Ngrams with coroutines in Python

This is how I define ngrams with coroutines

	def coroutine(func):
	""" A decorator function that takes care
	of starting a coroutine automatically on call """
	def start(*args,**kwargs):
	coro = func(*args,**kwargs)
	coro.next()
	return coro
	return start
	@coroutine
	def ngrams(n, target):
	""" A coroutine to generate ngrams.
	Accepts one char at a time """
	chars = collections.deque()
	while True:
	chars.append((yield))
	if len(chars) == n:
	target.send(chars)
	chars.popleft()

view raw gistfile1.py hosted with ❤ by GitHub

I need to filter text before generating ngrams and also, I want to process ngrams (in this case count bigrams)

	@coroutine
	def filter_chars(accepted_chars,target):
	""" A coroutine to filter out unaccepted chars.
	Accepts one char at a time """
	while True:
	c = (yield)
	if c.lower() in accepted_chars:
	target.send(c.lower())
	@coroutine
	def counter(matrix):
	""" A counter sink """
	while True:
	a, b = (yield)
	matrix[pos[a]][pos[b]] += 1

view raw gistfile1.py hosted with ❤ by GitHub

I combine my coroutines together

	counts = [[10 for i in xrange(k)] for i in xrange(k)]
	bigrams = filter_chars(accepted_chars, ngrams(2, counter(counts)))
	for c in open('big.txt').read().decode(enc): bigrams.send(c)

view raw gistfile1.py hosted with ❤ by GitHub

Full source can be found in my fork of rrenaud's gibberish detector.