Sorting Hash of Hashes by value (and return the hash, not an array)

Question

I have the following hash:

user = {
  'user' => {
    'title' => {'weight' => 1, .... }
    'body' => {'weight' => 4, ....}
     ....
     ....
  }
}

Is is possible to get the User sorted by the weight key of its child hashes?

I looked in the Hash.sort, but it looks like it returns array rather than my original hash sorted.

Answer 1

In Ruby 1.9, Hashes are sorted, but Hash#sort still returns an Array of Arrays. Imagine that! It does imply that you can build your own sorting method on top of it.

class Hash
  def sorted_hash(&block)
    self.class[sort(&block)]   # Hash[ [[key1, value1], [key2, value2]] ]
  end
end

Hashes are unsorted in Ruby 1.8. If you want Ruby 1.8 compatibility, you can use ActiveSupport's OrderedHash. It behaves like a 1.9-Hash, so you can define the same sorted_hash method on it:

class ActiveSupport::OrderedHash
  def sorted_hash(&block)
    self.class[sort(&block)]
  end
end

hash = ActiveSupport::OrderedHash.new
hash["b"] = "b"
hash["a"] = "a"
hash               #=> {"b"=>"b", "a"=>"a"}  => unsorted
hash.sorted_hash   #=> {"a"=>"a", "b"=>"b"}  => sorted!

You have to copy the sorted_hash method to your code, because it does not exist by default!

Update for deep sorting: If you're looking to sort on something else than the hash key, pass a block to the sorted_hash method as follows (assuming the implementation from above):

hash = ActiveSupport::OrderedHash.new
hash["a"] = { "attr" => "2", "..." => "..." }
hash["b"] = { "attr" => "1", "..." => "..." }

# Unsorted.
hash 
  #=> {"a"=>{"attr"=>"2", "..."=>"..."}, "b"=>{"attr"=>"1", "..."=>"..."}}

# Sort on the "attr" key. (Assuming every value is a Hash itself!)
hash.sorted_hash { |a, b| a[1]["attr"] <=> b[1]["attr"] }
  #=> {"b"=>{"attr"=>"1", "..."=>"..."}, "a"=>{"attr"=>"2", "..."=>"..."}}

Answer 2

Hashes are fundamentally unsorted data structures; Hash#sort is, indeed, what you want. Either that, or sort a list of keys and then use that to enumerate when it's time to output the hash, instead of enumerating directly over the hash using its own methods.