>"BigBinary Blog"}

irb> blog.slice(:name, :description)
  => {:name=>"Ruby 2.5", :description=>"BigBinary Blog"}

So, now we can use a simple method slice to select key value pairs from a hash with specified keys.

Here is the relevant commit and discussion.

Ruby 2.5 allows creating structs with keyword arguments

2018-01-16T00:00:00.000Z

In Ruby, structs can be created using positional arguments.

Customer = Struct.new(:name, :email)
Customer.new("John", "john@example.com")

This approach works when the arguments list is short. When arguments list increases then it gets harder to track which position maps to which value.

Here if we pass keyword argument then we won't get any error. But the values are not what we wanted.

Customer.new(name: "John", email: "john@example.com")
=> #"John", :email=>"john@example.com"}, email=nil>

Ruby 2.5 introduced creating structs using keyword arguments. Relevant pull request is here.

However this introduces a problem. How do we indicate to Struct if we want to pass arguments using position or keywords.

Takashi Kokubun suggested to use keyword_argument as an identifier.

Customer = Struct.new(:name, :email, keyword_argument: true)
Customer.create(name: "John", email: "john@example.com")

Matz suggested to change the name to keyword_init.

So in Ruby 2.5 we can create structs using keywords as long as we are passing keyword_init.

Customer = Struct.new(:name, :email, keyword_init: true)

Customer.new(name: "John", email: "john@example.com")
=> #

Ruby 2.5 adds Hash#transform_keys method

2018-01-09T00:00:00.000Z

Ruby 2.4 added Hash#transform_values method to transform values of the hash.

In Ruby 2.5, a similar method Hash#transform_keys is added for transforming keys of the hash.

>> h = { name: "John", email: "john@example.com" }
=> {:name=>"John", :email=>"john@example.com"}

>> h.transform_keys { |k| k.to_s }
=> {"name"=>"John", "email"=>"john@example.com"}

The bang sibling of this method, Hash#transform_keys! is also added which changes the hash in place.

These two methods are already present in Active Support from Rails and are natively supported in Ruby now.

Rails master is already supporting using the native methods if supported by the Ruby version.

Ruby 2.5 enumerable predicates accept pattern argument

2018-01-02T00:00:00.000Z

Ruby 2.5.0 was recently released.

Ruby has sequence predicates such as all?, none?, one? and any? which take a block and evaluate that by passing every element of the sequence to it.

if queries.any? { |sql| /LEFT OUTER JOIN/i =~ sql }
logger.log "Left outer join detected"
end

Ruby 2.5 allows using a shorthand for this by passing a pattern argument. Internally case equality operator(===) is used against every element of the sequence and the pattern argument.

if queries.any?(/LEFT OUTER JOIN/i)
logger.log "Left outer join detected"
end

# Translates to:

queries.any? { |sql| /LEFT OUTER JOIN/i === sql }

This allows us to write concise and shorthand expressions where block is only used for comparisons. This feature is applicable to all?, none?, one? and any? methods.

Similarities with Enumerable#grep

This feature is based on how Enumerable#grep works. grep returns an array of every element in the sequence for which the case equality operator(===) returns true by applying the pattern. In this case, the all? and friends return true or false.

There is a proposal to add it for select and reject as well.

Ruby 2.5 requires pp by default

2017-12-20T00:00:00.000Z

Ruby 2.5.0-preview1 was recently released.

Ruby allows pretty printing of objects using pp method.

Before Ruby 2.5, we had to require PP explicitly before using it. Even the official documentation states that "All examples assume you have loaded the PP class with require 'pp'".

>> months = %w(January February March)
=> ["January", "February", "March"]

>> pp months
NoMethodError: undefined method `pp' for main:Object
Did you mean?  p
	from (irb):5
	from /Users/prathamesh/.rbenv/versions/2.4.1/bin/irb:11:

>> require 'pp'
=> true

>> pp months
["January",
 "February",
 "March"]
=> ["January", "February", "March"]

In Ruby 2.5, we don't need to require pp. It gets required by default. We can use it directly.

>> months = %w(January February March)
=> ["January", "February", "March"]

>> pp months
["January",
 "February",
 "March"]
=> ["January", "February", "March"]

This feature was added after Ruby 2.5.0 preview 1 was released, so it's not present in the preview. It's present in Ruby trunk.

Array#prepend and Array#append in Ruby 2.5

2017-12-19T00:00:00.000Z

Ruby has Array#unshift to prepend an element to the start of an array and Array#push to append an element to the end of an array.

The names of these methods are not very intuitive. Active Support from Rails already has aliases for the unshift and push methods , namely prepend and append.

In Ruby 2.5, these methods are added in the Ruby language itself.

>> a = ["hello"]
=> ["hello"]
>> a.append "world"
=> ["hello", "world"]
>> a.prepend "Hey"
=> ["Hey", "hello", "world"]
>>

They are implemented as aliases to the original unshift and push methods so there is no change in the behavior.

Ruby 2.5 added yield_self

2017-12-12T00:00:00.000Z

Ruby 2.5 added a new method named yield_self. It yields the receiver to the given block and returns output of the last statement in the block.

irb> "Hello".yield_self { |str| str + " World" }
  => "Hello World"

How is it different from `try` in Rails ?

Without a method argument try behaves similar to yield_self. It would yield to the given block unless the receiver is nil and returns the output of the last statement in the block.

irb> "Hello".try { |str| str + " World" }
  => "Hello World"

Couple of differences to note are, try is not part of Ruby but Rails. Also try's main purpose is protection against nil hence it doesn't execute the block if receiver is nil.

irb> nil.yield_self { |obj| "Hello World" }
  => "Hello World"

irb> nil.try { |obj| "Hello World" }
  => nil

What about `tap`?

tap also is similar to yield_self. It's part of Ruby itself. The only difference is the value that is returned. tap returns the receiver itself while yield_self returns the output of the block.

irb> "Hello".yield_self { |str| str + " World" }
  => "Hello World"

irb> "Hello".tap { |str| str + " World" }
  => "Hello"

Overall, yield_self improves readability of the code by promoting chaining over nested function calls. Here is an example of both the styles.

irb> add_greeting = -> (str) { "HELLO " + str }
irb> to_upper = -> (str) { str.upcase }

# with new `yield_self`
irb> "world".yield_self(&to_upper)
            .yield_self(&add_greeting)
  => "HELLO WORLD"

# nested function calls
irb> add_greeting.call(to_upper.call("world"))
  => "HELLO WORLD"

yield_self is part of Kernel and hence it's available to all the objects.

Ruby 2.5 added delete_prefix and delete_suffix methods

2017-11-28T00:00:00.000Z

Ruby 2.4

Let's say that we have a string Projects::CategoriesController and we want to remove Controller. We can use chomp method.

irb> "Projects::CategoriesController".chomp("Controller")
=> "Projects::Categories"

However if we want to remove Projects:: from the string then there is no corresponding method of chomp. We need to resort to sub.

irb> "Projects::CategoriesController".sub(/Projects::/, '')
=> "CategoriesController"

Naotoshi Seo did not like using regular expression for such a simple task. He proposed that Ruby should have a method for taking care of such tasks.

Some of the names proposed were remove_prefix, deprefix, lchomp, remove_prefix and head_chomp.

Matz suggested the name delete_prefix and this method was born.

Ruby 2.5.0-preview1

irb> "Projects::CategoriesController".delete_prefix("Projects::")
=> "CategoriesController"

Now in order to delete prefix we can use delete_prefix and to delete suffix we could use chomp. This did not feel right. So for symmetry delete_suffix was added.

irb> "Projects::CategoriesController".delete_suffix("Controller")
=> "Projects::Categories"

Read up on this discussion to learn more about how elixir, go, python, and PHP deal with similar requirements.

Ruby 2.5 introduces Dir.children and Dir.each_child

2017-11-21T00:00:00.000Z

Dir.entries is a method present in Ruby 2.4. It returns the output of shell command ls -a in an array.


> Dir.entries("/Users/john/Desktop/test")
> => [".", "..", ".config", "program.rb", "group.txt"]
>

We also have method Dir.foreach which iterates and yields each value from the output of ls -a command to the block.


> Dir.foreach("/Users/john/Desktop/test") { |child| puts child }
> .
> ..
> .config
> program.rb
> group.txt
> test2
>

We can see that the output includes the directives for current directory and parent directory which are "." and "..".

When we want to have access only to the children files and directories, we do not need the [".", ".."] subarray.

This is a very common use case and we'll probably have to do something like Dir.entries(path) - [".", ".."] to achieve the desired output.

To overcome such issues, Ruby 2.5 introduced Dir.children. It returns the output of ls -a command without the directives for current and parent directories.


> Dir.children("/Users/mohitnatoo/Desktop/test")
> => [".config", "program.rb", "group.txt"]
>

Additionally, we can use Dir.each_child method to avoid yielding current and parent directory directives while iterating,


> Dir.each_child("/Users/mohitnatoo/Desktop/test") { |child| puts child }
> .config
> program.rb
> group.txt
> test2
>

As noted in the discussion the names were chosen to match with existing methods Pathname#children and Pathname#each_child.

These additions seem like simple features. Well the issue was posted more than two years ago.

Ruby 2.5 allows rescue/else/ensure inside do/end blocks

2017-10-24T00:00:00.000Z

Ruby 2.4

irb> array_from_user = [4, 2, 0, 1]
  => [4, 2, 0, 1]

irb> array_from_user.each do |number|
irb>   p 10 / number
irb> rescue ZeroDivisionError => exception
irb>   p exception
irb>   next
irb> end
SyntaxError: (irb):4: syntax error, unexpected keyword_rescue,
expecting keyword_end
rescue ZeroDivisionError => exception
      ^

Ruby 2.4 throws an error when we try to use rescue/else/ensure inside do/end blocks.

Ruby 2.5.0-preview1

irb> array_from_user = [4, 2, 0, 1]
  => [4, 2, 0, 1]
irb> array_from_user.each do |number|
irb>   p 10 / number
irb> rescue ZeroDivisionError => exception
irb>   p exception
irb>   next
irb> end
2
5
#
10
 => [4, 2, 0, 1]

Ruby 2.5 supports rescue/else/ensure inside do/end blocks.

Here is relevant commit and discussion.

Ruby 2.5 has removed top level constant lookup

2017-10-18T00:00:00.000Z

Ruby 2.4

irb> class Project
irb> end
=> nil

irb> class Category
irb> end
=> nil

irb> Project::Category
(irb):5: warning: toplevel constant Category referenced by Project::Category
 => Category

Ruby 2.4 returns the top level constant with a warning if it is unable to find a constant in the specified scope.

This does not work well in cases where we need constants to be defined with same name at top level and also in the same scope.

Ruby 2.5.0-preview1

irb> class Project
irb> end
=> nil

irb> class Category
irb> end
=> nil

irb> Project::Category
NameError: uninitialized constant Project::Category
Did you mean?  Category
	from (irb):5

Ruby 2.5 throws an error if it is unable to find a constant in the specified scope.

Here is the relevant commit and discussion.

BigBinary Blog: Ruby 2.5

Ruby 2.5 added lazy proc allocation for block parameters

Ruby 2.4

Ruby 2.5

Ruby 2.5 adds Thread.report_on_exception by default

Enabling abort_on_exception to stop on failure

Running program with debug flag to stop on failure

Running program with join on thread to stop on failure

Introduction of report_on_exception in Ruby 2.4

report_on_exception defaults to true in Ruby 2.5

Future Possibilities

Ruby 2.5 supports measuring branch and method coverages

Before Ruby 2.5

After Ruby 2.5

Line Coverage

Branch Coverage

Method Coverage

Conclusion

Ruby 2.5 adds Exception#full_message method

Ruby 2.5 prints backtrace & error message in reverse

Before Ruby 2.5

After Ruby 2.5

Ruby 2.5 added Hash#slice method

Ruby 2.4

Ruby 2.5.0

Ruby 2.5 allows creating structs with keyword arguments

Ruby 2.5 adds Hash#transform_keys method

Ruby 2.5 enumerable predicates accept pattern argument

Similarities with Enumerable#grep

Ruby 2.5 requires pp by default

Array#prepend and Array#append in Ruby 2.5

Ruby 2.5 added yield_self

How is it different from try in Rails ?

What about tap?

Ruby 2.5 added delete_prefix and delete_suffix methods

Ruby 2.4

Ruby 2.5.0-preview1

Ruby 2.5 introduces Dir.children and Dir.each_child

Ruby 2.5 allows rescue/else/ensure inside do/end blocks

Ruby 2.4

Ruby 2.5.0-preview1

Ruby 2.5 has removed top level constant lookup

Ruby 2.4

Ruby 2.5.0-preview1

How is it different from `try` in Rails ?

What about `tap`?