From: Stefano Crocco <stefano.crocco@...>
Date: 2007-08-02T01:55:03+09:00
Subject: Re: count : array to array of hashes

Alle mercoled� 1 agosto 2007, Jean-S�bastien ha scritto:
> Hi,
>   I'm tryind to count distinct ids from an array (array_of_ids =
> [1,2,3,1,2]), and want a result array of hashes looking like it :
> result = [{;id=>1,:count=>2},{;id=>2,:count=>2},{;id=>3,:count=>1}]
>
> i tried it, but it always returns errors :
> result = array_of_ids.inject(Array.new){ |a,x| elt = a.find{|h| h[:id]
> == x} || {:id => x, :count => 0}; elt[:count] += 1; elt}
>
> i took my information from this post :
> http://groups.google.com/group/comp.lang.ruby/browse_thread/thread/3122212f
>6db8d7f5/04a7b10da8195cec?lnk=gst&q=array++count&rnum=8#04a7b10da8195cec
>
> If someone could help.
>
> regards

There are two problems in your code:
1- in the inject block, you return elt, which is (or should be, if the code 
worked) the hash containing the id which is being processed. You should 
return a, that is the array which contains the hashes. Correcting this should 
give a piece of code which executes without errors, but which returns an 
empty array
2- you never insert the hashes you create inside the inject block into the 
array a: you only store them in the local variable elt, which gets destroyed 
after each iteration. The inject block should be:

  result = array_of_ids.inject(Array.new){ |a,x| 
    elt = ( a.find{|h| h[:id] == x} || a[a.size] = {:id => x, :count => 0} )
    elt[:count] += 1
    a
  }

As you can see, after the || operator a new hash is created and inserted at 
the end of a (corresponding to the index a.size). Since an assignment always 
return the value being assigned (this is why I didn't use <<, it returns the 
array, not the inserted element), elt is then set to the new hash. Of course, 
all this happens only if find returns nil.

If you can rely in the id to be positive integers, and don't care if the 
resulting array contains the hashes in the same order as the id are stored in 
the array, here's another approach you can consider:

  result = array_of_ids.inject([]) do |res, i|
    res[i-1] ||= {:id => i, :count => 0}
    res[i-1][:count] += 1
    res
  end

This code stores the data relative to the id i in the i-1 position in the 
array (the -1 is there to avoid a nil element at the beginning). This should 
make it faster, since you don't need to iterate all the array to check 
whether the data corresponding to an id is already there or not: either is in 
the position id - 1 or it itsn't there. A quick benchmark, done by creating 
an array of ids of 100_000 elements, with values randomly chosen between 1 
and 11 gives:

                              user     system      total        real
original approach         5.730000   1.490000   7.220000 (  7.310224)
                              user     system      total        real
alternative approach      1.250000   0.150000   1.400000 (  1.416871)

Changing the range of the ids from 11 to 101 gives:
                              user     system      total        real
alternative approach      1.270000   0.190000   1.460000 (  1.472353)
                              user     system      total        real
original approach        37.730000  11.360000  49.090000 ( 51.056527)

Increasing it to 1001 gives
                              user     system      total        real
alternative approach      1.500000   0.220000   1.720000 (  1.733568)

The original approach takes much more time (I didn't have the patience to wait 
for it to complete).

I hope this helps

Stefano