for in benchmark interested

[Charles G Waldman]

But won't this break apart words that happen to span across a
500000-byte blocks boundary?

Jeremy Hylton writes:
 > The Python version would be faster if you used sys.stdin.read instead
 > of sys.stdin.readlines.  I'm not sure why you need to split the input
 > into lines before you split it into words; it seems like an
 > unnecessary step.

 >         while 1:
 > 		buf = read(500000)
 >                 if buf:
 > 			for key in string_split(buf):
 > 				dict[key] = dict_get(key, 0) + 1
 >                 else:
 >                         return dict