How to update a Ruby hash item inside of a loop over that hash?

I am migrating my PHP code to Ruby and at some point I need to update hash elements inside of a loop. For example:

compositions.each_pair do |element,params|
  params['composition'].each_pair do |item,data| 
    data['af'] /= params['af sum']
    data['mf'] /= params['mass']
  end
end

I could make it using item indexes, but it will be ugly. Is there a nice way to link the loop variables to the corresponding hash items? In PHP I would write &$params and &$data in the corresponding loops. Or may be a better suggestion?

Update

Two tests to demonstrate the problem. The first one:

a={'a'=>1, 'b'=>2, 'c'=>3}

a.each_pair do |k,v|
  v += 1
end
p a # =>  {"a"=>1, "b"=>2, "c"=>3}

and the second

a.each_pair do |k,v|
  a[k] += 1
end
p a # =>  {"a"=>2, "b"=>3, "c"=>4}

Update2

Thanks to Mladen 开发者_如何学Go(see below), I have understood the difference between these two cases. However, I still have a question: how to update data item (not just its own items)? Let's say we add

data = data['af'] + data['mf']

to the inner loop.

It seems that the code from the question works OK:

compositions = {1 => {'af sum' => 100.0, 'mass' => 200.0, 'composition' => {1 => {'af' => 5.0, 'mf' => 6.0}, 2  => {'af' => 7.0, 'mf' => 8.0}, 3  => {'af' => 9.0, 'mf' => 16.0}}}}
#=> {1=>{"af sum"=>100.0, "mass"=>200.0, "composition"=>{1=>{"af"=>5.0, "mf"=>6.0}, 2=>{"af"=>7.0, "mf"=>8.0}, 3=>{"af"=>9.0, "mf"=>16.0}}}}
compositions.each_pair do |element,params|
  params['composition'].each_pair do |item,data| 
    data['af'] /= params['af sum']
    data['mf'] /= params['mass']
  end
end
#=> {1=>{"af sum"=>100.0, "mass"=>200.0, "composition"=>{1=>{"af"=>0.05, "mf"=>0.03}, 2=>{"af"=>0.07, "mf"=>0.04}, 3=>{"af"=>0.09, "mf"=>0.08}}}}

But the example structure is pretty much guessed from the code, OP should post an example he is working with so we could actually make some progress.

Edit:

When you perform + method on an integer, a new object is returned as a result. That object is assigned to v in your first example, but not assigned back to the hash, so it is not preserved. In your second example, you assign it back to the hash.

However, if you work with mutating methods, you alter the very objects in-place, so then you don't need to reassign them to the hash. For example:

{1 => 'foo', 2 => 'bar'}.each{|k,v| v.swapcase!}
#=> {1=>"FOO", 2=>"BAR"}

but

{1 => 'foo', 2 => 'bar'}.each{|k,v| v = 'baz'}
#=> {1=>"foo", 2=>"bar"}

Edit 2:

I think you have problem understanding what k and v in the examples are. They are just local variables in the block, they start by referencing whatever is in the hash, but what do they reference can be changed during course of a block just like with any other variable. That change won't reflect to the hash. You need to alter the very object, not the reference to it, to actually change it.

Take a look at this, even simpler example, no hashes, blocks mumbo-jumbo:

a = 5
b = a
b += 1
a # => 5
b # => 6

It's as simple as that.