From: David Alan Black <dblack@...>
Date: 2001-04-11T06:28:31+09:00
Subject: [ruby-talk:13599] Re: #U017 "partition" method -- followup

Hi --

On Sat, 7 Apr 2001, Guy N. Hurst wrote:

> Array#split would make an array of subarrays based on a block:
>
> [1,2,3,4,5].split{|x| x<=>3}    # => [[3],[4,5],[1,2]]
>
> ary=["an","a","the","when","if","can","do"]
> ary.split{|x| x.length }  # => [[],["a"],["an","if","do"],["the","can"],["when"]]

Here, just for the sake of seeing it (in some form) in action, is an
implementation of something close (see further comments):

   class Array
     def split
       results = {}
       ary2 = []
       each do |e|
	 y = yield(e)
	 results[y] ||= ary2.size
	 ary2[results[y]] ||= []
	 ary2[results[y]].push(e)
       end
       ary2
     end
   end

   [1,2,3,4,5].split{|x| x<=>3}
      # => [[1, 2], [3], [4, 5]]

   ary=["an","a","the","when","if","can","do"]
   ary.split{|x| x.length}
      # => [["an", "if", "do"], ["a"], ["the", "can"], ["when"]]

   # (Alternate #each:
   #   each do |e|
   #     (ary2[results[yield(e)] ||= ary2.size] ||= []) .push(e)
   #   end
   # :-)

Obviously my version goes through the array and creates (or appends to)
the appropriate subarray based on what it finds, in order.  I sense
that this might be good, since otherwise the method would have
to decide (for instance) that lengths should be returned in ascending
order, while spaceships should be returned in x=,x<,x> order (taking
your hypothetical output literally).

> It could also be made to generate a hash, to allow for block output
> other than just numbers:
>
> ary.split{|x| x[0..0]}  # => {"a"=>["an","a"], "t"=>["the"], etc... }

Funny -- I've been dealing with something quite like that, where I'm
taking a bunch of file names which cluster according to their first
four characters and hashing and arraying them in various ways.

Anyway, wouldn't this have to be a different method?  How would #split
(or whatever it's called) know to construct a hash, rather than an
array -- and, come to think of it, would one really want it returning
different types depending on the input?


David

-- 
David Alan Black
home: dblack@candle.superlink.net
work: blackdav@shu.edu
Web:  http://pirate.shu.edu/~blackdav