From: Florian Gross <flgr@...>
Date: 2005-03-15T22:57:38+09:00
Subject: Re: YAML obj merging

gga wrote:

> The problem is that I want hashes that contain other hashes to also
> follow the same rule.
> 
> That is:
> irb> a = { 'A' => { 'A1' => 'a' } }
> irb> b = { 'A' => { 'B1' => 'b' } }
> irb> b.merge(a)
> {"A"=>{"A1"=>"a"}}
> 
> When what I really want is this to be done recursively, so that:
> irb> a = { 'A' => { 'A1' => 'a' } }
> irb> b = { 'A' => { 'B1' => 'b' } }
> irb> b.mix(a) # ficticious method
> {"A"=>{"A1"=>"a", "B1" => 'b' }}
> 
> I guess I probably need to add my own method to do this.

I'm coming a bit late to this discussion (I've only spotted it after 
I've read through this weeks' Ruby Weekly News), but here's my version:

class Hash
   # Similar to Hash#merge, but this version works recursively so that
   # nested Hashs will be merged as well. You can specify a collision
   # block which will be invoked for conflicting items. (It takes the
   # same argument list as the collision block supplied to Hash#merge:
   # |hash, item1, item2|)
   #
   # Usage example:
   #   a = { "A" => { "A1" => "a" } }
   #   b = { "A" => { "B1" => "b" } }
   #   b.mix(a) # => {"A" => {"A1" => "a", "B1" => "b" }}
   def mix(other, &collision_block)
     merge_block = lambda do |hash, item1, item2|
       if item1.is_a?(Hash) and item2.is_a?(Hash) then
         item1.merge(item2, &merge_block)
       elsif collision_block then
         collision_block.call(hash, item1, item2)
       else
         item2
       end
     end

     self.merge(other, &merge_block)
   end
end

So basically the block that Hash#merge can take is insanely useful. I've 
also used for just detecting Hash collisions in the past as well.