DEV Community

Brandon Weaver
Brandon Weaver

Posted on • Originally published at Medium on

Ruby 2.7 — Pattern Matching — First Impressions

Ruby 2.7 — Pattern Matching — First Impressions

Shall we grab a sneak peak at what Pattern Matching looks like in Ruby 2.7? The code’s merged, nightly build is on the way, and I’m a rather impatient writer who wants to check out the new presents.

This is my first pass over pattern matching, and there will be more detailed articles coming out as I have more time to experiment with the features.

To be clear: This article will meander as it’s a first impression. I’ll be running more tests on it tomorrow, but wanted to write something on it tonight to see how well first impressions matched up.

The Short Version

This is going to be a long article, and the short version isn’t going to cover even a small portion of it. Each major section will have a single style it covers.

The test file covers a good portion of this if you want an at-a-glance read:

ruby/ruby

Literal Matches

Much like a regular case statement, you can perform literal matches:

case 0
in 0 then true
else false
end
# => true
Enter fullscreen mode Exit fullscreen mode

In these cases it would probably be best to use when instead.

The difference is that if there’s no match it’s going to raise aNoMatchingPattern error rather than returning nil like case / when statements would.

Multiple Matchers

In typical case statements, a comma would be used, but with some of the new syntax this would break some of the destructuring that will be mentioned later:

case 0
in 0 | 1
  true
end
# => true
Enter fullscreen mode Exit fullscreen mode

In the case of captured variables below, they can’t be mixed:

case 0
in a | 0
end
# Error: illegal variable in alternative pattern
Enter fullscreen mode Exit fullscreen mode

Captured Variables

One of the more interesting additions is captured variables:

case true
in a
  a
end
# => true
Enter fullscreen mode Exit fullscreen mode

In Guard Clauses

These can also be used in suffix conditionals or rather guard type clauses in what’s reminiscent of Haskell:

case 0
in a if a == 0
  true
end
# => true
Enter fullscreen mode Exit fullscreen mode

In Assignment Mappings

They can even be assigned with a Hash-like syntax:

case 0
in 0 => a
  a == 0
end
Enter fullscreen mode Exit fullscreen mode

With Placeholders

Most interesting in this section is that they appear to treat underscores as a special placeholder:

case 0
in _ | _a
  true
end
Enter fullscreen mode Exit fullscreen mode

Though this could also be seen as a literal variable assignment, it’s an interesting riff on Scala’s pattern matching wildcards.

Destructuring

Interestingly you can shadow the variables as well while destructuring an array:

case [0, 1]
in a, a
  a == 1
end
# => true
Enter fullscreen mode Exit fullscreen mode

It appears that the last assignment will be a in this case, and that commas are now used for destructuring rather than for denoting a separate pattern to match against as is the case for when.

Deconstruct

In the top of the file there’s a class defined:

class C
  class << self
    attr_accessor :keys
  end

  def initialize(obj)
    @obj = obj
  end

  def deconstruct
    @obj
  end

  def deconstruct_keys(keys)
    C.keys = keys
    @obj
  end
end
Enter fullscreen mode Exit fullscreen mode

It appears to indicate a way to define how an object should be destructured by pattern matching:

[[0, 1], C.new([0, 1])].all? do |i|
  case i
  in 0, 1
    true
  end
end
Enter fullscreen mode Exit fullscreen mode

including accounting for unbalanced matches:

[[0], C.new([0])].all? do |i|
  case i
  in 0, 1
  else
    true
  end
end
Enter fullscreen mode Exit fullscreen mode

I don’t understand how deconstruct_keys is working, but it feels really odd to me from first glance.

With Splats

It also will capture multiple arguments like destructuring currently works on Ruby arrays:

case []
in *a
  a == []
end

case [0, 1, 2]
in *a, 1, 2
  a == [0]
end
Enter fullscreen mode Exit fullscreen mode

Meaning that * could also be used as a bit of a wildcard:

case 0
in *
 true
end
Enter fullscreen mode Exit fullscreen mode

On Hashes

These appear to work a lot like keyword arguments:

case {a: 0}
in a: 0
  true
end
Enter fullscreen mode Exit fullscreen mode

That also means that kwsplats could be used here:

case {}
in **a
  a == {}
end
Enter fullscreen mode Exit fullscreen mode

…and with the keywords bit above, that means that classes could define their own destructuring using keywords, but I’m not clear on that.

I’m really hoping that this also uses === on each param, but the tests don’t indicate this. I’ll be experimenting with this later.

Triple Equals

Like our current case statements, it appears that the new method of pattern matching uses=== for comparisons:

case 'abc'
in /a/
  true
end

case 0
in -> i { i == 0 }
  true
end
Enter fullscreen mode Exit fullscreen mode

Combined with Destructuring

One thing I enjoyed in Qo was the ability to match against another array using === to write queries. It looks like the new syntax will compare each element and give us something very similar:

case [0, 1, 2, 3, 4, 5]
in [0..1, 0...2, 0.., 0..., (...5), (..5)]
  true
end
Enter fullscreen mode Exit fullscreen mode

It’s amusing that they’ve included beginningless and endless ranges in the example, as that could be very useful later on. I certainly hope these work with hashes, as I really really really want this to work:

case object
in method_a: 0..10, method_b: String
 true
end
Enter fullscreen mode Exit fullscreen mode

…because if we can define our own deconstructors just imagine the possibilities there. The class attr for keys is odd though, not sure what to think of that one.

Pin Operator

I’m going to have to come back to this one as I do not understand it. I assume it means not to assign and to “pin” a variable so it doesn’t get overwritten:

a = /a/
case 'abc'
in ^a
  true
end

case [0, 0]
in a, ^a
  a == 0
end
Enter fullscreen mode Exit fullscreen mode

I believe this can be used to capture one element in a variable and also reference it later like a back-reference of sorts.

Thoughts?

This hasn’t built with Nightly yet, so I intend to make a second pass on this after it clears to verify some behavior I have suspicions about. Mostly I want to see about the hash matchers, as there’s some amazing potential there.

Going to be digging through and playing with this over the next few weeks, here’s the next post in the series:

Ruby 2.7 — Pattern Matching — Destructuring on Point

Oldest comments (0)