crawlo 1.2.8__py3-none-any.whl → 1.3.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of crawlo might be problematic. Click here for more details.

Files changed (221) hide show
  1. crawlo/__init__.py +63 -61
  2. crawlo/__version__.py +1 -1
  3. crawlo/cli.py +75 -75
  4. crawlo/commands/__init__.py +14 -14
  5. crawlo/commands/check.py +594 -594
  6. crawlo/commands/genspider.py +151 -151
  7. crawlo/commands/help.py +138 -138
  8. crawlo/commands/list.py +155 -155
  9. crawlo/commands/run.py +314 -323
  10. crawlo/commands/startproject.py +436 -436
  11. crawlo/commands/stats.py +187 -187
  12. crawlo/commands/utils.py +186 -186
  13. crawlo/config.py +312 -312
  14. crawlo/config_validator.py +277 -251
  15. crawlo/core/__init__.py +2 -2
  16. crawlo/core/engine.py +365 -365
  17. crawlo/core/processor.py +40 -40
  18. crawlo/core/scheduler.py +256 -251
  19. crawlo/crawler.py +1097 -1099
  20. crawlo/data/__init__.py +5 -5
  21. crawlo/data/user_agents.py +194 -107
  22. crawlo/downloader/__init__.py +273 -266
  23. crawlo/downloader/aiohttp_downloader.py +226 -228
  24. crawlo/downloader/cffi_downloader.py +245 -256
  25. crawlo/downloader/httpx_downloader.py +259 -259
  26. crawlo/downloader/hybrid_downloader.py +212 -212
  27. crawlo/downloader/playwright_downloader.py +402 -402
  28. crawlo/downloader/selenium_downloader.py +472 -472
  29. crawlo/event.py +11 -11
  30. crawlo/exceptions.py +81 -81
  31. crawlo/extension/__init__.py +39 -39
  32. crawlo/extension/health_check.py +141 -141
  33. crawlo/extension/log_interval.py +57 -57
  34. crawlo/extension/log_stats.py +81 -81
  35. crawlo/extension/logging_extension.py +45 -43
  36. crawlo/extension/memory_monitor.py +104 -104
  37. crawlo/extension/performance_profiler.py +133 -133
  38. crawlo/extension/request_recorder.py +107 -107
  39. crawlo/filters/__init__.py +154 -154
  40. crawlo/filters/aioredis_filter.py +234 -234
  41. crawlo/filters/memory_filter.py +269 -269
  42. crawlo/items/__init__.py +23 -23
  43. crawlo/items/base.py +21 -21
  44. crawlo/items/fields.py +52 -52
  45. crawlo/items/items.py +104 -104
  46. crawlo/middleware/__init__.py +21 -21
  47. crawlo/middleware/default_header.py +132 -132
  48. crawlo/middleware/download_delay.py +104 -104
  49. crawlo/middleware/middleware_manager.py +136 -136
  50. crawlo/middleware/offsite.py +114 -114
  51. crawlo/middleware/proxy.py +386 -368
  52. crawlo/middleware/request_ignore.py +86 -86
  53. crawlo/middleware/response_code.py +163 -163
  54. crawlo/middleware/response_filter.py +136 -136
  55. crawlo/middleware/retry.py +124 -124
  56. crawlo/middleware/simple_proxy.py +65 -0
  57. crawlo/mode_manager.py +212 -211
  58. crawlo/network/__init__.py +21 -21
  59. crawlo/network/request.py +379 -338
  60. crawlo/network/response.py +359 -359
  61. crawlo/pipelines/__init__.py +21 -21
  62. crawlo/pipelines/bloom_dedup_pipeline.py +157 -157
  63. crawlo/pipelines/console_pipeline.py +39 -39
  64. crawlo/pipelines/csv_pipeline.py +316 -316
  65. crawlo/pipelines/database_dedup_pipeline.py +223 -223
  66. crawlo/pipelines/json_pipeline.py +218 -218
  67. crawlo/pipelines/memory_dedup_pipeline.py +115 -115
  68. crawlo/pipelines/mongo_pipeline.py +131 -131
  69. crawlo/pipelines/mysql_pipeline.py +317 -317
  70. crawlo/pipelines/pipeline_manager.py +74 -62
  71. crawlo/pipelines/redis_dedup_pipeline.py +167 -167
  72. crawlo/project.py +284 -315
  73. crawlo/queue/pqueue.py +37 -37
  74. crawlo/queue/queue_manager.py +379 -378
  75. crawlo/queue/redis_priority_queue.py +306 -306
  76. crawlo/settings/__init__.py +7 -7
  77. crawlo/settings/default_settings.py +216 -220
  78. crawlo/settings/setting_manager.py +175 -122
  79. crawlo/spider/__init__.py +639 -639
  80. crawlo/stats_collector.py +59 -59
  81. crawlo/subscriber.py +129 -129
  82. crawlo/task_manager.py +30 -30
  83. crawlo/templates/crawlo.cfg.tmpl +10 -10
  84. crawlo/templates/project/__init__.py.tmpl +3 -3
  85. crawlo/templates/project/items.py.tmpl +17 -17
  86. crawlo/templates/project/middlewares.py.tmpl +118 -118
  87. crawlo/templates/project/pipelines.py.tmpl +96 -96
  88. crawlo/templates/project/settings.py.tmpl +261 -288
  89. crawlo/templates/project/settings_distributed.py.tmpl +174 -157
  90. crawlo/templates/project/settings_gentle.py.tmpl +95 -100
  91. crawlo/templates/project/settings_high_performance.py.tmpl +125 -134
  92. crawlo/templates/project/settings_minimal.py.tmpl +30 -0
  93. crawlo/templates/project/settings_simple.py.tmpl +96 -98
  94. crawlo/templates/project/spiders/__init__.py.tmpl +5 -5
  95. crawlo/templates/run.py.tmpl +47 -47
  96. crawlo/templates/spider/spider.py.tmpl +143 -143
  97. crawlo/tools/__init__.py +200 -182
  98. crawlo/tools/anti_crawler.py +268 -268
  99. crawlo/tools/authenticated_proxy.py +240 -240
  100. crawlo/{cleaners → tools}/data_formatter.py +225 -225
  101. crawlo/tools/data_validator.py +180 -180
  102. crawlo/tools/date_tools.py +290 -36
  103. crawlo/tools/distributed_coordinator.py +388 -387
  104. crawlo/{cleaners → tools}/encoding_converter.py +127 -126
  105. crawlo/tools/request_tools.py +83 -0
  106. crawlo/tools/retry_mechanism.py +224 -221
  107. crawlo/tools/scenario_adapter.py +262 -262
  108. crawlo/{cleaners → tools}/text_cleaner.py +232 -232
  109. crawlo/utils/__init__.py +35 -35
  110. crawlo/utils/batch_processor.py +259 -259
  111. crawlo/utils/controlled_spider_mixin.py +439 -439
  112. crawlo/utils/db_helper.py +343 -343
  113. crawlo/utils/enhanced_error_handler.py +356 -356
  114. crawlo/utils/env_config.py +142 -142
  115. crawlo/utils/error_handler.py +123 -123
  116. crawlo/utils/func_tools.py +82 -82
  117. crawlo/utils/large_scale_config.py +286 -286
  118. crawlo/utils/large_scale_helper.py +344 -344
  119. crawlo/utils/log.py +146 -128
  120. crawlo/utils/performance_monitor.py +285 -285
  121. crawlo/utils/queue_helper.py +175 -175
  122. crawlo/utils/redis_connection_pool.py +351 -351
  123. crawlo/utils/redis_key_validator.py +198 -198
  124. crawlo/utils/request.py +267 -267
  125. crawlo/utils/request_serializer.py +218 -218
  126. crawlo/utils/spider_loader.py +61 -61
  127. crawlo/utils/system.py +11 -11
  128. crawlo/utils/tools.py +4 -4
  129. crawlo/utils/url.py +39 -39
  130. {crawlo-1.2.8.dist-info → crawlo-1.3.0.dist-info}/METADATA +1011 -764
  131. crawlo-1.3.0.dist-info/RECORD +219 -0
  132. examples/__init__.py +7 -7
  133. tests/DOUBLE_CRAWLO_PREFIX_FIX_REPORT.md +81 -81
  134. tests/__init__.py +7 -7
  135. tests/advanced_tools_example.py +275 -275
  136. tests/authenticated_proxy_example.py +107 -237
  137. tests/cleaners_example.py +160 -160
  138. tests/config_validation_demo.py +143 -103
  139. tests/controlled_spider_example.py +205 -205
  140. tests/date_tools_example.py +180 -180
  141. tests/debug_pipelines.py +67 -0
  142. tests/dynamic_loading_example.py +523 -523
  143. tests/dynamic_loading_test.py +104 -104
  144. tests/env_config_example.py +133 -133
  145. tests/error_handling_example.py +171 -171
  146. tests/redis_key_validation_demo.py +130 -130
  147. tests/request_params_example.py +151 -0
  148. tests/response_improvements_example.py +144 -144
  149. tests/test_advanced_tools.py +148 -148
  150. tests/test_all_redis_key_configs.py +145 -145
  151. tests/test_authenticated_proxy.py +141 -141
  152. tests/test_cleaners.py +54 -54
  153. tests/test_comprehensive.py +146 -146
  154. tests/test_config_consistency.py +80 -80
  155. tests/test_config_merge.py +153 -0
  156. tests/test_config_validator.py +182 -193
  157. tests/test_crawlo_proxy_integration.py +109 -173
  158. tests/test_date_tools.py +123 -123
  159. tests/test_default_header_middleware.py +158 -158
  160. tests/test_distributed.py +65 -0
  161. tests/test_double_crawlo_fix.py +207 -207
  162. tests/test_double_crawlo_fix_simple.py +124 -124
  163. tests/test_download_delay_middleware.py +221 -221
  164. tests/test_downloader_proxy_compatibility.py +268 -268
  165. tests/test_dynamic_downloaders_proxy.py +124 -124
  166. tests/test_dynamic_proxy.py +92 -92
  167. tests/test_dynamic_proxy_config.py +146 -146
  168. tests/test_dynamic_proxy_real.py +109 -109
  169. tests/test_edge_cases.py +303 -303
  170. tests/test_enhanced_error_handler.py +270 -270
  171. tests/test_env_config.py +121 -121
  172. tests/test_error_handler_compatibility.py +112 -112
  173. tests/test_final_validation.py +153 -153
  174. tests/test_framework_env_usage.py +103 -103
  175. tests/test_integration.py +169 -357
  176. tests/test_item_dedup_redis_key.py +122 -122
  177. tests/test_mode_consistency.py +51 -51
  178. tests/test_offsite_middleware.py +221 -221
  179. tests/test_parsel.py +29 -29
  180. tests/test_performance.py +327 -327
  181. tests/test_proxy_api.py +264 -264
  182. tests/test_proxy_health_check.py +32 -32
  183. tests/test_proxy_middleware.py +121 -121
  184. tests/test_proxy_middleware_enhanced.py +216 -216
  185. tests/test_proxy_middleware_integration.py +136 -136
  186. tests/test_proxy_middleware_refactored.py +185 -0
  187. tests/test_proxy_providers.py +56 -56
  188. tests/test_proxy_stats.py +19 -19
  189. tests/test_proxy_strategies.py +59 -59
  190. tests/test_queue_manager_double_crawlo.py +173 -173
  191. tests/test_queue_manager_redis_key.py +176 -176
  192. tests/test_random_user_agent.py +73 -0
  193. tests/test_real_scenario_proxy.py +195 -195
  194. tests/test_redis_config.py +28 -28
  195. tests/test_redis_connection_pool.py +294 -294
  196. tests/test_redis_key_naming.py +181 -181
  197. tests/test_redis_key_validator.py +123 -123
  198. tests/test_redis_queue.py +224 -224
  199. tests/test_request_ignore_middleware.py +182 -182
  200. tests/test_request_params.py +112 -0
  201. tests/test_request_serialization.py +70 -70
  202. tests/test_response_code_middleware.py +349 -349
  203. tests/test_response_filter_middleware.py +427 -427
  204. tests/test_response_improvements.py +152 -152
  205. tests/test_retry_middleware.py +241 -241
  206. tests/test_scheduler.py +252 -252
  207. tests/test_scheduler_config_update.py +133 -133
  208. tests/test_simple_response.py +61 -61
  209. tests/test_telecom_spider_redis_key.py +205 -205
  210. tests/test_template_content.py +87 -87
  211. tests/test_template_redis_key.py +134 -134
  212. tests/test_tools.py +159 -153
  213. tests/test_user_agents.py +97 -0
  214. tests/tools_example.py +260 -257
  215. tests/verify_distributed.py +117 -0
  216. crawlo/cleaners/__init__.py +0 -61
  217. crawlo/utils/date_tools.py +0 -290
  218. crawlo-1.2.8.dist-info/RECORD +0 -209
  219. {crawlo-1.2.8.dist-info → crawlo-1.3.0.dist-info}/WHEEL +0 -0
  220. {crawlo-1.2.8.dist-info → crawlo-1.3.0.dist-info}/entry_points.txt +0 -0
  221. {crawlo-1.2.8.dist-info → crawlo-1.3.0.dist-info}/top_level.txt +0 -0
crawlo/utils/log.py CHANGED
@@ -1,129 +1,147 @@
1
- # -*- coding: UTF-8 -*-
2
- """
3
- 日志管理器:安全版本,使用字符串化 key 避免 unhashable 问题
4
- """
5
- import os
6
- from logging import (
7
- Formatter,
8
- StreamHandler,
9
- FileHandler,
10
- Logger,
11
- DEBUG,
12
- INFO,
13
- WARNING,
14
- ERROR,
15
- CRITICAL,
16
- )
17
-
18
- LOG_FORMAT = '%(asctime)s - [%(name)s] - %(levelname)s: %(message)s'
19
-
20
-
21
- class LoggerManager:
22
- logger_cache = {}
23
- _default_filename = None
24
- _default_level = INFO
25
- _default_file_level = INFO
26
- _default_console_level = INFO
27
- _default_log_format = LOG_FORMAT
28
- _default_encoding = 'utf-8'
29
-
30
- _level_map = {
31
- 'DEBUG': DEBUG,
32
- 'INFO': INFO,
33
- 'WARNING': WARNING,
34
- 'ERROR': ERROR,
35
- 'CRITICAL': CRITICAL,
36
- }
37
-
38
- @classmethod
39
- def _to_level(cls, level):
40
- """安全转换为日志级别 int"""
41
- if level is None:
42
- return INFO
43
- if isinstance(level, int):
44
- return level
45
- if isinstance(level, str):
46
- return cls._level_map.get(level.upper(), INFO)
47
- if hasattr(level, 'get'): # 如 SettingManager 或 dict
48
- lv = level.get('LOG_LEVEL')
49
- if isinstance(lv, int):
50
- return lv
51
- if isinstance(lv, str):
52
- return cls._level_map.get(lv.upper(), INFO)
53
- return INFO
54
-
55
- @classmethod
56
- def configure(cls, settings=None, **kwargs):
57
- """
58
- 使用 settings 对象或关键字参数配置日志
59
- """
60
- # 优先使用 settings,否则用 kwargs
61
- get_val = settings.get if hasattr(settings, 'get') else (lambda k, d=None: kwargs.get(k, d))
62
-
63
- filename = get_val('LOG_FILE')
64
- level = get_val('LOG_LEVEL', 'INFO')
65
- file_level = get_val('LOG_FILE_LEVEL', level)
66
- console_level = get_val('LOG_CONSOLE_LEVEL', level)
67
- log_format = get_val('LOG_FORMAT', LOG_FORMAT)
68
- encoding = get_val('LOG_ENCODING', 'utf-8')
69
-
70
- cls._default_filename = filename
71
- cls._default_level = cls._to_level(level)
72
- cls._default_file_level = cls._to_level(file_level)
73
- cls._default_console_level = cls._to_level(console_level)
74
- cls._default_log_format = log_format
75
- cls._default_encoding = encoding
76
-
77
- @classmethod
78
- def get_logger(cls, name='default', level=None, filename=None):
79
- """
80
- 简化接口,只暴露必要参数
81
- """
82
- # 确定最终参数
83
- final_level = cls._to_level(level) if level is not None else cls._default_level
84
- final_filename = filename if filename is not None else cls._default_filename
85
-
86
- # 安全的字符串化 key,避免任何 unhashable 类型
87
- key_parts = [
88
- name,
89
- str(final_level),
90
- final_filename or 'no_file',
91
- ]
92
- key = '|'.join(key_parts) # 如 "my_spider|20|logs/app.log"
93
-
94
- if key in cls.logger_cache:
95
- return cls.logger_cache[key]
96
-
97
- # 创建 logger
98
- _logger = Logger(name=name)
99
- _logger.setLevel(final_level)
100
-
101
- formatter = Formatter(cls._default_log_format)
102
-
103
- # 控制台
104
- if cls._default_console_level is not False:
105
- ch = StreamHandler()
106
- ch.setFormatter(formatter)
107
- ch.setLevel(cls._default_console_level)
108
- _logger.addHandler(ch)
109
-
110
- # 文件
111
- if final_filename:
112
- try:
113
- log_dir = os.path.dirname(final_filename)
114
- if log_dir and not os.path.exists(log_dir):
115
- os.makedirs(log_dir, exist_ok=True)
116
-
117
- fh = FileHandler(final_filename, encoding=cls._default_encoding)
118
- fh.setFormatter(formatter)
119
- fh.setLevel(cls._default_file_level)
120
- _logger.addHandler(fh)
121
- except Exception as e:
122
- print(f"[Logger] 无法创建日志文件 {final_filename}: {e}")
123
-
124
- cls.logger_cache[key] = _logger
125
- return _logger
126
-
127
-
128
- # 全局快捷函数
1
+ import os
2
+ from logging import (
3
+ Formatter,
4
+ StreamHandler,
5
+ FileHandler,
6
+ Logger,
7
+ DEBUG,
8
+ INFO,
9
+ getLevelName,
10
+ )
11
+
12
+ LOG_FORMAT = '%(asctime)s - [%(name)s] - %(levelname)s: %(message)s'
13
+
14
+
15
+ class LoggerManager:
16
+ """日志管理器,提供统一的日志配置和获取接口"""
17
+ logger_cache = {}
18
+ _default_filename = None
19
+ _default_level = DEBUG # 设置为最低级别,由handler控制实际输出
20
+ _default_file_level = INFO # 默认为INFO级别
21
+ _default_console_level = INFO # 默认为INFO级别
22
+ _default_log_format = LOG_FORMAT
23
+ _default_encoding = 'utf-8'
24
+ _configured = False # 标记是否已配置
25
+
26
+ @classmethod
27
+ def _to_level(cls, level):
28
+ """安全转换为日志级别 int"""
29
+ if level is None:
30
+ return INFO
31
+ if isinstance(level, int):
32
+ return level
33
+ if isinstance(level, str):
34
+ # 使用logging模块内置的级别转换
35
+ level_value = getLevelName(level.upper())
36
+ # getLevelName在无效级别时返回字符串,我们需要返回数字
37
+ if isinstance(level_value, int):
38
+ return level_value
39
+ else:
40
+ return INFO
41
+ if hasattr(level, 'get'): # 如 SettingManager 或 dict
42
+ lv = level.get('LOG_LEVEL')
43
+ if isinstance(lv, int):
44
+ return lv
45
+ if isinstance(lv, str):
46
+ level_value = getLevelName(lv.upper())
47
+ if isinstance(level_value, int):
48
+ return level_value
49
+ else:
50
+ return INFO
51
+ return INFO
52
+
53
+ @classmethod
54
+ def configure(cls, settings=None, **kwargs):
55
+ """
56
+ 使用 settings 对象或关键字参数配置日志
57
+ """
58
+ # 优先使用 settings,否则用 kwargs
59
+ get_val = settings.get if hasattr(settings, 'get') else (lambda k, d=None: kwargs.get(k, d))
60
+
61
+ filename = get_val('LOG_FILE')
62
+ level = get_val('LOG_LEVEL', 'INFO') # 默认为INFO级别
63
+ file_level = get_val('LOG_FILE_LEVEL', level) # 默认继承LOG_LEVEL的值
64
+ # 根据项目规范,已完全移除LOG_CONSOLE_LEVEL支持,统一使用LOG_LEVEL控制控制台和文件的日志输出级别
65
+ console_level = level # 控制台日志级别直接使用LOG_LEVEL的值
66
+ log_format = get_val('LOG_FORMAT', LOG_FORMAT)
67
+ encoding = get_val('LOG_ENCODING', 'utf-8')
68
+
69
+ cls._default_filename = filename
70
+ cls._default_level = cls._to_level(level)
71
+ cls._default_file_level = cls._to_level(file_level)
72
+ cls._default_console_level = cls._to_level(console_level)
73
+ cls._default_log_format = log_format
74
+ cls._default_encoding = encoding
75
+
76
+ cls._configured = True
77
+
78
+ @classmethod
79
+ def get_logger(cls, name='default', level=None, filename=None):
80
+ """
81
+ 获取logger实例
82
+ """
83
+ # 确定最终参数
84
+ # 如果传入了level参数,则使用它,否则使用默认级别
85
+ if level is not None:
86
+ final_level = cls._to_level(level)
87
+ else:
88
+ # Logger级别设置为DEBUG(最低级别),由handler控制实际输出
89
+ final_level = DEBUG
90
+
91
+ final_filename = filename if filename is not None else cls._default_filename
92
+
93
+ # 安全的字符串化 key,避免任何 unhashable 类型
94
+ key_parts = [
95
+ name,
96
+ str(final_level),
97
+ final_filename or 'no_file',
98
+ ]
99
+ key = '|'.join(key_parts) # 如 "my_spider|20|logs/app.log"
100
+
101
+ if key in cls.logger_cache:
102
+ # 更新logger级别
103
+ cls.logger_cache[key].setLevel(final_level)
104
+ return cls.logger_cache[key]
105
+
106
+ # 创建 logger
107
+ _logger = Logger(name=name)
108
+ _logger.setLevel(final_level)
109
+
110
+ formatter = Formatter(cls._default_log_format)
111
+
112
+ # 控制台
113
+ if cls._default_console_level is not False:
114
+ ch = StreamHandler()
115
+ ch.setFormatter(formatter)
116
+ ch.setLevel(cls._default_console_level)
117
+ _logger.addHandler(ch)
118
+
119
+ # 文件
120
+ if final_filename:
121
+ try:
122
+ log_dir = os.path.dirname(final_filename)
123
+ if log_dir and not os.path.exists(log_dir):
124
+ os.makedirs(log_dir, exist_ok=True)
125
+
126
+ # 使用普通文件处理器(移除日志轮转功能)
127
+ fh = FileHandler(final_filename, mode='a', encoding=cls._default_encoding)
128
+
129
+ fh.setFormatter(formatter)
130
+ fh.setLevel(cls._default_file_level)
131
+ _logger.addHandler(fh)
132
+ except (PermissionError, FileNotFoundError) as e:
133
+ print(f"[Logger] 无法创建日志文件 {final_filename}: {e}")
134
+ except Exception as e:
135
+ print(f"[Logger] 创建日志文件时发生未知错误 {final_filename}: {e}")
136
+
137
+ cls.logger_cache[key] = _logger
138
+ return _logger
139
+
140
+ @classmethod
141
+ def is_configured(cls):
142
+ """检查日志系统是否已配置"""
143
+ return cls._configured
144
+
145
+
146
+ # 全局快捷函数
129
147
  get_logger = LoggerManager.get_logger