cobweb-launcher 1.2.11__tar.gz → 1.2.13__tar.gz

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cobweb-launcher
-Version: 1.2.11
+Version: 1.2.13
 Summary: spider_hole
 Home-page: https://github.com/Juannie-PP/cobweb
 Author: Juannie-PP

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb/crawlers/crawler.py

@@ -26,7 +26,8 @@ class Crawler(threading.Thread):
             launcher_queue: Union[Mapping[str, Queue]],
             custom_func: Union[Mapping[str, Callable]],
             thread_num: int,
-            max_retries: int
+            max_retries: int,
+            time_sleep: int
     ):
         super().__init__()
 
@@ -43,6 +44,7 @@ class Crawler(threading.Thread):
 
         self.thread_num = thread_num
         self.max_retries = max_retries
+        self.time_sleep = time_sleep
 
     @staticmethod
     def request(seed: Seed) -> Union[Request, BaseItem]:
@@ -150,6 +152,7 @@ class Crawler(threading.Thread):
                 ))
                 seed.params.retry += 1
                 self._todo.push(seed)
+                time.sleep(self.time_sleep * seed.params.retry)
             finally:
                 time.sleep(0.1)
         logger.info("spider thread close")

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb/launchers/launcher.py

@@ -94,10 +94,13 @@ 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._done_model = setting.DONE_MODEL
         self._task_model = setting.TASK_MODEL
 
+        self._filter_field = setting.FILTER_FIELD
+
     @property
     def request(self):
         """
@@ -151,7 +154,7 @@ 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 _execute(self):
         for func_name in self.__LAUNCHER_FUNC__:
@@ -168,7 +171,8 @@ class Launcher(threading.Thread):
             launcher_queue=self.__LAUNCHER_QUEUE__,
             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-launcher-1.2.11 → cobweb-launcher-1.2.13}/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
 
@@ -18,8 +19,11 @@ class LauncherPro(Launcher):
         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._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()
 

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb/setting.py

@@ -57,9 +57,15 @@ 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
 
 # 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-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb/utils/__init__.py

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

cobweb-launcher-1.2.13/cobweb/utils/bloom.py

@@ -0,0 +1,98 @@
+import time
+
+
+from redis 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.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))
+
+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
+
+        self._client = Redis(**redis_config).bf()
+        self._client.create(key=self.key, capacity=capacity, errorRate=error_rate)
+
+    def add(self, value):
+        return self._client.add(self.key, value)
+
+    def madd(self, items: list):
+        return self._client.madd(self.key, *items)
+
+    def exists(self, value):
+        return self._client.exists(self.key, value)
+
+    def mexists(self, items: list):
+        return self._client.mexists(self.key, *items)
+
+
+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 → cobweb-launcher-1.2.13}/cobweb_launcher.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cobweb-launcher
-Version: 1.2.11
+Version: 1.2.13
 Summary: spider_hole
 Home-page: https://github.com/Juannie-PP/cobweb
 Author: Juannie-PP

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb_launcher.egg-info/SOURCES.txt

@@ -27,6 +27,7 @@ cobweb/pipelines/pipeline.py
 cobweb/pipelines/pipeline_console.py
 cobweb/pipelines/pipeline_loghub.py
 cobweb/utils/__init__.py
+cobweb/utils/bloom.py
 cobweb/utils/oss.py
 cobweb/utils/tools.py
 cobweb_launcher.egg-info/PKG-INFO

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/cobweb_launcher.egg-info/requires.txt

@@ -2,3 +2,4 @@ requests>=2.19.1
 oss2>=2.18.1
 redis>=4.4.4
 aliyun-log-python-sdk
+mmh3

{cobweb-launcher-1.2.11 → cobweb-launcher-1.2.13}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setup(
     name="cobweb-launcher",
-    version="1.2.11",
+    version="1.2.13",
     packages=find_packages(),
     url="https://github.com/Juannie-PP/cobweb",
     license="MIT",
@@ -14,7 +14,7 @@ setup(
     description="spider_hole",
     long_description=long_description,
     long_description_content_type="text/markdown",
-    install_requires=["requests>=2.19.1", "oss2>=2.18.1", "redis>=4.4.4", "aliyun-log-python-sdk"],
+    install_requires=["requests>=2.19.1", "oss2>=2.18.1", "redis>=4.4.4", "aliyun-log-python-sdk", "mmh3"],
     classifiers=[
         "Programming Language :: Python :: 3",
     ],