From 57ccbff43c58d6a2f1dd5c7d27d2e42802754a6a Mon Sep 17 00:00:00 2001
From: Benjamin Clayman <bclayman@squareup.com>
Date: Wed, 6 Oct 2021 18:02:22 -0400
Subject: [PATCH] Proof of concept for improving ContainExactly matcher speed
 when elements obey transitivity

This is a proof of concept approach for addressing issue #1161.

The current implementation for ContainExactly runs in O(n!).  In practice,
it runs in O(n log n) when the elements are comparable and sorting result
in a match.

The crux of the problem is that some elements don't obey transitivity.  As
a result, knowing that sorting actual and expected doesn't result in a match
*doesn't* guarantee that expected and actual don't match.

This proof of concept provides a way for the user to indicate that the elements
in a particular example's expected and actual obey transitivity. That looks like this:

expect(a).to contain_exactly(*b).transitive

And runs in O(n log n) time.  More practically, this means that common
use cases for contains_exactly will enjoy a massive speedup.  Previously,
users have examples where comparing arrays of 30 integers "never finishes."
Using `.transitive` here with arrays of 10,000 integers runs in < 0.1s
on my machine.
---
 .../matchers/built_in/contain_exactly.rb      | 18 ++++++++-
 .../matchers/built_in/contain_exactly_spec.rb | 40 +++++++++++++++++++
 2 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/lib/rspec/matchers/built_in/contain_exactly.rb b/lib/rspec/matchers/built_in/contain_exactly.rb
index 1bf7f9831..f8d50d28b 100644
--- a/lib/rspec/matchers/built_in/contain_exactly.rb
+++ b/lib/rspec/matchers/built_in/contain_exactly.rb
@@ -6,6 +6,19 @@ module BuiltIn
       # Provides the implementation for `contain_exactly` and `match_array`.
       # Not intended to be instantiated directly.
       class ContainExactly < BaseMatcher
+        def initialize(expected=nil)
+          super
+          @transitive = false
+        end
+
+        # @api public
+        # Specifies that elements contained in actual and expected
+        # obey transitivity.  This lets match run much faster.
+        def transitive
+          @transitive = true
+          self
+        end
+
         # @api private
         # @return [String]
         def failure_message
@@ -72,7 +85,10 @@ def message_line(prefix, collection, surface_descriptions=false)
 
         def match(_expected, _actual)
           return false unless convert_actual_to_an_array
-          match_when_sorted? || (extra_items.empty? && missing_items.empty?)
+          matched_when_sorted = match_when_sorted?
+          return true if matched_when_sorted
+          return matched_when_sorted if @transitive
+          (extra_items.empty? && missing_items.empty?)
         end
 
         # This cannot always work (e.g. when dealing with unsortable items,
diff --git a/spec/rspec/matchers/built_in/contain_exactly_spec.rb b/spec/rspec/matchers/built_in/contain_exactly_spec.rb
index ac1996ae8..2723d1259 100644
--- a/spec/rspec/matchers/built_in/contain_exactly_spec.rb
+++ b/spec/rspec/matchers/built_in/contain_exactly_spec.rb
@@ -95,6 +95,46 @@ def array.send; :sent; end
 end
 
 RSpec.describe "using contain_exactly with expect" do
+  # users have reported using contains_exactly with 50 elements
+  # never finishing!
+  context "speeding up matching for elements that obey transitivity" do
+    require 'benchmark'
+    shared_examples "runs very fast" do
+      it do
+        time = Benchmark.realtime do
+          subject
+        end
+        # this is in seconds
+        expect(time).to be < 0.3
+      end
+    end
+
+    let(:a) { Array.new(10_000) { rand(10) } }
+    let(:b) { a.shuffle }
+
+    context "when expected and actual match" do
+      subject { expect(a).to contain_exactly(*b).transitive }
+
+      it "matches" do
+        subject
+      end
+
+      include_examples "runs very fast"
+    end
+
+    context "when expected and actual do not match" do
+      subject { expect(a).not_to contain_exactly(*b).transitive }
+
+      let(:b) { Array.new(10_000) { rand(10) } }
+
+      it "does not match" do
+        subject
+      end
+
+      include_examples "runs very fast"
+    end
+  end
+
   it "passes a valid positive expectation" do
     expect([1, 2]).to contain_exactly(2, 1)
   end