Welcome, guest | Sign In | My Account | Store | Cart

Shows how you can yield out of nested function calls. The version here works with methods.

Python, 123 lines
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
import types
class method_pseudo_threads(object):
   """
   Implements pseudo threads for class methods.
   """
   _result = None
   
   def __init__(self, caller, instance, *args, **kw):
      self.caller = caller
      self.this_function = self.func(instance, *args, **kw)
      
   def next(self):
      return self.this_function.next()

   def call(self, function, *args, **kwds):
      """
      Check if function is a method of an instance of pseudo_threads.
      If so, call this method via the pseudo_thread mechanism
      """
      special =  hasattr(function, "_parallel")
      if special:
         return (None, function(self.this_function, *args, **kwds))
      else:
         return (function(*args, **kwds), self.this_function)

   def Return(self, value):
      "Return value to the caller"
      return (value, self.caller)
      
   def start(self, thread, *args, **kwds):
      "Start a new pseudo_thread thread, which runs in parallel to the current thread"
      return (None, [self.this_function, thread(None, *args, **kwds)])

   def run(self):
      """
      Calls next for all running threads.
      """
      queue = [self.next()]
      iterations = ticks = 0
      while queue:
         iterations += 1
         newqueue = []
         for result, continuation in queue:
            ticks += 1
            method_pseudo_threads._result = result
            result, continuations = continuation.next()
            if type(continuations) == types.ListType:
               for continuation in continuations:
                  newqueue.append((None, continuation))
            else:
               if continuations:
                     newqueue.append((result, continuations))
         queue = newqueue
      self.iterations, self.ticks = iterations, ticks
      return self._result

def parallel(function):
   """
   Decorator to turn a method into a pseudo_thread.

   The method itself should be written:
   
   def (th, self, *args, **kwds):
      
   Use th.call to call another method which is a pseudo thread.
   Use th.Return to return a value to the caller.
   Use th.start 

   self is the reference to the inclosing instance.
   """
   class p(method_pseudo_threads):
      name = "parallel_class_" + function.func_name
   p.func = function
   def glue(self, caller, *args, **kwds):
      thread = p(caller, self, *args, **kwds)
      yield (None, thread)
   glue._parallel = True
   glue.func_name = "parallel_" + function.func_name
   return glue

class start(method_pseudo_threads):
   def func(self, instance, *args, **kw):
      yield (None, instance)
      yield (None, None)

# Example:

class ackermann(object):
   @parallel
   def acker(th, self, m, n):
      call, Return = th.call, th.Return
      if m == 0:
         yield th.Return(n+1)
      elif m > 0 and n == 0:
         yield call(self.acker, m-1, 1)
         yield Return(th._result)
      elif m > 0 and n > 0:
         yield call(self.acker, m, n-1)
         yield call(self.acker, m-1, th._result)
         yield Return(th._result)
      else:
         assert 0

   @parallel
   def print_ackermann(th, self, m, n):
      yield th.call(self.acker, m, n)
      print "Ackerman(", m, ",", n, ")=", th._result
      yield th.Return(None)
      
   @parallel
   def start(th, self, i, j):
      for i1 in range(i):
         for j1 in range(j):
            yield th.start(self.print_ackermann, i1, j1)
      yield th.Return(None)

def version2():
   ack = ackermann()
   th = ack.start(None, 4, 5)
   start(None, th).run()

if __name__ == '__main__':
   version2()

The main idea here is to make the caller of a method available to the function, so that the function can explicitly return control to the caller once the call has finished.

The ackermann class shows how to use this concept. It computes some values of this highly recursive function in a parallel manner.

In Python 2.5, yield can return a value, so that we don't need to method_pseudo_threads._result anymore to pass function call results.

With an improved AST implementation it should also be easier to hide the transformation of methods into generators in a metaclass.

1 comment

Jean-Paul Calderone 18 years, 1 month ago  # | flag

This is quite similar to Twisted Flow. You might be interested in comparing the two.

http://twistedmatrix.com/projects/flow/documentation/howto/flow.html