Question

我有一个生成器，它从有向无环图（DAG）中产生节点，深度优先：

def depth_first_search(self):
    yield self, 0 # root
    for child in self.get_child_nodes():
        for node, depth in child.depth_first_search():
            yield node, depth+1

我可以迭代这样的节点

for node, depth in graph.depth_first_search():
    # do something

如果满足某些条件，我希望能够从for循环告诉生成器停止在图中更深入。

我想出了以下使用外部函数的解决方案。

def depth_first_search(self, stop_crit=lambda n,d: False):
    yield self, 0 # root
    for child in self.get_child_nodes():
        for node, depth in child.depth_first_search():
            yield node, depth+1
            if stop_crit(node, depth): break

这个解决方案迫使我在定义stop_crit之前声明我需要的变量，以便可以从中访问它们。

在Ruby中，yield返回块中的最后一个表达式，因此可以方便地用它来告诉生成器继续或停止。

在Python中实现此功能的最佳方法是什么？

Answer 1

通常在Python中你只会停止使用生成器而忘记它。点。（因此以通常的方式将东西留给垃圾收集器）

然而，通过使用generator.close()，您可以立即强制立即启动发生器清理所有终结。

示例：

>>> def gen():
...     try: 
...         for i in range(10):
...             yield i
...     finally:
...         print "gen cleanup"
...         
>>> g = gen()
>>> next(g)
0
>>> for x in g:
...     print x
...     if x > 3:
...         g.close()
...         break
...        
1
2
3
4
gen cleanup
>>> g = gen()
>>> h = g
>>> next(g)
0
>>> del g
>>> del h   # last reference to generator code frame gets lost
gen cleanup

Answer 2

Coroutines（bassfriend提到他们）对于没有经验的人来说很棘手，所以这里有一个。我添加了一些测试代码，以便您可以看到它是如何工作的。

class Node(object):
    def __init__(self, val, left=None, right=None):
        self.val = val
        self.left = left
        self.right = right

    # the producing coroutine, it sends data to the consumer
    def depth_first_search(self, consumer, depth=0):
        """ `consumer` is a started coroutine that yields True to continue a branch """
        if consumer.send((self, depth)): # continue this branch?
            for child in self.get_child_nodes():
                child.depth_first_search(consumer, depth+1)

    def get_child_nodes(self):
        for node in (self.left, self.right):
            if node is not None:
                yield node

    def __repr__(self):
        return "Node(val=%d)" % self.val

def coroutine(func):
    """ decorator that declares `func` as a coroutine and starts it """
    def starter(*args, **kwargs):
        co = func(*args, **kwargs)
        next(co) # corotines need to be started/advanced to the first yield
        return co
    return starter

# the consumer takes data and yields if it wants to continue
@coroutine
def consumer( continue_branch=lambda n,d:True ):
    node, depth = (yield True) # first node
    while True:
        print node, depth # do stuff
        node, depth = (yield continue_branch(node, depth))


# testing
tree = Node(5, Node(2, Node(3), Node(4)), Node(6, Node(7), Node(8))) # 
cons = consumer()
tree.depth_first_search(cons)# yields all

print
stopper = consumer(lambda n,d: n.val > 2) # skips the children of Node 2
tree.depth_first_search(stopper)

诀窍在于，如果保留函数的角色，depth_first_search会产生节点，最终会出现可怕的混乱...而是生成节点并发送对消费者而言。

Python对协同程序的支持有点尴尬（@coroutine救援）。有一个非常nice tutorial for Python和大量资源用于依赖协程的语言，例如Lua。无论如何，这是一个值得探索的非常酷的概念： - ）

Answer 3

通常你不会告诉 iterable 检查条件，你在循环体中做到了这一点：

for node, depth in graph.depth_first_search():
    if node meets condition:
        # do something with node 
        break
# do something with node, its still referencing what you breaked on

此代码的优点是不会让任何人感到惊讶或混淆。

Answer 4

天真的解决方案：

def depth_first_search(self):
    yield self, 0 # root
    for child in self.get_child_nodes():
        for node, depth in child.depth_first_search():
            if(yield node, depth+1):
                yield None # for .send
                return

你可以正常调用它，但你必须将iterable保存为abort：

it = graph.depth_first_search()
for node, depth in it: #this is why there should be pronouns for loop iterables
    stuff(node,depth)
    if quit: it.send(1) 
    # it.next() should raise StopIteration on the next for iteration

我认为现在可行。

在Python中从块内停止生成器

4 个答案: