crawlo 1.2.0__py3-none-any.whl → 1.2.2__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 (220) hide show
  1. crawlo/__init__.py +61 -61
  2. crawlo/__version__.py +1 -1
  3. crawlo/cleaners/__init__.py +60 -60
  4. crawlo/cleaners/data_formatter.py +225 -225
  5. crawlo/cleaners/encoding_converter.py +125 -125
  6. crawlo/cleaners/text_cleaner.py +232 -232
  7. crawlo/cli.py +81 -65
  8. crawlo/commands/__init__.py +14 -14
  9. crawlo/commands/check.py +594 -594
  10. crawlo/commands/genspider.py +151 -151
  11. crawlo/commands/help.py +143 -133
  12. crawlo/commands/list.py +155 -155
  13. crawlo/commands/run.py +292 -292
  14. crawlo/commands/startproject.py +418 -418
  15. crawlo/commands/stats.py +188 -188
  16. crawlo/commands/utils.py +186 -186
  17. crawlo/config.py +312 -312
  18. crawlo/config_validator.py +252 -252
  19. crawlo/core/__init__.py +2 -2
  20. crawlo/core/engine.py +354 -354
  21. crawlo/core/processor.py +40 -40
  22. crawlo/core/scheduler.py +143 -143
  23. crawlo/crawler.py +1027 -1027
  24. crawlo/downloader/__init__.py +266 -266
  25. crawlo/downloader/aiohttp_downloader.py +220 -220
  26. crawlo/downloader/cffi_downloader.py +256 -256
  27. crawlo/downloader/httpx_downloader.py +259 -259
  28. crawlo/downloader/hybrid_downloader.py +213 -213
  29. crawlo/downloader/playwright_downloader.py +402 -402
  30. crawlo/downloader/selenium_downloader.py +472 -472
  31. crawlo/event.py +11 -11
  32. crawlo/exceptions.py +81 -81
  33. crawlo/extension/__init__.py +37 -37
  34. crawlo/extension/health_check.py +141 -141
  35. crawlo/extension/log_interval.py +57 -57
  36. crawlo/extension/log_stats.py +81 -81
  37. crawlo/extension/logging_extension.py +43 -43
  38. crawlo/extension/memory_monitor.py +104 -104
  39. crawlo/extension/performance_profiler.py +133 -133
  40. crawlo/extension/request_recorder.py +107 -107
  41. crawlo/filters/__init__.py +154 -154
  42. crawlo/filters/aioredis_filter.py +280 -280
  43. crawlo/filters/memory_filter.py +269 -269
  44. crawlo/items/__init__.py +23 -23
  45. crawlo/items/base.py +21 -21
  46. crawlo/items/fields.py +53 -53
  47. crawlo/items/items.py +104 -104
  48. crawlo/middleware/__init__.py +21 -21
  49. crawlo/middleware/default_header.py +132 -32
  50. crawlo/middleware/download_delay.py +105 -28
  51. crawlo/middleware/middleware_manager.py +135 -135
  52. crawlo/middleware/offsite.py +116 -0
  53. crawlo/middleware/proxy.py +366 -272
  54. crawlo/middleware/request_ignore.py +88 -30
  55. crawlo/middleware/response_code.py +164 -18
  56. crawlo/middleware/response_filter.py +138 -26
  57. crawlo/middleware/retry.py +124 -124
  58. crawlo/mode_manager.py +211 -211
  59. crawlo/network/__init__.py +21 -21
  60. crawlo/network/request.py +338 -338
  61. crawlo/network/response.py +359 -359
  62. crawlo/pipelines/__init__.py +21 -21
  63. crawlo/pipelines/bloom_dedup_pipeline.py +156 -156
  64. crawlo/pipelines/console_pipeline.py +39 -39
  65. crawlo/pipelines/csv_pipeline.py +316 -316
  66. crawlo/pipelines/database_dedup_pipeline.py +224 -224
  67. crawlo/pipelines/json_pipeline.py +218 -218
  68. crawlo/pipelines/memory_dedup_pipeline.py +115 -115
  69. crawlo/pipelines/mongo_pipeline.py +131 -131
  70. crawlo/pipelines/mysql_pipeline.py +316 -316
  71. crawlo/pipelines/pipeline_manager.py +61 -61
  72. crawlo/pipelines/redis_dedup_pipeline.py +167 -167
  73. crawlo/project.py +187 -187
  74. crawlo/queue/pqueue.py +37 -37
  75. crawlo/queue/queue_manager.py +337 -337
  76. crawlo/queue/redis_priority_queue.py +298 -298
  77. crawlo/settings/__init__.py +7 -7
  78. crawlo/settings/default_settings.py +226 -219
  79. crawlo/settings/setting_manager.py +122 -122
  80. crawlo/spider/__init__.py +639 -639
  81. crawlo/stats_collector.py +59 -59
  82. crawlo/subscriber.py +130 -130
  83. crawlo/task_manager.py +30 -30
  84. crawlo/templates/crawlo.cfg.tmpl +10 -10
  85. crawlo/templates/project/__init__.py.tmpl +3 -3
  86. crawlo/templates/project/items.py.tmpl +17 -17
  87. crawlo/templates/project/middlewares.py.tmpl +118 -109
  88. crawlo/templates/project/pipelines.py.tmpl +96 -96
  89. crawlo/templates/project/run.py.tmpl +45 -45
  90. crawlo/templates/project/settings.py.tmpl +327 -326
  91. crawlo/templates/project/settings_distributed.py.tmpl +119 -119
  92. crawlo/templates/project/settings_gentle.py.tmpl +94 -94
  93. crawlo/templates/project/settings_high_performance.py.tmpl +151 -151
  94. crawlo/templates/project/settings_simple.py.tmpl +68 -68
  95. crawlo/templates/project/spiders/__init__.py.tmpl +5 -5
  96. crawlo/templates/spider/spider.py.tmpl +143 -141
  97. crawlo/tools/__init__.py +182 -182
  98. crawlo/tools/anti_crawler.py +268 -268
  99. crawlo/tools/authenticated_proxy.py +240 -240
  100. crawlo/tools/data_validator.py +180 -180
  101. crawlo/tools/date_tools.py +35 -35
  102. crawlo/tools/distributed_coordinator.py +386 -386
  103. crawlo/tools/retry_mechanism.py +220 -220
  104. crawlo/tools/scenario_adapter.py +262 -262
  105. crawlo/utils/__init__.py +35 -35
  106. crawlo/utils/batch_processor.py +260 -260
  107. crawlo/utils/controlled_spider_mixin.py +439 -439
  108. crawlo/utils/date_tools.py +290 -290
  109. crawlo/utils/db_helper.py +343 -343
  110. crawlo/utils/enhanced_error_handler.py +359 -359
  111. crawlo/utils/env_config.py +105 -105
  112. crawlo/utils/error_handler.py +125 -125
  113. crawlo/utils/func_tools.py +82 -82
  114. crawlo/utils/large_scale_config.py +286 -286
  115. crawlo/utils/large_scale_helper.py +343 -343
  116. crawlo/utils/log.py +128 -128
  117. crawlo/utils/performance_monitor.py +284 -284
  118. crawlo/utils/queue_helper.py +175 -175
  119. crawlo/utils/redis_connection_pool.py +334 -334
  120. crawlo/utils/redis_key_validator.py +199 -199
  121. crawlo/utils/request.py +267 -267
  122. crawlo/utils/request_serializer.py +219 -219
  123. crawlo/utils/spider_loader.py +62 -62
  124. crawlo/utils/system.py +11 -11
  125. crawlo/utils/tools.py +4 -4
  126. crawlo/utils/url.py +39 -39
  127. {crawlo-1.2.0.dist-info → crawlo-1.2.2.dist-info}/METADATA +692 -697
  128. crawlo-1.2.2.dist-info/RECORD +220 -0
  129. examples/__init__.py +7 -7
  130. examples/aiohttp_settings.py +42 -0
  131. examples/curl_cffi_settings.py +41 -0
  132. examples/default_header_middleware_example.py +107 -0
  133. examples/default_header_spider_example.py +129 -0
  134. examples/download_delay_middleware_example.py +160 -0
  135. examples/httpx_settings.py +42 -0
  136. examples/multi_downloader_proxy_example.py +81 -0
  137. examples/offsite_middleware_example.py +55 -0
  138. examples/offsite_spider_example.py +107 -0
  139. examples/proxy_spider_example.py +166 -0
  140. examples/request_ignore_middleware_example.py +51 -0
  141. examples/request_ignore_spider_example.py +99 -0
  142. examples/response_code_middleware_example.py +52 -0
  143. examples/response_filter_middleware_example.py +67 -0
  144. examples/tong_hua_shun_settings.py +62 -0
  145. examples/tong_hua_shun_spider.py +170 -0
  146. tests/DOUBLE_CRAWLO_PREFIX_FIX_REPORT.md +81 -81
  147. tests/__init__.py +7 -7
  148. tests/advanced_tools_example.py +275 -275
  149. tests/authenticated_proxy_example.py +236 -236
  150. tests/cleaners_example.py +160 -160
  151. tests/config_validation_demo.py +102 -102
  152. tests/controlled_spider_example.py +205 -205
  153. tests/date_tools_example.py +180 -180
  154. tests/dynamic_loading_example.py +523 -523
  155. tests/dynamic_loading_test.py +104 -104
  156. tests/env_config_example.py +133 -133
  157. tests/error_handling_example.py +171 -171
  158. tests/redis_key_validation_demo.py +130 -130
  159. tests/response_improvements_example.py +144 -144
  160. tests/test_advanced_tools.py +148 -148
  161. tests/test_all_redis_key_configs.py +145 -145
  162. tests/test_authenticated_proxy.py +141 -141
  163. tests/test_cleaners.py +54 -54
  164. tests/test_comprehensive.py +146 -146
  165. tests/test_config_validator.py +193 -193
  166. tests/test_crawlo_proxy_integration.py +173 -0
  167. tests/test_date_tools.py +123 -123
  168. tests/test_default_header_middleware.py +159 -0
  169. tests/test_double_crawlo_fix.py +207 -207
  170. tests/test_double_crawlo_fix_simple.py +124 -124
  171. tests/test_download_delay_middleware.py +222 -0
  172. tests/test_downloader_proxy_compatibility.py +269 -0
  173. tests/test_dynamic_downloaders_proxy.py +124 -124
  174. tests/test_dynamic_proxy.py +92 -92
  175. tests/test_dynamic_proxy_config.py +146 -146
  176. tests/test_dynamic_proxy_real.py +109 -109
  177. tests/test_edge_cases.py +303 -303
  178. tests/test_enhanced_error_handler.py +270 -270
  179. tests/test_env_config.py +121 -121
  180. tests/test_error_handler_compatibility.py +112 -112
  181. tests/test_final_validation.py +153 -153
  182. tests/test_framework_env_usage.py +103 -103
  183. tests/test_integration.py +356 -356
  184. tests/test_item_dedup_redis_key.py +122 -122
  185. tests/test_offsite_middleware.py +222 -0
  186. tests/test_parsel.py +29 -29
  187. tests/test_performance.py +327 -327
  188. tests/test_proxy_api.py +265 -0
  189. tests/test_proxy_health_check.py +32 -32
  190. tests/test_proxy_middleware.py +122 -0
  191. tests/test_proxy_middleware_enhanced.py +217 -0
  192. tests/test_proxy_middleware_integration.py +136 -136
  193. tests/test_proxy_providers.py +56 -56
  194. tests/test_proxy_stats.py +19 -19
  195. tests/test_proxy_strategies.py +59 -59
  196. tests/test_queue_manager_double_crawlo.py +173 -173
  197. tests/test_queue_manager_redis_key.py +176 -176
  198. tests/test_real_scenario_proxy.py +196 -0
  199. tests/test_redis_config.py +28 -28
  200. tests/test_redis_connection_pool.py +294 -294
  201. tests/test_redis_key_naming.py +181 -181
  202. tests/test_redis_key_validator.py +123 -123
  203. tests/test_redis_queue.py +224 -224
  204. tests/test_request_ignore_middleware.py +183 -0
  205. tests/test_request_serialization.py +70 -70
  206. tests/test_response_code_middleware.py +350 -0
  207. tests/test_response_filter_middleware.py +428 -0
  208. tests/test_response_improvements.py +152 -152
  209. tests/test_retry_middleware.py +242 -0
  210. tests/test_scheduler.py +241 -241
  211. tests/test_simple_response.py +61 -61
  212. tests/test_telecom_spider_redis_key.py +205 -205
  213. tests/test_template_content.py +87 -87
  214. tests/test_template_redis_key.py +134 -134
  215. tests/test_tools.py +153 -153
  216. tests/tools_example.py +257 -257
  217. crawlo-1.2.0.dist-info/RECORD +0 -190
  218. {crawlo-1.2.0.dist-info → crawlo-1.2.2.dist-info}/WHEEL +0 -0
  219. {crawlo-1.2.0.dist-info → crawlo-1.2.2.dist-info}/entry_points.txt +0 -0
  220. {crawlo-1.2.0.dist-info → crawlo-1.2.2.dist-info}/top_level.txt +0 -0
crawlo/utils/log.py CHANGED
@@ -1,129 +1,129 @@
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
+ # -*- 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
129
129
  get_logger = LoggerManager.get_logger