[ruby-talk:6217] Re: Ruby hi(gh), and pointer to Jotto program

David Alan Black <[e-mail]> writes:

> On Thu, 9 Nov 2000, Dave Thomas wrote:
> >      def common4(w1, w2)
> >        c = 0
> >        w1.each_with_index { |ch, i| c += 1 if w2[i] == ch }
> >        c
> >      end
> 
> But doesn't that just split (inclusively) on $/, rather than
> interating through the characters?  If only $/ could be a regular
> expression.... but then it would perhaps only do as well as:

Yup - I'm totally wrong on that one. Also, I just tried it with
String#each_byte and it was slower than the others.

It occurs to me that a speedup might be possible in the Jotto case
because one of the words is always fixed (for each side of the game)
so you could create a counter object and just pass it the word to be
checked:

     class Common5
       def initialize(word)
         @word = word
         @len = word.length
       end

       def count(word)
         i = 0
         count = 0
         @len.times {|i| count += 1 if word[i] == @word[i] }
         count
       end
     end
     c = Common5.new("hello dave")

     bm(10) {|x|
       x.report { 10000.times { common1("hello dave", "wombo dxve") } }
       x.report { 10000.times { common2("hello dave", "wombo dxve") } }
       x.report { 10000.times { common3("hello dave", "wombo dxve") } }
       x.report { 10000.times { common4("hello dave", "wombo dxve") } }
       x.report { 10000.times { c.count("wombo dxve") } }
     }

For me, that runs as fast as the fastest function-based test. I'm sure
there's a clever algorithm that could be implemented by
pre-calculating stuff in Common5, much as Boyer-Moore speeds up string
searching by pre-computing gaps. However, getting a clever algorithm
would require me to be either lucky or clever, and I'm neither today.


Regards


Dave