criticism this python code (crawler with threadpoo

问题:

how good this python code ? need criticism) there is a error in this code, some times script do print "ALL WAIT - CAN FINISH!" and freeze (no more actions are happend..) but i can't find reason why this happend?

site crawler with threadpool:

import sys
from urllib import urlopen
from BeautifulSoup import BeautifulSoup, SoupStrainer
import re
from Queue import Queue, Empty
from threading import Thread

W_WAIT = 1
W_WORK = 0

class Worker(Thread):
    """Thread executing tasks from a given tasks queue"""
    def __init__(self, pool, tasks):
        Thread.__init__(self)
        self.tasks = tasks
        self.daemon = True
        self.start()
        self.pool = pool
        self.state = None

    def is_wait(self):
        return self.state == W_WAIT


    def run(self):
        while True:
            #if all workers wait - time to exsit
            print "CHECK WAIT: !!! ",self.pool.is_all_wait()
            if self.pool.is_all_wait():
                print "ALL WAIT - CAN FINISH!"
                return
            try:
                func, args, kargs = self.tasks.get(timeout=3)
            except Empty:
                print "task wait timeout"
                continue

            self.state = W_WORK
            print "START !!! in thread %s" % str(self)
            #print args

            try: func(*args, **kargs)
            except Exception, e: print e
            print "!!! STOP in thread %s", str(self)
            self.tasks.task_done()
            self.state = W_WAIT
            #threads can fast empty it!
            #if self.tasks.qsize() == 0:
            #    print "QUIT!!!!!!"
            #    break

class ThreadPool:
    """Pool of threads consuming tasks from a queue"""
    def __init__(self, num_threads):
        #self.tasks = Queue(num_threads)
        self.tasks = Queue()
        self.workers = []
        for _ in range(num_threads): 
            self.workers.append(Worker(self,self.tasks))


    def add_task(self, func, *args, **kargs):
        """Add a task to the queue"""
        self.tasks.put((func, args, kargs))

    def wait_completion(self):
        """Wait for completion of all the tasks in the queue"""
        self.tasks.join()

    def is_all_wait(self):
        for w in self.workers:
            if not w.is_wait():
                return False
        return True

visited = set()
queue = Queue()
external_links_set = set()
internal_links_set = set()
external_links = 0

def process(pool,host,url):

    try:

        content = urlopen(url).read()
    except UnicodeDecodeError:
        return


    for link in BeautifulSoup(content, parseOnlyThese=SoupStrainer('a')):
        try:
            href = link['href']
        except KeyError:
            continue


        if not href.startswith('http://'):
            href = 'http://%s%s' % (host, href)
        if not href.startswith('http://%s%s' % (host, '/')):
            continue

        internal_links_set.add(href)


        if href not in visited:
            visited.add(href)
            pool.add_task(process,pool,host,href)

        else:
            pass

def start(host,charset):
    pool = ThreadPool(20)
    pool.add_task(process,pool,host,'http://%s/' % (host))
    pool.wait_completion()

start('evgenm.com','utf8') 

Thanx for help! i make new implementation: What you can say about this code#2 ? ==================================TRY #2=======================================

    import sys
    from urllib import urlopen
    from BeautifulSoup import BeautifulSoup, SoupStrainer
    import re
    from Queue import Queue, Empty
    from threading import Thread


    W_STOP = 1

class Worker(Thread):
    """Thread executing tasks from a given tasks queue"""
    def __init__(self, pool, tasks):
        Thread.__init__(self)
        self.tasks = tasks
        self.daemon = True
        self.pool = pool
        self.state = None
        self.start()



    def stop(self):
        self.state = W_STOP

    def run(self):
        while True:
            if self.state == W_STOP:
                print "\ncalled stop"
                break
            try:
                func, args, kargs = self.tasks.get(timeout=3)
            except Empty:
                continue
            print "\n***START*** %s" % str(self)
            try: 
                func(*args, **kargs)
            except Exception, e: 
                print e
            print "\n***STOP*** %s", str(self)
            self.tasks.task_done()



class ThreadPool:
    """Pool of threads consuming tasks from a queue"""
    def __init__(self, num_threads):
        #self.tasks = Queue(num_threads)
        self.tasks = Queue()
        self.workers = []
        for _ in range(num_threads): 
            self.workers.append(Worker(self,self.tasks))


    def add_task(self, func, *args, **kargs):
        """Add a task to the queue"""
        self.tasks.put((func, args, kargs))

    def wait_completion(self):
        """Wait for completion of all the tasks in the queue"""
        self.tasks.join()

    def stop_threads(self):
        for w in self.workers:
            w.stop()

    def wait_stop(self):
        self.wait_completion()
        self.stop_threads()



    visited = set()
    queue = Queue()
    external_links_set = set()
    internal_links_set = set()
    external_links = 0

    def process(pool,host,url):

        try:

            content = urlopen(url).read()
        except UnicodeDecodeError:
            return


        for link in BeautifulSoup(content, parseOnlyThese=SoupStrainer('a')):
            try:
                href = link['href']
            except KeyError:
                continue


            if not href.startswith('http://'):
                href = 'http://%s%s' % (host, href)
            if not href.startswith('http://%s%s' % (host, '/')):
                continue

            internal_links_set.add(href)


            if href not in visited:
                visited.add(href)
                pool.add_task(process,pool,host,href)

            else:
                pass

    def start(host,charset):
        pool = ThreadPool(20)
        pool.add_task(process,pool,host,'http://%s/' % (host))
        pool.wait_stop()

    start('evgenm.com','utf8') 

回答1:

You are sharing state between threads (i.e., in is_all_wait) without synchronization. Plus, the fact that all threads are "waiting" is not a reliable indicator that the queue is empty (for instance, they could all be in the process of getting a task). I suspect that, occasionally, threads are exiting before the queue is truly empty. If this happens often enough, you will be left with tasks in the queue but no threads to run them. So queue.join() will wait forever.

My recomendation is:

1. Get rid of is_all_wait -- it's not a reliable indicator
2. Get rid of the task state -- it's not really necessary
3. Rely on queue.join to let you know when everything is processed

If you need to kill the threads (for example, this is part of a larger, long-running program), then do so after the queue.join().

回答2:

I have basic python knowledge but threading in python isn't useless? I've seen tons of articles criticizing the global lock interpreter for this.