Differ wrongly highlights BigDecimals in a hash where another element doesn't match.

[tomekn]

Hi!

I've encountered an issue with the differ that feels wrong.

Let's take this class as an example...

require 'bigdecimal'

class HashGenerator
  def self.generate
    {
      hello: 'world',
      decimal1: BigDecimal.new("10"),
      decimal2: BigDecimal.new("29.32")
    }
  end
end

Simple enough, let's test that our hash has the right values.

require 'hash_generator'
require 'bigdecimal'

RSpec.describe HashGenerator, '#self.generate' do
  it 'returns the expected hash' do
    expected_hash = {
      hello: 'world',
      decimal1: BigDecimal.new("10"),
      decimal2: BigDecimal.new("29.32")
    }
    expect(HashGenerator.generate).to eq(expected_hash)
  end
end

Green - all good!

Now let's change the expectation by altering the hello string, note that the bigdecimals are unchanged...

require 'hash_generator'
require 'bigdecimal'

RSpec.describe HashGenerator, '#self.generate' do
  it 'returns the expected hash' do
    expected_hash = {
      hello: 'SOME DIFFERENT STRING',
      decimal1: BigDecimal.new("10"),
      decimal2: BigDecimal.new("29.32")
    }
    expect(HashGenerator.generate).to eq(expected_hash)
  end
end

The test fails as expected - but the diff also highlights the bigdecimals as a problem. I understand this is because the BigDecimal inspect contains the object id and the differ probably looks at the stringified version of the hash when it's doing its thing?

Anyway, a more complex permutation of this has led me down a wild goose chase yesterday and I thought I'd raise it here. I'm happy to work on a solution if there is consensus that this is undesired behaviour.

Thanks,

Tom

[myronmarston]

RSpec performs a textual diff using diff-lcs. In this case, the textual representation of the two BigDecimal objects is different, even though they are equal. I agree it's an undesirable behavior but as long as we perform textual diffs there's not much we can do about it.

I think that to fix this we'd probably have to write our own differ from scratch, which is not something we're willing to take on and maintain (nor do we have the time to do so).

[benoittgt]

Closing the issue because of inactivity. I think at the moment you can deal with this with custom matchers. Here is an example:

# run with `ruby rspec_bigdecimal.rb`
require "bundler/inline"
require 'bigdecimal'

gemfile(true) do
  source "https://rubygems.org"
  gem "rspec"
end

require 'rspec/autorun'

RSpec::Matchers.define :a_big_decimal_of do |expected|
  match do |actual|
    BigDecimal(actual.to_s) == BigDecimal(expected.to_s)
  end
end

RSpec::Matchers.define :a_big_decimal_close_to do |expected|
  match do |actual|
    BigDecimal(actual.to_s).round(3) == BigDecimal(expected.to_s).round(3)
  end
end

RSpec.describe "BigDecimal" do
  it "validate big decimales" do
    hash_with_bigdecimal = {
      hello: "world",
      decimal: BigDecimal("10"),
      decimal_rounded: BigDecimal("10.123456789")
    }

    expect(hash_with_bigdecimal).to include({
      decimal: a_big_decimal_of("10"),
      decimal_rounded: a_big_decimal_close_to("10.123")
    })
  end
end

If wanted we can maybe add a matcher.

[JonRowe]

Where did we get to with your differ @benoittgt? That would solve this right?

[benoittgt]

If you speak about rspec/rspec#94

I see it as more dedicated to complicated structure like hash, very long string, deeply nested array. But this case is something the new differ should manage.