Hashdiff is a ruby library to compute the smallest difference between two hashes.

It also supports comparing two arrays.

Hashdiff does not monkey-patch any existing class. All features are contained inside the Hashdiff module.

Docs: Documentation

WARNING: Don't use the library for comparing large arrays, say ~10K (see #49).

Given two Hashes A and B, sometimes you face the question: what's the smallest modification that can be made to change A into B?

An algorithm that responds to this question has to do following:

• Generate a list of additions, deletions and changes, so that A + ChangeSet = B and B - ChangeSet = A.
• Compute recursively -- Arrays and Hashes may be nested arbitrarily in A or B.
• Compute the smallest change -- it should recognize similar child Hashes or child Arrays between A and B.

Hashdiff answers the question above using an opinionated approach:

• Hash can be represented as a list of (dot-syntax-path, value) pairs. For example, {a:[{c:2}]} can be represented as ["a[0].c", 2].
• The change set can be represented using the dot-syntax representation. For example, [['-', 'b.x', 3], ['~', 'b.z', 45, 30], ['+', 'b.y', 3]].
• It compares Arrays using the LCS(longest common subsequence) algorithm.
• It recognizes similar Hashes in an Array using a similarity value (0 < similarity <= 1).

To use the gem, add the following to your Gemfile:

gem 'hashdiff'

Two simple hashes:

a = {a:3, b:2}
b = {}

diff = Hashdiff.diff(a, b)
diff.should == [['-', 'a', 3], ['-', 'b', 2]]

More complex hashes:

a = {a:{x:2, y:3, z:4}, b:{x:3, z:45}}
b = {a:{y:3}, b:{y:3, z:30}}

diff = Hashdiff.diff(a, b)
diff.should == [['-', 'a.x', 2], ['-', 'a.z', 4], ['-', 'b.x', 3], ['~', 'b.z', 45, 30], ['+', 'b.y', 3]]

Arrays in hashes:

a = {a:[{x:2, y:3, z:4}, {x:11, y:22, z:33}], b:{x:3, z:45}}
b = {a:[{y:3}, {x:11, z:33}], b:{y:22}}

diff = Hashdiff.best_diff(a, b)
diff.should == [['-', 'a[0].x', 2], ['-', 'a[0].z', 4], ['-', 'a[1].y', 22], ['-', 'b.x', 3], ['-', 'b.z', 45], ['+', 'b.y', 22]]

patch example:

a = {'a' => 3}
b = {'a' => {'a1' => 1, 'a2' => 2}}

diff = Hashdiff.diff(a, b)
Hashdiff.patch!(a, diff).should == b

unpatch example:

a = [{'a' => 1, 'b' => 2, 'c' => 3, 'd' => 4, 'e' => 5}, {'x' => 5, 'y' => 6, 'z' => 3}, 1]
b = [1, {'a' => 1, 'b' => 2, 'c' => 3, 'e' => 5}]

diff = Hashdiff.diff(a, b) # diff two array is OK
Hashdiff.unpatch!(b, diff).should == a

The following options are available: :delimiter, :similarity, :strict, :ignore_keys, :indifferent, :numeric_tolerance, :strip, :case_insensitive, :array_path and :use_lcs

You can specify :delimiter to be something other than the default dot. For example:

a = {a:{x:2, y:3, z:4}, b:{x:3, z:45}}
b = {a:{y:3}, b:{y:3, z:30}}

diff = Hashdiff.diff(a, b, delimiter: '\t')
diff.should == [['-', 'a\tx', 2], ['-', 'a\tz', 4], ['-', 'b\tx', 3], ['~', 'b\tz', 45, 30], ['+', 'b\ty', 3]]

In cases where you have similar hash objects in arrays, you can pass a custom value for :similarity instead of the default 0.8. This is interpreted as a ratio of similarity (default is 80% similar, whereas :similarity => 0.5 would look for at least a 50% similarity).

The :strict option, which defaults to true, specifies whether numeric types are compared on type as well as value. By default, an Integer will never be equal to a Float (e.g. 4 != 4.0). Setting :strict to false makes the comparison looser (e.g. 4 == 4.0).

The :ignore_keys option allows you to specify one or more keys to ignore, which defaults to [] (none). Ignored keys are ignored at all levels. For example:

a = { a: 1, b: { d: 2, a: 3 }, c: 4 }
b = { a: 2, b: { d: 2, a: 7 }, c: 5 }
diff = Hashdiff.diff(a, b, ignore_keys: :a)
diff.should == [['~', 'c', 4, 5]]

If you wish instead to ignore keys at a particlar level you should use a custom comparison method instead. For example:

a = { a: 1, b: { d: 2, a: 3 }, c: 4 }
b = { a: 2, b: { d: 2, a: 7 }, c: 5 }
diff = Hashdiff.diff(a, b) { |path, _e, _a| true if path == 'b.a' } # note '.' is the default delimiter
diff.should == [['~', 'a', 1, 2], ['~', 'c', 4, 5]]

The :indifferent option, which defaults to false, specifies whether to treat hash keys indifferently. Setting :indifferent to true has the effect of ignoring differences between symbol keys (ie. {a: 1} ~= {'a' => 1})

The :numeric_tolerance option allows for a small numeric tolerance.

a = {x:5, y:3.75, z:7}
b = {x:6, y:3.76, z:7}

diff = Hashdiff.diff(a, b, numeric_tolerance: 0.1)
diff.should == [["~", "x", 5, 6]]

The :strip option strips all strings before comparing.

a = {x:5, s:'foo '}
b = {x:6, s:'foo'}

diff = Hashdiff.diff(a, b, numeric_tolerance: 0.1, strip: true)
diff.should == [["~", "x", 5, 6]]

The :case_insensitive option makes string comparisons ignore case.

a = {x:5, s:'FooBar'}
b = {x:6, s:'foobar'}

diff = Hashdiff.diff(a, b, numeric_tolerance: 0.1, case_insensitive: true)
diff.should == [["~", "x", 5, 6]]

The :array_path option represents the path of the diff in an array rather than a string. This can be used to show differences in between hash key types and is useful for patch! when used on hashes without string keys.

a = {x:5}
b = {'x'=>6}

diff = Hashdiff.diff(a, b, array_path: true)
diff.should == [['-', [:x], 5], ['+', ['x'], 6]]

For cases where there are arrays in paths their index will be added to the path.

a = {x:[0,1]}
b = {x:[0,2]}

diff = Hashdiff.diff(a, b, array_path: true)
diff.should == [["-", [:x, 1], 1], ["+", [:x, 1], 2]]

This shouldn't cause problems if you are comparing an array with a hash:

a = {x:{0=>1}}
b = {x:[1]}

diff = Hashdiff.diff(a, b, array_path: true)
diff.should == [["~", [:x], {0=>1}, [1]]]

The :use_lcs option is used to specify whether a Longest common subsequence (LCS) algorithm is used to determine differences in arrays. This defaults to true but can be changed to false for significantly faster array comparisons (O(n) complexity rather than O(n²) for LCS).

When :use_lcs is false the results of array comparisons have a tendency to show changes at indexes rather than additions and subtractions when :use_lcs is true.

Note, currently the :similarity option has no effect when :use_lcs is false.

a = {x: [0, 1, 2]}
b = {x: [0, 2, 2, 3]}

diff = Hashdiff.diff(a, b, use_lcs: false)
diff.should == [["~", "x[1]", 1, 2], ["+", "x[3]", 3]]

It's possible to specify how the values of a key should be compared.

a = {a:'car', b:'boat', c:'plane'}
b = {a:'bus', b:'truck', c:' plan'}

diff = Hashdiff.diff(a, b) do |path, obj1, obj2|
  case path
  when  /a|b|c/
    obj1.length == obj2.length
  end
end

diff.should == [['~', 'b', 'boat', 'truck']]

The yielded params of the comparison block is |path, obj1, obj2|, in which path is the key (or delimited compound key) to the value being compared. When comparing elements in array, the path is with the format array[*]. For example:

a = {a:'car', b:['boat', 'plane'] }
b = {a:'bus', b:['truck', ' plan'] }

diff = Hashdiff.diff(a, b) do |path, obj1, obj2|
  case path
  when 'b[*]'
    obj1.length == obj2.length
  end
end

diff.should == [["~", "a", "car", "bus"], ["~", "b[1]", "plane", " plan"], ["-", "b[0]", "boat"], ["+", "b[0]", "truck"]]

When a comparison block is given, it'll be given priority over other specified options. If the block returns value other than true or false, then the two values will be compared with other specified options.

When used in conjunction with the array_path option, the path passed in as an argument will be an array. When determining the ordering of an array a key of "*" will be used in place of the key[*] field. It is possible, if you have hashes with integer or "*" keys, to have problems distinguishing between arrays and hashes - although this shouldn't be an issue unless your data is very difficult to predict and/or your custom rules are very specific.

An order difference alone between two arrays can create too many diffs to be useful. Consider sorting them prior to diffing.

a = {a:'car', b:['boat', 'plane'] }
b = {a:'car', b:['plane', 'boat'] }

Hashdiff.diff(a, b).should == [["+", "b[0]", "plane"], ["-", "b[2]", "plane"]]

b[:b].sort!

Hashdiff.diff(a, b).should == []

• Krzysztof Rybka (@krzysiek1507)
• Fengyun Liu (@liufengyun)

Hashdiff is distributed under the MIT-LICENSE.