crawlo 1.3.5__py3-none-any.whl → 1.3.6__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 (289) hide show
  1. crawlo/__init__.py +87 -87
  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 +341 -341
  10. crawlo/commands/startproject.py +436 -436
  11. crawlo/commands/stats.py +187 -187
  12. crawlo/commands/utils.py +196 -196
  13. crawlo/config.py +312 -312
  14. crawlo/config_validator.py +277 -277
  15. crawlo/core/__init__.py +45 -45
  16. crawlo/core/engine.py +439 -439
  17. crawlo/core/processor.py +40 -40
  18. crawlo/core/scheduler.py +257 -257
  19. crawlo/crawler.py +638 -638
  20. crawlo/data/__init__.py +5 -5
  21. crawlo/data/user_agents.py +194 -194
  22. crawlo/downloader/__init__.py +273 -273
  23. crawlo/downloader/aiohttp_downloader.py +228 -228
  24. crawlo/downloader/cffi_downloader.py +245 -245
  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 +61 -61
  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/factories/__init__.py +27 -27
  40. crawlo/factories/base.py +68 -68
  41. crawlo/factories/crawler.py +103 -103
  42. crawlo/factories/registry.py +84 -84
  43. crawlo/filters/__init__.py +154 -154
  44. crawlo/filters/aioredis_filter.py +257 -257
  45. crawlo/filters/memory_filter.py +269 -269
  46. crawlo/framework.py +292 -291
  47. crawlo/initialization/__init__.py +39 -39
  48. crawlo/initialization/built_in.py +425 -425
  49. crawlo/initialization/context.py +141 -141
  50. crawlo/initialization/core.py +193 -193
  51. crawlo/initialization/phases.py +148 -148
  52. crawlo/initialization/registry.py +145 -145
  53. crawlo/items/__init__.py +23 -23
  54. crawlo/items/base.py +23 -23
  55. crawlo/items/fields.py +52 -52
  56. crawlo/items/items.py +104 -104
  57. crawlo/logging/__init__.py +37 -37
  58. crawlo/logging/config.py +96 -96
  59. crawlo/logging/factory.py +128 -128
  60. crawlo/logging/manager.py +111 -111
  61. crawlo/middleware/__init__.py +21 -21
  62. crawlo/middleware/default_header.py +132 -132
  63. crawlo/middleware/download_delay.py +104 -104
  64. crawlo/middleware/middleware_manager.py +135 -135
  65. crawlo/middleware/offsite.py +123 -123
  66. crawlo/middleware/proxy.py +386 -386
  67. crawlo/middleware/request_ignore.py +86 -86
  68. crawlo/middleware/response_code.py +163 -163
  69. crawlo/middleware/response_filter.py +136 -136
  70. crawlo/middleware/retry.py +124 -124
  71. crawlo/middleware/simple_proxy.py +65 -65
  72. crawlo/mode_manager.py +212 -212
  73. crawlo/network/__init__.py +21 -21
  74. crawlo/network/request.py +379 -379
  75. crawlo/network/response.py +359 -359
  76. crawlo/pipelines/__init__.py +21 -21
  77. crawlo/pipelines/bloom_dedup_pipeline.py +156 -156
  78. crawlo/pipelines/console_pipeline.py +39 -39
  79. crawlo/pipelines/csv_pipeline.py +316 -316
  80. crawlo/pipelines/database_dedup_pipeline.py +222 -222
  81. crawlo/pipelines/json_pipeline.py +218 -218
  82. crawlo/pipelines/memory_dedup_pipeline.py +115 -115
  83. crawlo/pipelines/mongo_pipeline.py +131 -131
  84. crawlo/pipelines/mysql_pipeline.py +325 -325
  85. crawlo/pipelines/pipeline_manager.py +76 -76
  86. crawlo/pipelines/redis_dedup_pipeline.py +166 -166
  87. crawlo/project.py +327 -327
  88. crawlo/queue/pqueue.py +42 -42
  89. crawlo/queue/queue_manager.py +503 -503
  90. crawlo/queue/redis_priority_queue.py +326 -326
  91. crawlo/settings/__init__.py +7 -7
  92. crawlo/settings/default_settings.py +321 -321
  93. crawlo/settings/setting_manager.py +214 -214
  94. crawlo/spider/__init__.py +657 -657
  95. crawlo/stats_collector.py +73 -73
  96. crawlo/subscriber.py +129 -129
  97. crawlo/task_manager.py +138 -138
  98. crawlo/templates/crawlo.cfg.tmpl +10 -10
  99. crawlo/templates/project/__init__.py.tmpl +3 -3
  100. crawlo/templates/project/items.py.tmpl +17 -17
  101. crawlo/templates/project/middlewares.py.tmpl +118 -118
  102. crawlo/templates/project/pipelines.py.tmpl +96 -96
  103. crawlo/templates/project/settings.py.tmpl +167 -167
  104. crawlo/templates/project/settings_distributed.py.tmpl +166 -166
  105. crawlo/templates/project/settings_gentle.py.tmpl +166 -166
  106. crawlo/templates/project/settings_high_performance.py.tmpl +167 -167
  107. crawlo/templates/project/settings_minimal.py.tmpl +65 -65
  108. crawlo/templates/project/settings_simple.py.tmpl +164 -164
  109. crawlo/templates/project/spiders/__init__.py.tmpl +9 -9
  110. crawlo/templates/run.py.tmpl +34 -34
  111. crawlo/templates/spider/spider.py.tmpl +143 -143
  112. crawlo/templates/spiders_init.py.tmpl +9 -9
  113. crawlo/tools/__init__.py +200 -200
  114. crawlo/tools/anti_crawler.py +268 -268
  115. crawlo/tools/authenticated_proxy.py +240 -240
  116. crawlo/tools/data_formatter.py +225 -225
  117. crawlo/tools/data_validator.py +180 -180
  118. crawlo/tools/date_tools.py +289 -289
  119. crawlo/tools/distributed_coordinator.py +388 -388
  120. crawlo/tools/encoding_converter.py +127 -127
  121. crawlo/tools/network_diagnostic.py +364 -364
  122. crawlo/tools/request_tools.py +82 -82
  123. crawlo/tools/retry_mechanism.py +224 -224
  124. crawlo/tools/scenario_adapter.py +262 -262
  125. crawlo/tools/text_cleaner.py +232 -232
  126. crawlo/utils/__init__.py +34 -34
  127. crawlo/utils/batch_processor.py +259 -259
  128. crawlo/utils/class_loader.py +25 -25
  129. crawlo/utils/controlled_spider_mixin.py +439 -439
  130. crawlo/utils/db_helper.py +343 -343
  131. crawlo/utils/enhanced_error_handler.py +356 -356
  132. crawlo/utils/env_config.py +142 -142
  133. crawlo/utils/error_handler.py +165 -165
  134. crawlo/utils/func_tools.py +82 -82
  135. crawlo/utils/large_scale_config.py +286 -286
  136. crawlo/utils/large_scale_helper.py +344 -344
  137. crawlo/utils/log.py +79 -79
  138. crawlo/utils/performance_monitor.py +285 -285
  139. crawlo/utils/queue_helper.py +175 -175
  140. crawlo/utils/redis_connection_pool.py +388 -388
  141. crawlo/utils/redis_key_validator.py +198 -198
  142. crawlo/utils/request.py +267 -267
  143. crawlo/utils/request_serializer.py +225 -225
  144. crawlo/utils/spider_loader.py +61 -61
  145. crawlo/utils/system.py +11 -11
  146. crawlo/utils/tools.py +4 -4
  147. crawlo/utils/url.py +39 -39
  148. {crawlo-1.3.5.dist-info → crawlo-1.3.6.dist-info}/METADATA +1126 -1126
  149. crawlo-1.3.6.dist-info/RECORD +290 -0
  150. examples/__init__.py +7 -7
  151. tests/__init__.py +7 -7
  152. tests/advanced_tools_example.py +275 -275
  153. tests/authenticated_proxy_example.py +106 -106
  154. tests/baidu_performance_test.py +108 -108
  155. tests/baidu_test.py +59 -59
  156. tests/cleaners_example.py +160 -160
  157. tests/comprehensive_framework_test.py +212 -212
  158. tests/comprehensive_test.py +81 -81
  159. tests/comprehensive_testing_summary.md +186 -186
  160. tests/config_validation_demo.py +142 -142
  161. tests/controlled_spider_example.py +205 -205
  162. tests/date_tools_example.py +180 -180
  163. tests/debug_configure.py +69 -69
  164. tests/debug_framework_logger.py +84 -84
  165. tests/debug_log_config.py +126 -126
  166. tests/debug_log_levels.py +63 -63
  167. tests/debug_pipelines.py +66 -66
  168. tests/detailed_log_test.py +233 -233
  169. tests/distributed_test.py +66 -66
  170. tests/distributed_test_debug.py +76 -76
  171. tests/dynamic_loading_example.py +523 -523
  172. tests/dynamic_loading_test.py +104 -104
  173. tests/env_config_example.py +133 -133
  174. tests/error_handling_example.py +171 -171
  175. tests/final_comprehensive_test.py +151 -151
  176. tests/final_log_test.py +260 -260
  177. tests/final_validation_test.py +182 -182
  178. tests/fix_log_test.py +142 -142
  179. tests/framework_performance_test.py +202 -202
  180. tests/log_buffering_test.py +111 -111
  181. tests/log_generation_timing_test.py +153 -153
  182. tests/optimized_performance_test.py +211 -211
  183. tests/performance_comparison.py +245 -245
  184. tests/queue_blocking_test.py +113 -113
  185. tests/queue_test.py +89 -89
  186. tests/redis_key_validation_demo.py +130 -130
  187. tests/request_params_example.py +150 -150
  188. tests/response_improvements_example.py +144 -144
  189. tests/scrapy_comparison/ofweek_scrapy.py +138 -138
  190. tests/scrapy_comparison/scrapy_test.py +133 -133
  191. tests/simple_command_test.py +119 -119
  192. tests/simple_crawlo_test.py +127 -127
  193. tests/simple_log_test.py +57 -57
  194. tests/simple_log_test2.py +137 -137
  195. tests/simple_optimization_test.py +128 -128
  196. tests/simple_queue_type_test.py +42 -0
  197. tests/simple_spider_test.py +49 -49
  198. tests/simple_test.py +47 -47
  199. tests/spider_log_timing_test.py +177 -177
  200. tests/test_advanced_tools.py +148 -148
  201. tests/test_all_commands.py +230 -230
  202. tests/test_all_redis_key_configs.py +145 -145
  203. tests/test_authenticated_proxy.py +141 -141
  204. tests/test_batch_processor.py +178 -178
  205. tests/test_cleaners.py +54 -54
  206. tests/test_component_factory.py +174 -174
  207. tests/test_comprehensive.py +146 -146
  208. tests/test_config_consistency.py +80 -80
  209. tests/test_config_merge.py +152 -152
  210. tests/test_config_validator.py +182 -182
  211. tests/test_controlled_spider_mixin.py +79 -79
  212. tests/test_crawlo_proxy_integration.py +108 -108
  213. tests/test_date_tools.py +123 -123
  214. tests/test_default_header_middleware.py +158 -158
  215. tests/test_distributed.py +65 -65
  216. tests/test_double_crawlo_fix.py +207 -207
  217. tests/test_double_crawlo_fix_simple.py +124 -124
  218. tests/test_download_delay_middleware.py +221 -221
  219. tests/test_downloader_proxy_compatibility.py +268 -268
  220. tests/test_dynamic_downloaders_proxy.py +124 -124
  221. tests/test_dynamic_proxy.py +92 -92
  222. tests/test_dynamic_proxy_config.py +146 -146
  223. tests/test_dynamic_proxy_real.py +109 -109
  224. tests/test_edge_cases.py +303 -303
  225. tests/test_enhanced_error_handler.py +270 -270
  226. tests/test_enhanced_error_handler_comprehensive.py +245 -245
  227. tests/test_env_config.py +121 -121
  228. tests/test_error_handler_compatibility.py +112 -112
  229. tests/test_factories.py +252 -252
  230. tests/test_final_validation.py +153 -153
  231. tests/test_framework_env_usage.py +103 -103
  232. tests/test_framework_logger.py +66 -66
  233. tests/test_framework_startup.py +64 -64
  234. tests/test_get_component_logger.py +83 -83
  235. tests/test_integration.py +169 -169
  236. tests/test_item_dedup_redis_key.py +122 -122
  237. tests/test_large_scale_config.py +112 -112
  238. tests/test_large_scale_helper.py +235 -235
  239. tests/test_logging_system.py +282 -282
  240. tests/test_mode_change.py +72 -72
  241. tests/test_mode_consistency.py +51 -51
  242. tests/test_offsite_middleware.py +221 -221
  243. tests/test_parsel.py +29 -29
  244. tests/test_performance.py +327 -327
  245. tests/test_performance_monitor.py +115 -115
  246. tests/test_proxy_api.py +264 -264
  247. tests/test_proxy_health_check.py +32 -32
  248. tests/test_proxy_middleware.py +121 -121
  249. tests/test_proxy_middleware_enhanced.py +216 -216
  250. tests/test_proxy_middleware_integration.py +136 -136
  251. tests/test_proxy_middleware_refactored.py +184 -184
  252. tests/test_proxy_providers.py +56 -56
  253. tests/test_proxy_stats.py +19 -19
  254. tests/test_proxy_strategies.py +59 -59
  255. tests/test_queue_empty_check.py +41 -41
  256. tests/test_queue_manager_double_crawlo.py +173 -173
  257. tests/test_queue_manager_redis_key.py +176 -176
  258. tests/test_queue_type.py +107 -0
  259. tests/test_random_user_agent.py +72 -72
  260. tests/test_real_scenario_proxy.py +195 -195
  261. tests/test_redis_config.py +28 -28
  262. tests/test_redis_connection_pool.py +294 -294
  263. tests/test_redis_key_naming.py +181 -181
  264. tests/test_redis_key_validator.py +123 -123
  265. tests/test_redis_queue.py +224 -224
  266. tests/test_request_ignore_middleware.py +182 -182
  267. tests/test_request_params.py +111 -111
  268. tests/test_request_serialization.py +70 -70
  269. tests/test_response_code_middleware.py +349 -349
  270. tests/test_response_filter_middleware.py +427 -427
  271. tests/test_response_improvements.py +152 -152
  272. tests/test_retry_middleware.py +241 -241
  273. tests/test_scheduler.py +252 -252
  274. tests/test_scheduler_config_update.py +133 -133
  275. tests/test_simple_response.py +61 -61
  276. tests/test_telecom_spider_redis_key.py +205 -205
  277. tests/test_template_content.py +87 -87
  278. tests/test_template_redis_key.py +134 -134
  279. tests/test_tools.py +159 -159
  280. tests/test_user_agents.py +96 -96
  281. tests/tools_example.py +260 -260
  282. tests/untested_features_report.md +138 -138
  283. tests/verify_debug.py +51 -51
  284. tests/verify_distributed.py +117 -117
  285. tests/verify_log_fix.py +111 -111
  286. crawlo-1.3.5.dist-info/RECORD +0 -288
  287. {crawlo-1.3.5.dist-info → crawlo-1.3.6.dist-info}/WHEEL +0 -0
  288. {crawlo-1.3.5.dist-info → crawlo-1.3.6.dist-info}/entry_points.txt +0 -0
  289. {crawlo-1.3.5.dist-info → crawlo-1.3.6.dist-info}/top_level.txt +0 -0
@@ -1,426 +1,426 @@
1
- #!/usr/bin/python
2
- # -*- coding: UTF-8 -*-
3
- """
4
- 内置初始化器 - 提供框架核心组件的初始化实现
5
- """
6
-
7
- import time
8
- from .registry import BaseInitializer, register_initializer
9
- from .phases import InitializationPhase, PhaseResult
10
- from .context import InitializationContext
11
-
12
-
13
- class LoggingInitializer(BaseInitializer):
14
- """日志系统初始化器"""
15
-
16
- def __init__(self):
17
- super().__init__(InitializationPhase.LOGGING)
18
-
19
- def initialize(self, context: InitializationContext) -> PhaseResult:
20
- """初始化日志系统"""
21
- start_time = time.time()
22
-
23
- try:
24
- # 导入日志模块
25
- from crawlo.logging import configure_logging, LogConfig
26
-
27
- # 获取日志配置
28
- log_config = self._get_log_config(context)
29
-
30
- # 确保日志目录存在
31
- if log_config.file_path and log_config.file_enabled:
32
- import os
33
- log_dir = os.path.dirname(log_config.file_path)
34
- if log_dir and not os.path.exists(log_dir):
35
- os.makedirs(log_dir, exist_ok=True)
36
-
37
- # 配置日志系统
38
- configure_logging(log_config)
39
-
40
- # 存储到共享数据
41
- context.add_shared_data('log_config', log_config)
42
-
43
- # 创建框架logger
44
- from crawlo.logging import get_logger
45
- framework_logger = get_logger('crawlo.framework')
46
- context.add_shared_data('framework_logger', framework_logger)
47
-
48
- return self._create_result(
49
- success=True,
50
- duration=time.time() - start_time,
51
- artifacts={'log_config': log_config}
52
- )
53
-
54
- except Exception as e:
55
- return self._create_result(
56
- success=False,
57
- duration=time.time() - start_time,
58
- error=e
59
- )
60
-
61
- def _get_log_config(self, context: InitializationContext) -> 'LogConfig':
62
- """
63
- 获取日志配置
64
-
65
- Args:
66
- context: 初始化上下文
67
-
68
- Returns:
69
- LogConfig: 日志配置对象
70
- """
71
- # 导入日志配置类
72
- from crawlo.logging import LogConfig
73
-
74
- # 按优先级获取配置:自定义配置 > 上下文配置 > 项目配置 > 默认配置
75
- config_sources = [
76
- context.custom_settings,
77
- context.settings,
78
- self._load_project_config()
79
- ]
80
-
81
- # 遍历配置源
82
- for config_source in config_sources:
83
- if config_source:
84
- log_config = self._create_log_config_from_source(config_source)
85
- if log_config:
86
- return log_config
87
-
88
- # 使用默认配置
89
- return LogConfig()
90
-
91
- def _create_log_config_from_source(self, config_source) -> 'LogConfig':
92
- """
93
- 从配置源创建日志配置
94
-
95
- Args:
96
- config_source: 配置源
97
-
98
- Returns:
99
- LogConfig: 日志配置对象,如果配置源无效则返回None
100
- """
101
- # 导入日志配置类
102
- from crawlo.logging import LogConfig
103
-
104
- # 检查配置源是否有效
105
- if not config_source:
106
- return None
107
-
108
- # 检查是否有日志相关配置
109
- has_keys_method = hasattr(config_source, 'keys')
110
- if has_keys_method:
111
- has_log_config = any(key.startswith('LOG_') for key in config_source.keys())
112
- else:
113
- has_log_config = any(key.startswith('LOG_') for key in dir(config_source))
114
-
115
- if not has_log_config:
116
- return None
117
-
118
- # 从配置源获取日志配置
119
- log_level = config_source.get('LOG_LEVEL', 'INFO')
120
- log_file = config_source.get('LOG_FILE')
121
- log_format = config_source.get('LOG_FORMAT', '%(asctime)s - [%(name)s] - %(levelname)s: %(message)s')
122
- log_encoding = config_source.get('LOG_ENCODING', 'utf-8')
123
- log_max_bytes = config_source.get('LOG_MAX_BYTES', 10 * 1024 * 1024)
124
- log_backup_count = config_source.get('LOG_BACKUP_COUNT', 5)
125
- log_console_enabled = config_source.get('LOG_CONSOLE_ENABLED', True)
126
- log_file_enabled = config_source.get('LOG_FILE_ENABLED', True)
127
-
128
- # 创建日志配置
129
- return LogConfig(
130
- level=log_level,
131
- format=log_format,
132
- encoding=log_encoding,
133
- file_path=log_file,
134
- max_bytes=log_max_bytes,
135
- backup_count=log_backup_count,
136
- console_enabled=log_console_enabled,
137
- file_enabled=log_file_enabled
138
- )
139
-
140
- def _load_project_config(self):
141
- """
142
- 自动加载项目配置以获取日志设置
143
- """
144
- try:
145
- # 查找项目根目录
146
- import os
147
- import sys
148
- import configparser
149
-
150
- current_path = os.getcwd()
151
-
152
- # 向上查找直到找到crawlo.cfg
153
- checked_paths = set()
154
- path = current_path
155
-
156
- while path not in checked_paths:
157
- checked_paths.add(path)
158
-
159
- # 检查crawlo.cfg
160
- cfg_file = os.path.join(path, "crawlo.cfg")
161
- if os.path.exists(cfg_file):
162
- # 读取配置文件
163
- config_parser = configparser.ConfigParser()
164
- config_parser.read(cfg_file, encoding="utf-8")
165
-
166
- if config_parser.has_section("settings") and config_parser.has_option("settings", "default"):
167
- # 获取settings模块路径
168
- settings_module_path = config_parser.get("settings", "default")
169
-
170
- # 添加项目根目录到Python路径
171
- if path not in sys.path:
172
- sys.path.insert(0, path)
173
-
174
- # 导入项目配置模块
175
- import importlib
176
- settings_module = importlib.import_module(settings_module_path)
177
-
178
- # 创建配置字典
179
- project_config = {}
180
- for key in dir(settings_module):
181
- if key.isupper():
182
- project_config[key] = getattr(settings_module, key)
183
-
184
- return project_config
185
-
186
- # 向上一级目录
187
- parent = os.path.dirname(path)
188
- if parent == path:
189
- break
190
- path = parent
191
-
192
- return {}
193
-
194
- except Exception as e:
195
- return {}
196
-
197
-
198
- class SettingsInitializer(BaseInitializer):
199
- """配置系统初始化器"""
200
-
201
- def __init__(self):
202
- super().__init__(InitializationPhase.SETTINGS)
203
-
204
- def initialize(self, context: InitializationContext) -> PhaseResult:
205
- """初始化配置系统"""
206
- start_time = time.time()
207
-
208
- try:
209
- # 导入配置管理器
210
- from crawlo.settings.setting_manager import SettingManager
211
- from crawlo.project import _load_project_settings
212
-
213
- # 创建配置管理器并加载项目配置
214
- settings = _load_project_settings(context.custom_settings)
215
-
216
- # 存储到上下文
217
- context.settings = settings
218
- context.add_shared_data('settings', settings)
219
-
220
- return self._create_result(
221
- success=True,
222
- duration=time.time() - start_time,
223
- artifacts={'settings': settings}
224
- )
225
-
226
- except Exception as e:
227
- return self._create_result(
228
- success=False,
229
- duration=time.time() - start_time,
230
- error=e
231
- )
232
-
233
-
234
- class CoreComponentsInitializer(BaseInitializer):
235
- """核心组件初始化器"""
236
-
237
- def __init__(self):
238
- super().__init__(InitializationPhase.CORE_COMPONENTS)
239
-
240
- def initialize(self, context: InitializationContext) -> PhaseResult:
241
- """初始化核心组件"""
242
- start_time = time.time()
243
-
244
- try:
245
- # 在核心组件初始化阶段,大多数组件需要crawler参数
246
- # 因此我们只初始化那些完全独立的组件
247
- # 或者只记录需要初始化的组件类型,实际初始化在crawler创建时进行
248
-
249
- return self._create_result(
250
- success=True,
251
- duration=time.time() - start_time,
252
- artifacts={}
253
- )
254
-
255
- except Exception as e:
256
- return self._create_result(
257
- success=False,
258
- duration=time.time() - start_time,
259
- error=e
260
- )
261
-
262
- def _initialize_engine(self, context: InitializationContext):
263
- """初始化引擎"""
264
- try:
265
- # 注意:Engine需要crawler参数,不能在此阶段初始化
266
- pass
267
- except Exception as e:
268
- context.add_error(f"Failed to initialize engine: {e}")
269
- raise
270
-
271
- def _initialize_scheduler(self, context: InitializationContext):
272
- """初始化调度器"""
273
- try:
274
- # 注意:Scheduler需要很多参数,不能在此阶段初始化
275
- pass
276
- except Exception as e:
277
- context.add_error(f"Failed to initialize scheduler: {e}")
278
- raise
279
-
280
- def _initialize_downloader(self, context: InitializationContext):
281
- """初始化下载器"""
282
- try:
283
- # 注意:下载器类需要crawler参数,不能在此阶段初始化实例
284
- pass
285
- except Exception as e:
286
- context.add_error(f"Failed to initialize downloader: {e}")
287
- raise
288
-
289
- def _initialize_pipeline_manager(self, context: InitializationContext):
290
- """初始化管道管理器"""
291
- try:
292
- # 注意:PipelineManager需要crawler参数,不能在此阶段初始化
293
- pass
294
- except Exception as e:
295
- context.add_error(f"Failed to initialize pipeline manager: {e}")
296
- raise
297
-
298
- def _initialize_middleware_manager(self, context: InitializationContext):
299
- """初始化中间件管理器"""
300
- try:
301
- # 注意:MiddlewareManager需要crawler参数,不能在此阶段初始化
302
- pass
303
- except Exception as e:
304
- context.add_error(f"Failed to initialize middleware manager: {e}")
305
- raise
306
-
307
-
308
- class ExtensionsInitializer(BaseInitializer):
309
- """扩展组件初始化器"""
310
-
311
- def __init__(self):
312
- super().__init__(InitializationPhase.EXTENSIONS)
313
-
314
- def initialize(self, context: InitializationContext) -> PhaseResult:
315
- """初始化扩展组件"""
316
- start_time = time.time()
317
-
318
- try:
319
- # 初始化扩展组件
320
- self._initialize_extensions(context)
321
-
322
- return self._create_result(
323
- success=True,
324
- duration=time.time() - start_time,
325
- artifacts={}
326
- )
327
-
328
- except Exception as e:
329
- return self._create_result(
330
- success=False,
331
- duration=time.time() - start_time,
332
- error=e
333
- )
334
-
335
- def _initialize_extensions(self, context: InitializationContext):
336
- """初始化扩展组件"""
337
- try:
338
- # 获取扩展配置
339
- extensions = []
340
- if context.settings:
341
- extensions = context.settings.get('EXTENSIONS', [])
342
- elif context.custom_settings:
343
- extensions = context.custom_settings.get('EXTENSIONS', [])
344
-
345
- # 初始化每个扩展
346
- initialized_extensions = []
347
- for extension_path in extensions:
348
- try:
349
- from crawlo.utils.class_loader import load_class
350
- extension_class = load_class(extension_path)
351
- extension_instance = extension_class()
352
- initialized_extensions.append(extension_instance)
353
- except Exception as e:
354
- if context.settings and context.settings.get('EXTENSIONS_STRICT', False):
355
- raise
356
- else:
357
- # 非严格模式下记录警告但继续
358
- context.add_warning(f"Failed to initialize extension {extension_path}: {e}")
359
-
360
- # 存储到上下文
361
- context.add_shared_data('extensions', initialized_extensions)
362
- except Exception as e:
363
- context.add_error(f"Failed to initialize extensions: {e}")
364
- raise
365
-
366
-
367
- class FrameworkStartupLogger(BaseInitializer):
368
- """框架启动日志记录器"""
369
-
370
- def __init__(self):
371
- # 使用新的FRAMEWORK_STARTUP_LOG阶段
372
- super().__init__(InitializationPhase.FRAMEWORK_STARTUP_LOG)
373
-
374
- def initialize(self, context: InitializationContext) -> PhaseResult:
375
- """记录框架启动日志"""
376
- start_time = time.time()
377
-
378
- try:
379
- # 获取框架logger
380
- from crawlo.logging import get_logger
381
- logger = get_logger('crawlo.framework')
382
-
383
- # 获取框架版本
384
- version = self._get_framework_version()
385
-
386
- # 记录框架启动信息(符合规范要求)
387
- logger.info(f"Crawlo Framework Started {version}")
388
-
389
- # 获取运行模式
390
- run_mode = "unknown"
391
- if context.settings:
392
- run_mode = context.settings.get('RUN_MODE', 'standalone')
393
- logger.info(f"Run mode: {run_mode}")
394
-
395
- # 注意:爬虫名称信息将在实际启动爬虫时记录,而不是在框架初始化时
396
-
397
- return self._create_result(
398
- success=True,
399
- duration=time.time() - start_time,
400
- artifacts={}
401
- )
402
-
403
- except Exception as e:
404
- # 即使日志记录失败,也不应该影响框架初始化
405
- return self._create_result(
406
- success=True, # 不影响初始化成功与否
407
- duration=time.time() - start_time,
408
- error=e
409
- )
410
-
411
- def _get_framework_version(self):
412
- """获取框架版本"""
413
- try:
414
- from crawlo import __version__
415
- return __version__
416
- except Exception:
417
- return "unknown"
418
-
419
-
420
- def register_built_in_initializers():
421
- """注册所有内置初始化器"""
422
- register_initializer(LoggingInitializer())
423
- register_initializer(SettingsInitializer())
424
- register_initializer(CoreComponentsInitializer())
425
- register_initializer(ExtensionsInitializer())
1
+ #!/usr/bin/python
2
+ # -*- coding: UTF-8 -*-
3
+ """
4
+ 内置初始化器 - 提供框架核心组件的初始化实现
5
+ """
6
+
7
+ import time
8
+ from .registry import BaseInitializer, register_initializer
9
+ from .phases import InitializationPhase, PhaseResult
10
+ from .context import InitializationContext
11
+
12
+
13
+ class LoggingInitializer(BaseInitializer):
14
+ """日志系统初始化器"""
15
+
16
+ def __init__(self):
17
+ super().__init__(InitializationPhase.LOGGING)
18
+
19
+ def initialize(self, context: InitializationContext) -> PhaseResult:
20
+ """初始化日志系统"""
21
+ start_time = time.time()
22
+
23
+ try:
24
+ # 导入日志模块
25
+ from crawlo.logging import configure_logging, LogConfig
26
+
27
+ # 获取日志配置
28
+ log_config = self._get_log_config(context)
29
+
30
+ # 确保日志目录存在
31
+ if log_config.file_path and log_config.file_enabled:
32
+ import os
33
+ log_dir = os.path.dirname(log_config.file_path)
34
+ if log_dir and not os.path.exists(log_dir):
35
+ os.makedirs(log_dir, exist_ok=True)
36
+
37
+ # 配置日志系统
38
+ configure_logging(log_config)
39
+
40
+ # 存储到共享数据
41
+ context.add_shared_data('log_config', log_config)
42
+
43
+ # 创建框架logger
44
+ from crawlo.logging import get_logger
45
+ framework_logger = get_logger('crawlo.framework')
46
+ context.add_shared_data('framework_logger', framework_logger)
47
+
48
+ return self._create_result(
49
+ success=True,
50
+ duration=time.time() - start_time,
51
+ artifacts={'log_config': log_config}
52
+ )
53
+
54
+ except Exception as e:
55
+ return self._create_result(
56
+ success=False,
57
+ duration=time.time() - start_time,
58
+ error=e
59
+ )
60
+
61
+ def _get_log_config(self, context: InitializationContext) -> 'LogConfig':
62
+ """
63
+ 获取日志配置
64
+
65
+ Args:
66
+ context: 初始化上下文
67
+
68
+ Returns:
69
+ LogConfig: 日志配置对象
70
+ """
71
+ # 导入日志配置类
72
+ from crawlo.logging import LogConfig
73
+
74
+ # 按优先级获取配置:自定义配置 > 上下文配置 > 项目配置 > 默认配置
75
+ config_sources = [
76
+ context.custom_settings,
77
+ context.settings,
78
+ self._load_project_config()
79
+ ]
80
+
81
+ # 遍历配置源
82
+ for config_source in config_sources:
83
+ if config_source:
84
+ log_config = self._create_log_config_from_source(config_source)
85
+ if log_config:
86
+ return log_config
87
+
88
+ # 使用默认配置
89
+ return LogConfig()
90
+
91
+ def _create_log_config_from_source(self, config_source) -> 'LogConfig':
92
+ """
93
+ 从配置源创建日志配置
94
+
95
+ Args:
96
+ config_source: 配置源
97
+
98
+ Returns:
99
+ LogConfig: 日志配置对象,如果配置源无效则返回None
100
+ """
101
+ # 导入日志配置类
102
+ from crawlo.logging import LogConfig
103
+
104
+ # 检查配置源是否有效
105
+ if not config_source:
106
+ return None
107
+
108
+ # 检查是否有日志相关配置
109
+ has_keys_method = hasattr(config_source, 'keys')
110
+ if has_keys_method:
111
+ has_log_config = any(key.startswith('LOG_') for key in config_source.keys())
112
+ else:
113
+ has_log_config = any(key.startswith('LOG_') for key in dir(config_source))
114
+
115
+ if not has_log_config:
116
+ return None
117
+
118
+ # 从配置源获取日志配置
119
+ log_level = config_source.get('LOG_LEVEL', 'INFO')
120
+ log_file = config_source.get('LOG_FILE')
121
+ log_format = config_source.get('LOG_FORMAT', '%(asctime)s - [%(name)s] - %(levelname)s: %(message)s')
122
+ log_encoding = config_source.get('LOG_ENCODING', 'utf-8')
123
+ log_max_bytes = config_source.get('LOG_MAX_BYTES', 10 * 1024 * 1024)
124
+ log_backup_count = config_source.get('LOG_BACKUP_COUNT', 5)
125
+ log_console_enabled = config_source.get('LOG_CONSOLE_ENABLED', True)
126
+ log_file_enabled = config_source.get('LOG_FILE_ENABLED', True)
127
+
128
+ # 创建日志配置
129
+ return LogConfig(
130
+ level=log_level,
131
+ format=log_format,
132
+ encoding=log_encoding,
133
+ file_path=log_file,
134
+ max_bytes=log_max_bytes,
135
+ backup_count=log_backup_count,
136
+ console_enabled=log_console_enabled,
137
+ file_enabled=log_file_enabled
138
+ )
139
+
140
+ def _load_project_config(self):
141
+ """
142
+ 自动加载项目配置以获取日志设置
143
+ """
144
+ try:
145
+ # 查找项目根目录
146
+ import os
147
+ import sys
148
+ import configparser
149
+
150
+ current_path = os.getcwd()
151
+
152
+ # 向上查找直到找到crawlo.cfg
153
+ checked_paths = set()
154
+ path = current_path
155
+
156
+ while path not in checked_paths:
157
+ checked_paths.add(path)
158
+
159
+ # 检查crawlo.cfg
160
+ cfg_file = os.path.join(path, "crawlo.cfg")
161
+ if os.path.exists(cfg_file):
162
+ # 读取配置文件
163
+ config_parser = configparser.ConfigParser()
164
+ config_parser.read(cfg_file, encoding="utf-8")
165
+
166
+ if config_parser.has_section("settings") and config_parser.has_option("settings", "default"):
167
+ # 获取settings模块路径
168
+ settings_module_path = config_parser.get("settings", "default")
169
+
170
+ # 添加项目根目录到Python路径
171
+ if path not in sys.path:
172
+ sys.path.insert(0, path)
173
+
174
+ # 导入项目配置模块
175
+ import importlib
176
+ settings_module = importlib.import_module(settings_module_path)
177
+
178
+ # 创建配置字典
179
+ project_config = {}
180
+ for key in dir(settings_module):
181
+ if key.isupper():
182
+ project_config[key] = getattr(settings_module, key)
183
+
184
+ return project_config
185
+
186
+ # 向上一级目录
187
+ parent = os.path.dirname(path)
188
+ if parent == path:
189
+ break
190
+ path = parent
191
+
192
+ return {}
193
+
194
+ except Exception as e:
195
+ return {}
196
+
197
+
198
+ class SettingsInitializer(BaseInitializer):
199
+ """配置系统初始化器"""
200
+
201
+ def __init__(self):
202
+ super().__init__(InitializationPhase.SETTINGS)
203
+
204
+ def initialize(self, context: InitializationContext) -> PhaseResult:
205
+ """初始化配置系统"""
206
+ start_time = time.time()
207
+
208
+ try:
209
+ # 导入配置管理器
210
+ from crawlo.settings.setting_manager import SettingManager
211
+ from crawlo.project import _load_project_settings
212
+
213
+ # 创建配置管理器并加载项目配置
214
+ settings = _load_project_settings(context.custom_settings)
215
+
216
+ # 存储到上下文
217
+ context.settings = settings
218
+ context.add_shared_data('settings', settings)
219
+
220
+ return self._create_result(
221
+ success=True,
222
+ duration=time.time() - start_time,
223
+ artifacts={'settings': settings}
224
+ )
225
+
226
+ except Exception as e:
227
+ return self._create_result(
228
+ success=False,
229
+ duration=time.time() - start_time,
230
+ error=e
231
+ )
232
+
233
+
234
+ class CoreComponentsInitializer(BaseInitializer):
235
+ """核心组件初始化器"""
236
+
237
+ def __init__(self):
238
+ super().__init__(InitializationPhase.CORE_COMPONENTS)
239
+
240
+ def initialize(self, context: InitializationContext) -> PhaseResult:
241
+ """初始化核心组件"""
242
+ start_time = time.time()
243
+
244
+ try:
245
+ # 在核心组件初始化阶段,大多数组件需要crawler参数
246
+ # 因此我们只初始化那些完全独立的组件
247
+ # 或者只记录需要初始化的组件类型,实际初始化在crawler创建时进行
248
+
249
+ return self._create_result(
250
+ success=True,
251
+ duration=time.time() - start_time,
252
+ artifacts={}
253
+ )
254
+
255
+ except Exception as e:
256
+ return self._create_result(
257
+ success=False,
258
+ duration=time.time() - start_time,
259
+ error=e
260
+ )
261
+
262
+ def _initialize_engine(self, context: InitializationContext):
263
+ """初始化引擎"""
264
+ try:
265
+ # 注意:Engine需要crawler参数,不能在此阶段初始化
266
+ pass
267
+ except Exception as e:
268
+ context.add_error(f"Failed to initialize engine: {e}")
269
+ raise
270
+
271
+ def _initialize_scheduler(self, context: InitializationContext):
272
+ """初始化调度器"""
273
+ try:
274
+ # 注意:Scheduler需要很多参数,不能在此阶段初始化
275
+ pass
276
+ except Exception as e:
277
+ context.add_error(f"Failed to initialize scheduler: {e}")
278
+ raise
279
+
280
+ def _initialize_downloader(self, context: InitializationContext):
281
+ """初始化下载器"""
282
+ try:
283
+ # 注意:下载器类需要crawler参数,不能在此阶段初始化实例
284
+ pass
285
+ except Exception as e:
286
+ context.add_error(f"Failed to initialize downloader: {e}")
287
+ raise
288
+
289
+ def _initialize_pipeline_manager(self, context: InitializationContext):
290
+ """初始化管道管理器"""
291
+ try:
292
+ # 注意:PipelineManager需要crawler参数,不能在此阶段初始化
293
+ pass
294
+ except Exception as e:
295
+ context.add_error(f"Failed to initialize pipeline manager: {e}")
296
+ raise
297
+
298
+ def _initialize_middleware_manager(self, context: InitializationContext):
299
+ """初始化中间件管理器"""
300
+ try:
301
+ # 注意:MiddlewareManager需要crawler参数,不能在此阶段初始化
302
+ pass
303
+ except Exception as e:
304
+ context.add_error(f"Failed to initialize middleware manager: {e}")
305
+ raise
306
+
307
+
308
+ class ExtensionsInitializer(BaseInitializer):
309
+ """扩展组件初始化器"""
310
+
311
+ def __init__(self):
312
+ super().__init__(InitializationPhase.EXTENSIONS)
313
+
314
+ def initialize(self, context: InitializationContext) -> PhaseResult:
315
+ """初始化扩展组件"""
316
+ start_time = time.time()
317
+
318
+ try:
319
+ # 初始化扩展组件
320
+ self._initialize_extensions(context)
321
+
322
+ return self._create_result(
323
+ success=True,
324
+ duration=time.time() - start_time,
325
+ artifacts={}
326
+ )
327
+
328
+ except Exception as e:
329
+ return self._create_result(
330
+ success=False,
331
+ duration=time.time() - start_time,
332
+ error=e
333
+ )
334
+
335
+ def _initialize_extensions(self, context: InitializationContext):
336
+ """初始化扩展组件"""
337
+ try:
338
+ # 获取扩展配置
339
+ extensions = []
340
+ if context.settings:
341
+ extensions = context.settings.get('EXTENSIONS', [])
342
+ elif context.custom_settings:
343
+ extensions = context.custom_settings.get('EXTENSIONS', [])
344
+
345
+ # 初始化每个扩展
346
+ initialized_extensions = []
347
+ for extension_path in extensions:
348
+ try:
349
+ from crawlo.utils.class_loader import load_class
350
+ extension_class = load_class(extension_path)
351
+ extension_instance = extension_class()
352
+ initialized_extensions.append(extension_instance)
353
+ except Exception as e:
354
+ if context.settings and context.settings.get('EXTENSIONS_STRICT', False):
355
+ raise
356
+ else:
357
+ # 非严格模式下记录警告但继续
358
+ context.add_warning(f"Failed to initialize extension {extension_path}: {e}")
359
+
360
+ # 存储到上下文
361
+ context.add_shared_data('extensions', initialized_extensions)
362
+ except Exception as e:
363
+ context.add_error(f"Failed to initialize extensions: {e}")
364
+ raise
365
+
366
+
367
+ class FrameworkStartupLogger(BaseInitializer):
368
+ """框架启动日志记录器"""
369
+
370
+ def __init__(self):
371
+ # 使用新的FRAMEWORK_STARTUP_LOG阶段
372
+ super().__init__(InitializationPhase.FRAMEWORK_STARTUP_LOG)
373
+
374
+ def initialize(self, context: InitializationContext) -> PhaseResult:
375
+ """记录框架启动日志"""
376
+ start_time = time.time()
377
+
378
+ try:
379
+ # 获取框架logger
380
+ from crawlo.logging import get_logger
381
+ logger = get_logger('crawlo.framework')
382
+
383
+ # 获取框架版本
384
+ version = self._get_framework_version()
385
+
386
+ # 记录框架启动信息(符合规范要求)
387
+ logger.info(f"Crawlo Framework Started {version}")
388
+
389
+ # 获取运行模式
390
+ run_mode = "unknown"
391
+ if context.settings:
392
+ run_mode = context.settings.get('RUN_MODE', 'standalone')
393
+ logger.info(f"Run mode: {run_mode}")
394
+
395
+ # 注意:爬虫名称信息将在实际启动爬虫时记录,而不是在框架初始化时
396
+
397
+ return self._create_result(
398
+ success=True,
399
+ duration=time.time() - start_time,
400
+ artifacts={}
401
+ )
402
+
403
+ except Exception as e:
404
+ # 即使日志记录失败,也不应该影响框架初始化
405
+ return self._create_result(
406
+ success=True, # 不影响初始化成功与否
407
+ duration=time.time() - start_time,
408
+ error=e
409
+ )
410
+
411
+ def _get_framework_version(self):
412
+ """获取框架版本"""
413
+ try:
414
+ from crawlo import __version__
415
+ return __version__
416
+ except Exception:
417
+ return "unknown"
418
+
419
+
420
+ def register_built_in_initializers():
421
+ """注册所有内置初始化器"""
422
+ register_initializer(LoggingInitializer())
423
+ register_initializer(SettingsInitializer())
424
+ register_initializer(CoreComponentsInitializer())
425
+ register_initializer(ExtensionsInitializer())
426
426
  register_initializer(FrameworkStartupLogger()) # 添加框架启动日志记录器