Re: the telegram problem: concurrency simplifies it (in Python)

Granted, decoupling producer and consumer can often be useful (what  
if, instead of enjoying a file abstraction, we were formatting words  
which could be broken across network packets or disk blocks?) but I'm  
not sure it applies in this case.  The 1960's logic seems relatively  
straightforward even without generators (although it does make use of  
"magic" behavior from print, presumably exactly for this  
application), and the functional approach still iterates from left-to- 
right, even though the word-appending operation is associative and  
hence in principle could be approached more concurrently.

-Dave

via "control break" programming:

:: :: ::

#!/usr/bin/python
"Telegram problem, aka fmt."
import sys

def telegram_control_break(infile, maxwidth):
     """A procedural version with minimal loop state

     This is what you get if you're writing a program to solve the
     problem about 50 years ago.

     """

     xpos = 0
     for line in infile:
	for word in line.split():
		if xpos + len(word) > maxwidth:
			print
			xpos = 0
		print word,
		xpos += len(word) + 1
     print

if __name__ == '__main__':
     telegram_control_break(sys.stdin, 72)

:: :: ::

or more functionally:

:: :: ::

#!/usr/bin/python
"Telegram problem, aka fmt."

bol	= lambda s	: ('\n'+s).rindex('\n')-1
fold	= lambda (s,m),w: (s+" \n"[len(s)+len(w)-bol(s)>m]+w, m)
tgram_	= lambda ws,m	: reduce(fold, ws[1:], (ws[0], m))[0]
tgram	= lambda f,m	: tgram_(f.read().split(),m)

if __name__ == '__main__':
     import sys
     print tgram(sys.stdin, 72)

:: :: ::

-- 
To unsubscribe: http://lists.canonical.org/mailman/listinfo/kragen-discuss