From: Robert Klemme <bob.news@...>
Date: 2005-06-22T16:05:36+09:00
Subject: Re: Histogram-type Data


"Florian Frank" <flori@nixe.ping.de> schrieb im Newsbeitrag 
news:42B8D1A3.4040005@nixe.ping.de...
> Charles L. Snyder wrote:
>
>> I have a text-parsing question -
>>
>>array1 = ["Joe lives in France", "Sue goes to Japan", "Hiro walked around
>>France", "Bill moved to England"]
>>a_countries = ["France", "Japan", "England", "Spain"]
>>
> puts a_countries.inject({}) {|h,c| h[c] = array1.inject(0) { |s,a| s + 
> a.scan(c).size }; h }.sort_by { |_,v| -v }.map { |a| "%10s%10u" % a }

If I'm not mistaken your variant iterates array1 a_countries.size number of 
times.  For large arrays this might be quite inefficient.  This is what I'd 
do:

counts = a_countries.inject({}){|h,c| h[c]=0; h}
array1.each do |text|
  text.scan(/\w+/i) {|m| c = counts[m] and counts[m] = c+1}
end

You could even optimize away one hash lookups per found word by doing:

class Counter
  def initialize(n=0) @n=n end
  def inc() @n+=1 end
  def to_int() @n end
end

counts = a_countries.inject({}){|h,c| h[c]=Counter.new; h}
array1.each do |text|
  text.scan(/\w+/i) {|m| c = counts[m] and c.inc}
end

Kind regards

    robert