cobweb-launcher 1.2.11__py3-none-any.whl → 1.2.14__py3-none-any.whl

cobweb/crawlers/crawler.py

@@ -23,25 +23,33 @@ class Crawler(threading.Thread):
             self,
             stop: threading.Event,
             pause: threading.Event,
-            launcher_queue: Union[Mapping[str, Queue]],
+            # launcher_queue: Union[Mapping[str, Queue]],
+            get_seed: Callable,
+            set_seed: Callable,
+            add_seed: Callable,
+            delete_seed: Callable,
+            upload_data: Callable,
             custom_func: Union[Mapping[str, Callable]],
             thread_num: int,
-            max_retries: int
+            max_retries: int,
+            time_sleep: int,
     ):
         super().__init__()
 
         self._stop = stop
         self._pause = pause
-        self._new = launcher_queue["new"]
-        self._todo = launcher_queue["todo"]
-        self._done = launcher_queue["done"]
-        self._upload = launcher_queue["upload"]
+        self._get_seed = get_seed
+        self._set_seed = set_seed
+        self._add_seed = add_seed
+        self._delete_seed = delete_seed
+        self._upload_data = upload_data
 
         for func_name, _callable in custom_func.items():
             if isinstance(_callable, Callable):
                 self.__setattr__(func_name, _callable)
 
         self.thread_num = thread_num
+        self.time_sleep = time_sleep
         self.max_retries = max_retries
 
     @staticmethod
@@ -64,23 +72,23 @@ class Crawler(threading.Thread):
 
     def distribute(self, item, seed):
         if isinstance(item, BaseItem):
-            self._upload.push(item)
+            self._upload_data(item)
         elif isinstance(item, Seed):
-            self._new.push(item)
+            self._add_seed(item)
         elif isinstance(item, str) and item == DealModel.poll:
-            self._todo.push(seed)
+            self._set_seed(seed)
         elif isinstance(item, str) and item == DealModel.done:
-            self._done.push(seed)
+            self._delete_seed(seed)
         elif isinstance(item, str) and item == DealModel.fail:
             seed.params.seed_status = DealModel.fail
-            self._done.push(seed)
+            self._delete_seed(seed)
         else:
             raise TypeError("yield value type error!")
 
     def spider(self):
         while not self._stop.is_set():
 
-            seed = self._todo.pop()
+            seed = self._get_seed()
 
             if not seed:
                 time.sleep(1)
@@ -88,7 +96,7 @@ class Crawler(threading.Thread):
 
             elif seed.params.retry > self.max_retries:
                 seed.params.seed_status = DealModel.fail
-                self._done.push(seed)
+                self._delete_seed(seed)
                 continue
 
             seed_detail_log_info = LogTemplate.log_info(seed.to_dict)
@@ -149,7 +157,9 @@ class Crawler(threading.Thread):
                     exception=''.join(traceback.format_exception(type(e), e, e.__traceback__))
                 ))
                 seed.params.retry += 1
-                self._todo.push(seed)
+                # self._todo.push(seed)
+                self._set_seed(seed)
+                time.sleep(self.time_sleep * seed.params.retry)
             finally:
                 time.sleep(0.1)
         logger.info("spider thread close")

cobweb/db/redis_db.py

@@ -63,7 +63,7 @@ class RedisDB:
 
     def lock(self, key, t=15) -> bool:
         lua_script = """
-        local status = redis.call('setnx', KEYS[1], ARGV[1])
+        local status = redis.call('setnx', KEYS[1], 1)
         if ( status == 1 ) then
             redis.call('expire', KEYS[1], ARGV[1])
         end 

cobweb/launchers/launcher.py

@@ -17,6 +17,8 @@ def check_pause(func):
                 func(self, *args, **kwargs)
             except Exception as e:
                 logger.info(f"{func.__name__}: " + str(e))
+            finally:
+                time.sleep(0.1)
 
     return wrapper
 
@@ -94,10 +96,14 @@ class Launcher(threading.Thread):
 
         self._spider_max_retries = setting.SPIDER_MAX_RETRIES
         self._spider_thread_num = setting.SPIDER_THREAD_NUM
+        self._spider_time_sleep = setting.SPIDER_TIME_SLEEP
+        self._spider_max_speed = setting.SPIDER_MAX_SPEED
 
         self._done_model = setting.DONE_MODEL
         self._task_model = setting.TASK_MODEL
 
+        self._filter_field = setting.FILTER_FIELD
+
     @property
     def request(self):
         """
@@ -151,7 +157,22 @@ class Launcher(threading.Thread):
     def _remove_doing_seeds(self, seeds):
         for seed in seeds:
             self.__DOING__.pop(seed, None)
-        logger.info("remove %s seeds from __DOING__" % len(seeds))
+        # logger.info("remove %s seeds from __DOING__" % len(seeds))
+
+    def _get_seed(self) -> Seed:
+        return self.__LAUNCHER_QUEUE__["todo"].pop()
+
+    def _set_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["todo"].push(seed, **kwargs)
+
+    def _upload_data(self, data, **kwargs):
+        self.__LAUNCHER_QUEUE__["upload"].push(data, **kwargs)
+
+    def _add_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["new"].push(seed, **kwargs)
+
+    def _delete_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["done"].push(seed, **kwargs)
 
     def _execute(self):
         for func_name in self.__LAUNCHER_FUNC__:
@@ -165,10 +186,16 @@ class Launcher(threading.Thread):
 
         self._Crawler(
             stop=self._stop, pause=self._pause,
-            launcher_queue=self.__LAUNCHER_QUEUE__,
+            # launcher_queue=self.__LAUNCHER_QUEUE__,
+            get_seed=self._get_seed,
+            set_seed=self._set_seed,
+            add_seed=self._add_seed,
+            delete_seed=self._delete_seed,
+            upload_data=self._upload_data,
             custom_func=self.__CUSTOM_FUNC__,
             thread_num = self._spider_thread_num,
-            max_retries = self._spider_max_retries
+            max_retries = self._spider_max_retries,
+            time_sleep=self._spider_time_sleep
         ).start()
 
         self._Pipeline(

cobweb/launchers/launcher_pro.py

@@ -3,6 +3,7 @@ import threading
 
 from cobweb.db import RedisDB
 from cobweb.base import Seed, logger
+from cobweb.utils import BloomFilter
 from cobweb.constant import DealModel, LogTemplate
 from .launcher import Launcher, check_pause
 
@@ -15,11 +16,19 @@ class LauncherPro(Launcher):
         self._done_key = "{%s:%s}:done" % (project, task)
         self._fail_key = "{%s:%s}:fail" % (project, task)
         self._heartbeat_key = "heartbeat:%s_%s" % (project, task)
-        self._reset_lock_key = "lock:reset:%s_%s" % (project, task)
+
         self._statistics_done_key = "statistics:%s:%s:done" % (project, task)
         self._statistics_fail_key = "statistics:%s:%s:fail" % (project, task)
+        self._speed_control_key = "speed_control:%s_%s" % (project, task)
+
+        self._reset_lock_key = "lock:reset:%s_%s" % (project, task)
+
+        self._bf_key = "bloom_%s_%s" % (project, task)
+
         self._db = RedisDB()
 
+        self._bf = BloomFilter(self._bf_key)
+
         self._heartbeat_start_event = threading.Event()
         self._redis_queue_empty_event = threading.Event()
 
@@ -33,6 +42,21 @@ class LauncherPro(Launcher):
         else:
             self._db._client.incrby(key, count)
 
+    def _get_seed(self) -> Seed:
+        spider_speed = self._db._client.get(self._speed_control_key)
+        if int(spider_speed or 0) > self._spider_max_speed:
+            expire_time = self._db.ttl(self._speed_control_key)
+            if expire_time == -1:
+                self._db.delete(self._speed_control_key)
+            else:
+                logger.info(f"Too fast! Please wait {expire_time} seconds...")
+                time.sleep(expire_time / 2)
+            return None
+        seed = self.__LAUNCHER_QUEUE__["todo"].pop()
+        if seed and not self._db.lock(self._speed_control_key, t=60):
+            self._db._client.incrby(self._speed_control_key, 1)
+        return seed
+
     @check_pause
     def _execute_heartbeat(self):
         if self._heartbeat_start_event.is_set():
@@ -110,7 +134,7 @@ class LauncherPro(Launcher):
         """
         删除队列种子，根据状态添加至成功或失败队列，移除doing字典种子索引
         """
-        seeds, s_seeds, f_seeds = [], [], []
+        seed_info = {"count": 0, "failed": [], "succeed": [], "common": []}
         status = self.__LAUNCHER_QUEUE__['done'].length < self._done_queue_max_size
 
         for _ in range(self._done_queue_max_size):
@@ -118,26 +142,25 @@ class LauncherPro(Launcher):
             if not seed:
                 break
             if seed.params.seed_status == DealModel.fail:
-                f_seeds.append(seed.to_string)
+                seed_info["failed"].append(seed.to_string)
             elif self._done_model == 1:
-                s_seeds.append(seed.to_string)
+                seed_info["succeed"].append(seed.to_string)
             else:
-                seeds.append(seed.to_string)
-        if seeds:
-            count = self._db.zrem(self._todo_key, *seeds)
-            if count:
-                self.statistics(self._statistics_done_key, count)
-            self._remove_doing_seeds(seeds)
-        if s_seeds:
-            count = self._db.done([self._todo_key, self._done_key], *s_seeds)
-            if count:
-                self.statistics(self._statistics_done_key, count)
-            self._remove_doing_seeds(s_seeds)
-        if f_seeds:
-            count = self._db.done([self._todo_key, self._fail_key], *f_seeds)
-            if count:
-                self.statistics(self._statistics_fail_key, count)
-            self._remove_doing_seeds(f_seeds)
+                seed_info["common"].append(seed.to_string)
+            seed_info['count'] += 1
+
+        if seed_info["count"]:
+
+            succeed_count = self._db.zrem(self._todo_key, *seed_info["common"])
+            succeed_count += self._db.done([self._todo_key, self._done_key], *seed_info["succeed"])
+            failed_count = self._db.done([self._todo_key, self._fail_key], *seed_info["failed"])
+
+            if failed_count:
+                self.statistics(self._statistics_fail_key, failed_count)
+            if succeed_count:
+                self.statistics(self._statistics_done_key, succeed_count)
+
+            self._remove_doing_seeds(seed_info["common"] + seed_info["succeed"] + seed_info["failed"])
 
         if status:
             time.sleep(self._done_queue_wait_seconds)

cobweb/setting.py

@@ -57,9 +57,16 @@ DONE_MODEL = 0   # 0:种子消费成功直接从队列移除，失败则添加
 # spider
 SPIDER_THREAD_NUM = 10
 SPIDER_MAX_RETRIES = 5
+SPIDER_TIME_SLEEP = 10
+SPIDER_MAX_SPEED = 1000   # 一分钟最大采集数
 
 # 任务模式
 TASK_MODEL = 0  # 0:单次，1:常驻
 
+
+# bloom过滤器
+CAPACITY = 100000000
+ERROR_RATE = 0.001
+FILTER_FIELD = "url"
 # 文件下载响应类型过滤
 # FILE_FILTER_CONTENT_TYPE = ["text/html", "application/xhtml+xml"]

cobweb/utils/__init__.py

@@ -1,3 +1,4 @@
 from .oss import OssUtil
 from .tools import *
+from .bloom import BloomFilter
 

cobweb/utils/bloom.py

@@ -0,0 +1,76 @@
+import math
+import time
+
+import mmh3
+import redis
+from cobweb import setting
+
+
+class BloomFilter:
+
+    def __init__(self, key, redis_config=None, capacity=None, error_rate=None):
+        redis_config = redis_config or setting.REDIS_CONFIG
+        capacity = capacity or setting.CAPACITY
+        error_rate = error_rate or setting.ERROR_RATE
+        redis_config['db'] = 3
+
+        self.key = key
+
+        pool = redis.ConnectionPool(**redis_config)
+        self._client = redis.Redis(connection_pool=pool)
+        self.bit_size = self.get_bit_size(capacity, error_rate)
+        self.hash_count = self.get_hash_count(self.bit_size, capacity)
+        self._init_bloom_key()
+
+    def add(self, value):
+        for seed in range(self.hash_count):
+            result = mmh3.hash(value, seed) % self.bit_size
+            self._client.setbit(self.key, result, 1)
+        return True
+
+    def exists(self, value):
+        if not self._client.exists(self.key):
+            return False
+        for seed in range(self.hash_count):
+            result = mmh3.hash(value, seed) % self.bit_size
+            if not self._client.getbit(self.key, result):
+                return False
+        return True
+
+    def _init_bloom_key(self):
+        lua_script = """
+        redis.call("SETBIT", KEYS[1], ARGV[1], ARGV[2])
+        redis.call("EXPIRE", KEYS[1], 604800)
+        """
+        if self._client.exists(self.key):
+            return True
+        execute = self._client.register_script(lua_script)
+        execute(keys=[self.key], args=[self.bit_size-1, 1])
+
+    @classmethod
+    def get_bit_size(cls, n, p):
+        return int(-(n * math.log(p)) / (math.log(2) ** 2))
+
+    @classmethod
+    def get_hash_count(cls, m, n):
+        return int((m / n) * math.log(2))
+
+
+if __name__ == '__main__':
+    testBLF = BloomFilter("test", {
+        "host": "r-j6c1t3etiefpmz7cwdpd.redis.rds.aliyuncs.com",
+        "password": "SpiderLinux666",
+    })
+
+    print("start")
+    start_time = time.time()
+    testBLF.add("test")
+    add_time = time.time()
+    print("add time::: ")
+    print(add_time - start_time)
+    print("get::: ")
+    print(testBLF.exists("test"))
+    exist_time = time.time()
+    print("get time::: ")
+    print(exist_time - add_time)
+

{cobweb_launcher-1.2.11.dist-info → cobweb_launcher-1.2.14.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cobweb-launcher
-Version: 1.2.11
+Version: 1.2.14
 Summary: spider_hole
 Home-page: https://github.com/Juannie-PP/cobweb
 Author: Juannie-PP
@@ -16,6 +16,7 @@ Requires-Dist: requests (>=2.19.1)
 Requires-Dist: oss2 (>=2.18.1)
 Requires-Dist: redis (>=4.4.4)
 Requires-Dist: aliyun-log-python-sdk
+Requires-Dist: mmh3
 
 # cobweb 
 cobweb是一个基于python的分布式爬虫调度框架，目前支持分布式爬虫，单机爬虫，支持自定义数据库，支持自定义数据存储，支持自定义数据处理等操作。  

{cobweb_launcher-1.2.11.dist-info → cobweb_launcher-1.2.14.dist-info}/RECORD

@@ -1,6 +1,6 @@
 cobweb/__init__.py,sha256=uMHyf4Fekbyw2xBCbkA8R0LwCpBJf5p_7pWbh60ZWYk,83
 cobweb/constant.py,sha256=zy3XYsc1qp2B76_Fn_hVQ8eGHlPBd3OFlZK2cryE6FY,2839
-cobweb/setting.py,sha256=Wev0clo4ZETI5cRvBnzTWnJWo0Nowv_uvNCqlzYPSiE,1990
+cobweb/setting.py,sha256=MGe4QGnE5XOTh9Z7NhakaTFK7f-lZtzlA9PFcuc1qoY,2145
 cobweb/base/__init__.py,sha256=4gwWWQ0Q8cYG9cD7Lwf4XMqRGc5M_mapS3IczR6zeCE,222
 cobweb/base/common_queue.py,sha256=W7PPZZFl52j3Mc916T0imHj7oAUelA6aKJwW-FecDPE,872
 cobweb/base/decorators.py,sha256=wDCaQ94aAZGxks9Ljc0aXq6omDXT1_yzFy83ZW6VbVI,930
@@ -11,27 +11,28 @@ cobweb/base/response.py,sha256=eB1DWMXFCpn3cJ3yzgCRU1WeZAdayGDohRgdjdMUFN4,406
 cobweb/base/seed.py,sha256=Uz_VBRlAxNYQcFHk3tsZFMlU96yPOedHaWGTvk-zKd8,2908
 cobweb/crawlers/__init__.py,sha256=msvkB9mTpsgyj8JfNMsmwAcpy5kWk_2NrO1Adw2Hkw0,29
 cobweb/crawlers/base_crawler.py,sha256=ee_WSDnPQpPTk6wlFuY2UEx5L3hcsAZFcr6i3GLSry8,5751
-cobweb/crawlers/crawler.py,sha256=MJ2rFTrVjQ4Q9TmX0dpbOvT4tdokQwh7dKeFY_W7pPE,6002
+cobweb/crawlers/crawler.py,sha256=xiFNM0t69f5xlm59hPbO2MpqtdirVAUhD84-CLpyHPM,6349
 cobweb/crawlers/file_crawler.py,sha256=2Sjbdgxzqd41WykKUQE3QQlGai3T8k-pmHNmPlTchjQ,4454
 cobweb/db/__init__.py,sha256=ut0iEyBLjcJL06WNG_5_d4hO5PJWvDrKWMkDOdmgh2M,30
-cobweb/db/redis_db.py,sha256=NNI2QkRV1hEZI-z-COEncXt88z3pZN6wusKlcQzc8V4,4304
+cobweb/db/redis_db.py,sha256=fumNZJiio-uQqRcSrymx8eJ1PqsdOwITe_Y-9JOXxrQ,4298
 cobweb/exceptions/__init__.py,sha256=E9SHnJBbhD7fOgPFMswqyOf8SKRDrI_i25L0bSpohvk,32
 cobweb/exceptions/oss_db_exception.py,sha256=iP_AImjNHT3-Iv49zCFQ3rdLnlvuHa3h2BXApgrOYpA,636
 cobweb/launchers/__init__.py,sha256=af0Y6wrGX8SQZ7w7XL2sOtREjCT3dwad-uCc3nIontY,76
-cobweb/launchers/launcher.py,sha256=mBpq0CmxXXv-KdXQ2x7vwOkkAvaKECLkZLGhraafkQA,5953
+cobweb/launchers/launcher.py,sha256=AbkrytfJEyj8FhTbLgjmOOIuvOYV3cpVknE9yt31WbM,6930
 cobweb/launchers/launcher_air.py,sha256=KAk_M8F3029cXYe7m4nn3Nzyi89lbxJ2cqZjqW8iZ0E,2832
-cobweb/launchers/launcher_pro.py,sha256=OmlC5o3BdcsTFwOVAKNJdnwdlTl-fUiXC2kViWEpcoU,7677
+cobweb/launchers/launcher_pro.py,sha256=8QKhToKoD2WonIaqRQAhUWRhbNOIgYXzGFRK1id_3yM,8638
 cobweb/pipelines/__init__.py,sha256=zSUsGtx6smbs2iXBXvYynReKSgky-3gjqaAtKVnA_OU,105
 cobweb/pipelines/base_pipeline.py,sha256=fYnWf79GmhufXpcnMa3te18SbmnVeYLwxfyo-zLd9CY,1577
 cobweb/pipelines/loghub_pipeline.py,sha256=cjPO6w6UJ0jNw2fVvdX0BCdlm58T7dmYXlxzXOBpvfY,1027
 cobweb/pipelines/pipeline.py,sha256=4TJLX0sUHRxYndF5A4Vs5btUGI-wigkOcFvhTW1hLXI,2009
 cobweb/pipelines/pipeline_console.py,sha256=NEh-4zhuVAQOqwXLsqeb-rcNZ9_KXFUpL3otUTL5qBs,754
 cobweb/pipelines/pipeline_loghub.py,sha256=xZ6D55BGdiM71WUv83jyLGbEyUwhBHLJRZoXthBxxTs,1019
-cobweb/utils/__init__.py,sha256=JTE4sBfHnKHhD6w9Auk0MIT7O9BMOamCeryhlHNx3Zg,47
+cobweb/utils/__init__.py,sha256=vBtZTy3EfRE0MmH43URhmr7nw6_oOWTEbGOM9xR_9o8,78
+cobweb/utils/bloom.py,sha256=K3ACqQOIw9drxeVb6DKLA2iZYFn_rE16I781gXAsyqA,2228
 cobweb/utils/oss.py,sha256=gyt8-UB07tVphZLQXMOf-JTJwU-mWq8KZkOXKkAf3uk,3513
 cobweb/utils/tools.py,sha256=5JEaaAwYoV9Sdla2UBIJn6faUBuXmxUMagm9ck6FVqs,1253
-cobweb_launcher-1.2.11.dist-info/LICENSE,sha256=z1rxSIGOyzcSb3orZxFPxzx-0C1vTocmswqBNxpKfEk,1063
-cobweb_launcher-1.2.11.dist-info/METADATA,sha256=GSDZ3aWvattuW3Md-faK0cSBPyGDA8bJcVZRlSVp4ag,6490
-cobweb_launcher-1.2.11.dist-info/WHEEL,sha256=ewwEueio1C2XeHTvT17n8dZUJgOvyCWCt0WVNLClP9o,92
-cobweb_launcher-1.2.11.dist-info/top_level.txt,sha256=4GETBGNsKqiCUezmT-mJn7tjhcDlu7nLIV5gGgHBW4I,7
-cobweb_launcher-1.2.11.dist-info/RECORD,,
+cobweb_launcher-1.2.14.dist-info/LICENSE,sha256=z1rxSIGOyzcSb3orZxFPxzx-0C1vTocmswqBNxpKfEk,1063
+cobweb_launcher-1.2.14.dist-info/METADATA,sha256=ofgY9uUtCBfQn-LkEc0eZelQpJxWoB8Aza_u53qrytk,6510
+cobweb_launcher-1.2.14.dist-info/WHEEL,sha256=ewwEueio1C2XeHTvT17n8dZUJgOvyCWCt0WVNLClP9o,92
+cobweb_launcher-1.2.14.dist-info/top_level.txt,sha256=4GETBGNsKqiCUezmT-mJn7tjhcDlu7nLIV5gGgHBW4I,7
+cobweb_launcher-1.2.14.dist-info/RECORD,,