crawlo 1.4.7__py3-none-any.whl → 1.4.8__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 (348) hide show
  1. crawlo/__init__.py +90 -90
  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 +186 -186
  7. crawlo/commands/help.py +140 -140
  8. crawlo/commands/list.py +155 -155
  9. crawlo/commands/run.py +379 -379
  10. crawlo/commands/startproject.py +460 -460
  11. crawlo/commands/stats.py +187 -187
  12. crawlo/commands/utils.py +196 -196
  13. crawlo/config.py +320 -320
  14. crawlo/config_validator.py +277 -277
  15. crawlo/core/__init__.py +52 -52
  16. crawlo/core/engine.py +451 -451
  17. crawlo/core/processor.py +47 -47
  18. crawlo/core/scheduler.py +290 -290
  19. crawlo/crawler.py +698 -698
  20. crawlo/data/__init__.py +5 -5
  21. crawlo/data/user_agents.py +194 -194
  22. crawlo/downloader/__init__.py +280 -280
  23. crawlo/downloader/aiohttp_downloader.py +233 -233
  24. crawlo/downloader/cffi_downloader.py +250 -250
  25. crawlo/downloader/httpx_downloader.py +265 -265
  26. crawlo/downloader/hybrid_downloader.py +212 -212
  27. crawlo/downloader/playwright_downloader.py +425 -425
  28. crawlo/downloader/selenium_downloader.py +486 -486
  29. crawlo/event.py +45 -45
  30. crawlo/exceptions.py +214 -214
  31. crawlo/extension/__init__.py +64 -64
  32. crawlo/extension/health_check.py +141 -141
  33. crawlo/extension/log_interval.py +94 -94
  34. crawlo/extension/log_stats.py +70 -70
  35. crawlo/extension/logging_extension.py +53 -53
  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 +104 -104
  42. crawlo/factories/registry.py +84 -84
  43. crawlo/factories/utils.py +134 -134
  44. crawlo/filters/__init__.py +170 -170
  45. crawlo/filters/aioredis_filter.py +347 -347
  46. crawlo/filters/memory_filter.py +261 -261
  47. crawlo/framework.py +306 -306
  48. crawlo/initialization/__init__.py +44 -44
  49. crawlo/initialization/built_in.py +391 -391
  50. crawlo/initialization/context.py +141 -141
  51. crawlo/initialization/core.py +240 -240
  52. crawlo/initialization/phases.py +229 -229
  53. crawlo/initialization/registry.py +143 -143
  54. crawlo/initialization/utils.py +48 -48
  55. crawlo/interfaces.py +23 -23
  56. crawlo/items/__init__.py +23 -23
  57. crawlo/items/base.py +23 -23
  58. crawlo/items/fields.py +52 -52
  59. crawlo/items/items.py +104 -104
  60. crawlo/logging/__init__.py +42 -42
  61. crawlo/logging/config.py +280 -276
  62. crawlo/logging/factory.py +175 -175
  63. crawlo/logging/manager.py +104 -104
  64. crawlo/middleware/__init__.py +87 -87
  65. crawlo/middleware/default_header.py +132 -132
  66. crawlo/middleware/download_delay.py +104 -104
  67. crawlo/middleware/middleware_manager.py +142 -142
  68. crawlo/middleware/offsite.py +123 -123
  69. crawlo/middleware/proxy.py +209 -209
  70. crawlo/middleware/request_ignore.py +86 -86
  71. crawlo/middleware/response_code.py +150 -150
  72. crawlo/middleware/response_filter.py +136 -136
  73. crawlo/middleware/retry.py +124 -124
  74. crawlo/mode_manager.py +287 -287
  75. crawlo/network/__init__.py +21 -21
  76. crawlo/network/request.py +408 -376
  77. crawlo/network/response.py +598 -569
  78. crawlo/pipelines/__init__.py +52 -52
  79. crawlo/pipelines/base_pipeline.py +452 -452
  80. crawlo/pipelines/bloom_dedup_pipeline.py +145 -146
  81. crawlo/pipelines/console_pipeline.py +39 -39
  82. crawlo/pipelines/csv_pipeline.py +316 -316
  83. crawlo/pipelines/database_dedup_pipeline.py +196 -197
  84. crawlo/pipelines/json_pipeline.py +218 -218
  85. crawlo/pipelines/memory_dedup_pipeline.py +104 -105
  86. crawlo/pipelines/mongo_pipeline.py +140 -139
  87. crawlo/pipelines/mysql_pipeline.py +468 -469
  88. crawlo/pipelines/pipeline_manager.py +100 -100
  89. crawlo/pipelines/redis_dedup_pipeline.py +155 -155
  90. crawlo/project.py +347 -347
  91. crawlo/queue/__init__.py +9 -9
  92. crawlo/queue/pqueue.py +38 -38
  93. crawlo/queue/queue_manager.py +591 -591
  94. crawlo/queue/redis_priority_queue.py +518 -518
  95. crawlo/settings/__init__.py +7 -7
  96. crawlo/settings/default_settings.py +287 -284
  97. crawlo/settings/setting_manager.py +219 -219
  98. crawlo/spider/__init__.py +658 -657
  99. crawlo/stats_collector.py +81 -81
  100. crawlo/subscriber.py +129 -129
  101. crawlo/task_manager.py +138 -138
  102. crawlo/templates/crawlo.cfg.tmpl +10 -10
  103. crawlo/templates/project/__init__.py.tmpl +1 -1
  104. crawlo/templates/project/items.py.tmpl +13 -13
  105. crawlo/templates/project/middlewares.py.tmpl +38 -38
  106. crawlo/templates/project/pipelines.py.tmpl +35 -35
  107. crawlo/templates/project/settings.py.tmpl +113 -109
  108. crawlo/templates/project/settings_distributed.py.tmpl +160 -156
  109. crawlo/templates/project/settings_gentle.py.tmpl +174 -170
  110. crawlo/templates/project/settings_high_performance.py.tmpl +175 -171
  111. crawlo/templates/project/settings_minimal.py.tmpl +102 -98
  112. crawlo/templates/project/settings_simple.py.tmpl +172 -168
  113. crawlo/templates/project/spiders/__init__.py.tmpl +9 -9
  114. crawlo/templates/run.py.tmpl +23 -23
  115. crawlo/templates/spider/spider.py.tmpl +32 -32
  116. crawlo/templates/spiders_init.py.tmpl +4 -4
  117. crawlo/tools/__init__.py +86 -86
  118. crawlo/tools/date_tools.py +289 -289
  119. crawlo/tools/distributed_coordinator.py +384 -384
  120. crawlo/tools/scenario_adapter.py +262 -262
  121. crawlo/tools/text_cleaner.py +232 -232
  122. crawlo/utils/__init__.py +74 -50
  123. crawlo/utils/batch_processor.py +276 -276
  124. crawlo/utils/config_manager.py +442 -442
  125. crawlo/utils/controlled_spider_mixin.py +439 -439
  126. crawlo/utils/db_helper.py +250 -250
  127. crawlo/utils/encoding_helper.py +190 -0
  128. crawlo/utils/error_handler.py +410 -410
  129. crawlo/utils/fingerprint.py +121 -121
  130. crawlo/utils/func_tools.py +82 -82
  131. crawlo/utils/large_scale_helper.py +344 -344
  132. crawlo/utils/leak_detector.py +335 -335
  133. crawlo/utils/misc.py +81 -81
  134. crawlo/utils/mongo_connection_pool.py +157 -157
  135. crawlo/utils/mysql_connection_pool.py +197 -197
  136. crawlo/utils/performance_monitor.py +285 -285
  137. crawlo/utils/queue_helper.py +175 -175
  138. crawlo/utils/redis_checker.py +90 -90
  139. crawlo/utils/redis_connection_pool.py +578 -578
  140. crawlo/utils/redis_key_validator.py +198 -198
  141. crawlo/utils/request.py +278 -278
  142. crawlo/utils/request_serializer.py +225 -225
  143. crawlo/utils/resource_manager.py +337 -337
  144. crawlo/utils/response_helper.py +113 -0
  145. crawlo/utils/selector_helper.py +138 -137
  146. crawlo/utils/singleton.py +69 -69
  147. crawlo/utils/spider_loader.py +201 -201
  148. crawlo/utils/text_helper.py +94 -94
  149. {crawlo-1.4.7.dist-info → crawlo-1.4.8.dist-info}/METADATA +831 -689
  150. crawlo-1.4.8.dist-info/RECORD +347 -0
  151. examples/__init__.py +7 -7
  152. tests/__init__.py +7 -7
  153. tests/advanced_tools_example.py +217 -217
  154. tests/authenticated_proxy_example.py +110 -110
  155. tests/baidu_performance_test.py +108 -108
  156. tests/baidu_test.py +59 -59
  157. tests/bug_check_test.py +250 -250
  158. tests/cleaners_example.py +160 -160
  159. tests/comprehensive_framework_test.py +212 -212
  160. tests/comprehensive_test.py +81 -81
  161. tests/comprehensive_testing_summary.md +186 -186
  162. tests/config_validation_demo.py +142 -142
  163. tests/controlled_spider_example.py +205 -205
  164. tests/date_tools_example.py +180 -180
  165. tests/debug_configure.py +69 -69
  166. tests/debug_framework_logger.py +84 -84
  167. tests/debug_log_config.py +126 -126
  168. tests/debug_log_levels.py +63 -63
  169. tests/debug_pipelines.py +66 -66
  170. tests/detailed_log_test.py +233 -233
  171. tests/direct_selector_helper_test.py +96 -96
  172. tests/distributed_dedup_test.py +467 -467
  173. tests/distributed_test.py +66 -66
  174. tests/distributed_test_debug.py +76 -76
  175. tests/dynamic_loading_example.py +523 -523
  176. tests/dynamic_loading_test.py +104 -104
  177. tests/error_handling_example.py +171 -171
  178. tests/explain_mysql_update_behavior.py +76 -76
  179. tests/final_comprehensive_test.py +151 -151
  180. tests/final_log_test.py +260 -260
  181. tests/final_validation_test.py +182 -182
  182. tests/fix_log_test.py +142 -142
  183. tests/framework_performance_test.py +202 -202
  184. tests/log_buffering_test.py +111 -111
  185. tests/log_generation_timing_test.py +153 -153
  186. tests/monitor_redis_dedup.sh +72 -72
  187. tests/ofweek_scrapy/ofweek_scrapy/items.py +12 -12
  188. tests/ofweek_scrapy/ofweek_scrapy/middlewares.py +100 -100
  189. tests/ofweek_scrapy/ofweek_scrapy/pipelines.py +13 -13
  190. tests/ofweek_scrapy/ofweek_scrapy/settings.py +84 -84
  191. tests/ofweek_scrapy/ofweek_scrapy/spiders/__init__.py +4 -4
  192. tests/ofweek_scrapy/scrapy.cfg +11 -11
  193. tests/optimized_performance_test.py +211 -211
  194. tests/performance_comparison.py +244 -244
  195. tests/queue_blocking_test.py +113 -113
  196. tests/queue_test.py +89 -89
  197. tests/redis_key_validation_demo.py +130 -130
  198. tests/request_params_example.py +150 -150
  199. tests/response_improvements_example.py +144 -144
  200. tests/scrapy_comparison/ofweek_scrapy.py +138 -138
  201. tests/scrapy_comparison/scrapy_test.py +133 -133
  202. tests/simple_cli_test.py +54 -54
  203. tests/simple_command_test.py +119 -119
  204. tests/simple_crawlo_test.py +126 -126
  205. tests/simple_follow_test.py +38 -38
  206. tests/simple_log_test2.py +137 -137
  207. tests/simple_optimization_test.py +128 -128
  208. tests/simple_queue_type_test.py +41 -41
  209. tests/simple_response_selector_test.py +94 -94
  210. tests/simple_selector_helper_test.py +154 -154
  211. tests/simple_selector_test.py +207 -207
  212. tests/simple_spider_test.py +49 -49
  213. tests/simple_url_test.py +73 -73
  214. tests/simulate_mysql_update_test.py +139 -139
  215. tests/spider_log_timing_test.py +177 -177
  216. tests/test_advanced_tools.py +148 -148
  217. tests/test_all_commands.py +230 -230
  218. tests/test_all_pipeline_fingerprints.py +133 -133
  219. tests/test_all_redis_key_configs.py +145 -145
  220. tests/test_asyncmy_usage.py +56 -56
  221. tests/test_batch_processor.py +178 -178
  222. tests/test_cleaners.py +54 -54
  223. tests/test_cli_arguments.py +118 -118
  224. tests/test_component_factory.py +174 -174
  225. tests/test_config_consistency.py +80 -80
  226. tests/test_config_merge.py +152 -152
  227. tests/test_config_validator.py +182 -182
  228. tests/test_controlled_spider_mixin.py +79 -79
  229. tests/test_crawler_process_import.py +38 -38
  230. tests/test_crawler_process_spider_modules.py +47 -47
  231. tests/test_crawlo_proxy_integration.py +114 -114
  232. tests/test_date_tools.py +123 -123
  233. tests/test_dedup_fix.py +220 -220
  234. tests/test_dedup_pipeline_consistency.py +124 -124
  235. tests/test_default_header_middleware.py +313 -313
  236. tests/test_distributed.py +65 -65
  237. tests/test_double_crawlo_fix.py +204 -204
  238. tests/test_double_crawlo_fix_simple.py +124 -124
  239. tests/test_download_delay_middleware.py +221 -221
  240. tests/test_downloader_proxy_compatibility.py +272 -272
  241. tests/test_edge_cases.py +305 -305
  242. tests/test_encoding_core.py +56 -56
  243. tests/test_encoding_detection.py +126 -126
  244. tests/test_enhanced_error_handler.py +270 -270
  245. tests/test_enhanced_error_handler_comprehensive.py +245 -245
  246. tests/test_error_handler_compatibility.py +112 -112
  247. tests/test_factories.py +252 -252
  248. tests/test_factory_compatibility.py +196 -196
  249. tests/test_final_validation.py +153 -153
  250. tests/test_fingerprint_consistency.py +135 -135
  251. tests/test_fingerprint_simple.py +51 -51
  252. tests/test_get_component_logger.py +83 -83
  253. tests/test_hash_performance.py +99 -99
  254. tests/test_integration.py +169 -169
  255. tests/test_item_dedup_redis_key.py +122 -122
  256. tests/test_large_scale_helper.py +235 -235
  257. tests/test_logging_enhancements.py +374 -374
  258. tests/test_logging_final.py +184 -184
  259. tests/test_logging_integration.py +312 -312
  260. tests/test_logging_system.py +282 -282
  261. tests/test_middleware_debug.py +141 -141
  262. tests/test_mode_consistency.py +51 -51
  263. tests/test_multi_directory.py +67 -67
  264. tests/test_multiple_spider_modules.py +80 -80
  265. tests/test_mysql_pipeline_config.py +164 -164
  266. tests/test_mysql_pipeline_error.py +98 -98
  267. tests/test_mysql_pipeline_init_log.py +82 -82
  268. tests/test_mysql_pipeline_integration.py +132 -132
  269. tests/test_mysql_pipeline_refactor.py +143 -143
  270. tests/test_mysql_pipeline_refactor_simple.py +85 -85
  271. tests/test_mysql_pipeline_robustness.py +195 -195
  272. tests/test_mysql_pipeline_types.py +88 -88
  273. tests/test_mysql_update_columns.py +93 -93
  274. tests/test_offsite_middleware.py +244 -244
  275. tests/test_offsite_middleware_simple.py +203 -203
  276. tests/test_optimized_selector_naming.py +100 -100
  277. tests/test_parsel.py +29 -29
  278. tests/test_performance.py +327 -327
  279. tests/test_performance_monitor.py +115 -115
  280. tests/test_pipeline_fingerprint_consistency.py +86 -86
  281. tests/test_priority_behavior.py +211 -211
  282. tests/test_priority_consistency.py +151 -151
  283. tests/test_priority_consistency_fixed.py +249 -249
  284. tests/test_proxy_health_check.py +32 -32
  285. tests/test_proxy_middleware.py +217 -217
  286. tests/test_proxy_middleware_enhanced.py +212 -212
  287. tests/test_proxy_middleware_integration.py +142 -142
  288. tests/test_proxy_middleware_refactored.py +207 -207
  289. tests/test_proxy_only.py +83 -83
  290. tests/test_proxy_providers.py +56 -56
  291. tests/test_proxy_stats.py +19 -19
  292. tests/test_proxy_strategies.py +59 -59
  293. tests/test_proxy_with_downloader.py +152 -152
  294. tests/test_queue_empty_check.py +41 -41
  295. tests/test_queue_manager_double_crawlo.py +173 -173
  296. tests/test_queue_manager_redis_key.py +179 -179
  297. tests/test_queue_naming.py +154 -154
  298. tests/test_queue_type.py +106 -106
  299. tests/test_queue_type_redis_config_consistency.py +130 -130
  300. tests/test_random_headers_default.py +322 -322
  301. tests/test_random_headers_necessity.py +308 -308
  302. tests/test_random_user_agent.py +72 -72
  303. tests/test_redis_config.py +28 -28
  304. tests/test_redis_connection_pool.py +294 -294
  305. tests/test_redis_key_naming.py +181 -181
  306. tests/test_redis_key_validator.py +123 -123
  307. tests/test_redis_queue.py +224 -224
  308. tests/test_redis_queue_name_fix.py +175 -175
  309. tests/test_redis_queue_type_fallback.py +129 -129
  310. tests/test_request_ignore_middleware.py +182 -182
  311. tests/test_request_params.py +111 -111
  312. tests/test_request_serialization.py +70 -70
  313. tests/test_response_code_middleware.py +349 -349
  314. tests/test_response_filter_middleware.py +427 -427
  315. tests/test_response_follow.py +104 -104
  316. tests/test_response_improvements.py +152 -152
  317. tests/test_response_selector_methods.py +92 -92
  318. tests/test_response_url_methods.py +70 -70
  319. tests/test_response_urljoin.py +86 -86
  320. tests/test_retry_middleware.py +333 -333
  321. tests/test_retry_middleware_realistic.py +273 -273
  322. tests/test_scheduler.py +252 -252
  323. tests/test_scheduler_config_update.py +133 -133
  324. tests/test_scrapy_style_encoding.py +112 -112
  325. tests/test_selector_helper.py +100 -100
  326. tests/test_selector_optimizations.py +146 -146
  327. tests/test_simple_response.py +61 -61
  328. tests/test_spider_loader.py +49 -49
  329. tests/test_spider_loader_comprehensive.py +69 -69
  330. tests/test_spider_modules.py +84 -84
  331. tests/test_spiders/test_spider.py +9 -9
  332. tests/test_telecom_spider_redis_key.py +205 -205
  333. tests/test_template_content.py +87 -87
  334. tests/test_template_redis_key.py +134 -134
  335. tests/test_tools.py +159 -159
  336. tests/test_user_agent_randomness.py +176 -176
  337. tests/test_user_agents.py +96 -96
  338. tests/untested_features_report.md +138 -138
  339. tests/verify_debug.py +51 -51
  340. tests/verify_distributed.py +117 -117
  341. tests/verify_log_fix.py +111 -111
  342. tests/verify_mysql_warnings.py +109 -109
  343. crawlo/utils/log.py +0 -80
  344. crawlo/utils/url_utils.py +0 -40
  345. crawlo-1.4.7.dist-info/RECORD +0 -347
  346. {crawlo-1.4.7.dist-info → crawlo-1.4.8.dist-info}/WHEEL +0 -0
  347. {crawlo-1.4.7.dist-info → crawlo-1.4.8.dist-info}/entry_points.txt +0 -0
  348. {crawlo-1.4.7.dist-info → crawlo-1.4.8.dist-info}/top_level.txt +0 -0
@@ -1,278 +1,278 @@
1
- #!/usr/bin/python
2
- # -*- coding: UTF-8 -*-
3
- """
4
- 配置验证器
5
- ==========
6
- 提供配置项的验证和默认值设置功能,确保配置的合理性和一致性。
7
- """
8
- from typing import Dict, Any, List, Tuple
9
-
10
- from crawlo.logging import get_logger
11
-
12
-
13
- class ConfigValidator:
14
- """配置验证器"""
15
-
16
- def __init__(self):
17
- self.logger = get_logger(self.__class__.__name__)
18
- self.errors = []
19
- self.warnings = []
20
-
21
- def validate(self, config: Dict[str, Any]) -> Tuple[bool, List[str], List[str]]:
22
- """
23
- 验证配置
24
-
25
- Args:
26
- config: 配置字典
27
-
28
- Returns:
29
- Tuple[bool, List[str], List[str]]: (是否有效, 错误列表, 警告列表)
30
- """
31
- self.errors = []
32
- self.warnings = []
33
-
34
- # 验证各个配置项
35
- self._validate_basic_settings(config)
36
- self._validate_network_settings(config)
37
- self._validate_concurrency_settings(config)
38
- self._validate_queue_settings(config)
39
- self._validate_storage_settings(config)
40
- self._validate_redis_settings(config)
41
- self._validate_middleware_settings(config)
42
- self._validate_pipeline_settings(config)
43
- self._validate_extension_settings(config)
44
- self._validate_logging_settings(config)
45
-
46
- is_valid = len(self.errors) == 0
47
- return is_valid, self.errors, self.warnings
48
-
49
- def _validate_basic_settings(self, config: Dict[str, Any]):
50
- """验证基本设置"""
51
- project_name = config.get('PROJECT_NAME', 'crawlo')
52
- if not isinstance(project_name, str) or not project_name.strip():
53
- self.errors.append("PROJECT_NAME 必须是非空字符串")
54
-
55
- version = config.get('VERSION', '1.0')
56
- if not isinstance(version, str):
57
- self.errors.append("VERSION 必须是字符串")
58
-
59
- def _validate_network_settings(self, config: Dict[str, Any]):
60
- """验证网络设置"""
61
- # 下载器验证
62
- downloader = config.get('DOWNLOADER', 'crawlo.downloader.aiohttp_downloader.AioHttpDownloader')
63
- if not isinstance(downloader, str):
64
- self.errors.append("DOWNLOADER 必须是字符串")
65
-
66
- # 超时设置验证
67
- timeout = config.get('DOWNLOAD_TIMEOUT', 30)
68
- if not isinstance(timeout, (int, float)) or timeout <= 0:
69
- self.errors.append("DOWNLOAD_TIMEOUT 必须是正数")
70
-
71
- # 延迟设置验证
72
- delay = config.get('DOWNLOAD_DELAY', 1.0)
73
- if not isinstance(delay, (int, float)) or delay < 0:
74
- self.errors.append("DOWNLOAD_DELAY 必须是非负数")
75
-
76
- # 重试次数验证
77
- max_retries = config.get('MAX_RETRY_TIMES', 3)
78
- if not isinstance(max_retries, int) or max_retries < 0:
79
- self.errors.append("MAX_RETRY_TIMES 必须是非负整数")
80
-
81
- # 连接池限制验证
82
- pool_limit = config.get('CONNECTION_POOL_LIMIT', 50)
83
- if not isinstance(pool_limit, int) or pool_limit <= 0:
84
- self.errors.append("CONNECTION_POOL_LIMIT 必须是正整数")
85
-
86
- def _validate_concurrency_settings(self, config: Dict[str, Any]):
87
- """验证并发设置"""
88
- concurrency = config.get('CONCURRENCY', 8)
89
- if not isinstance(concurrency, int) or concurrency <= 0:
90
- self.errors.append("CONCURRENCY 必须是正整数")
91
-
92
- max_running_spiders = config.get('MAX_RUNNING_SPIDERS', 3)
93
- if not isinstance(max_running_spiders, int) or max_running_spiders <= 0:
94
- self.errors.append("MAX_RUNNING_SPIDERS 必须是正整数")
95
-
96
- def _validate_queue_settings(self, config: Dict[str, Any]):
97
- """验证队列设置"""
98
- queue_type = config.get('QUEUE_TYPE', 'memory')
99
- valid_queue_types = ['memory', 'redis', 'auto']
100
- if queue_type not in valid_queue_types:
101
- self.errors.append(f"QUEUE_TYPE 必须是以下值之一: {valid_queue_types}")
102
-
103
- # 队列大小验证
104
- max_queue_size = config.get('SCHEDULER_MAX_QUEUE_SIZE', 2000)
105
- if not isinstance(max_queue_size, int) or max_queue_size <= 0:
106
- self.errors.append("SCHEDULER_MAX_QUEUE_SIZE 必须是正整数")
107
-
108
- # 队列名称验证(如果是Redis队列)
109
- if queue_type == 'redis':
110
- queue_name = config.get('SCHEDULER_QUEUE_NAME', '')
111
- if not queue_name:
112
- self.errors.append("使用Redis队列时,SCHEDULER_QUEUE_NAME 不能为空")
113
- elif not self._is_valid_redis_key(queue_name):
114
- self.warnings.append(f"Redis队列名称 '{queue_name}' 不符合命名规范,建议使用 'crawlo:{config.get('PROJECT_NAME', 'project')}:queue:requests' 格式")
115
-
116
- def _validate_storage_settings(self, config: Dict[str, Any]):
117
- """验证存储设置"""
118
- # MySQL设置验证
119
- mysql_host = config.get('MYSQL_HOST')
120
- if mysql_host is not None and not isinstance(mysql_host, str):
121
- self.errors.append("MYSQL_HOST 必须是字符串")
122
-
123
- mysql_port = config.get('MYSQL_PORT')
124
- if mysql_port is not None and (not isinstance(mysql_port, int) or mysql_port <= 0 or mysql_port > 65535):
125
- self.errors.append("MYSQL_PORT 必须是1-65535之间的整数")
126
-
127
- # MongoDB设置验证
128
- mongo_uri = config.get('MONGO_URI')
129
- if mongo_uri is not None and not isinstance(mongo_uri, str):
130
- self.errors.append("MONGO_URI 必须是字符串")
131
-
132
- def _validate_redis_settings(self, config: Dict[str, Any]):
133
- """验证Redis设置"""
134
- queue_type = config.get('QUEUE_TYPE', 'memory')
135
- if queue_type == 'redis':
136
- # Redis主机验证
137
- redis_host = config.get('REDIS_HOST', '127.0.0.1')
138
- if not isinstance(redis_host, str) or not redis_host.strip():
139
- self.errors.append("REDIS_HOST 必须是非空字符串")
140
-
141
- # Redis端口验证
142
- redis_port = config.get('REDIS_PORT', 6379)
143
- if not isinstance(redis_port, int) or redis_port <= 0 or redis_port > 65535:
144
- self.errors.append("REDIS_PORT 必须是1-65535之间的整数")
145
-
146
- # Redis URL验证
147
- redis_url = config.get('REDIS_URL')
148
- if redis_url is not None and not isinstance(redis_url, str):
149
- self.errors.append("REDIS_URL 必须是字符串")
150
-
151
- # Redis队列名称验证(提供默认值)
152
- scheduler_queue_name = config.get('SCHEDULER_QUEUE_NAME')
153
- project_name = config.get('PROJECT_NAME', 'crawlo')
154
- if scheduler_queue_name is None:
155
- # 如果没有设置,使用默认值
156
- scheduler_queue_name = f'crawlo:{project_name}:queue:requests'
157
-
158
- if not scheduler_queue_name:
159
- self.errors.append("使用Redis队列时,SCHEDULER_QUEUE_NAME 不能为空")
160
- elif not self._is_valid_redis_key(scheduler_queue_name):
161
- self.warnings.append(f"Redis队列名称 '{scheduler_queue_name}' 不符合命名规范,建议使用 'crawlo:{project_name}:queue:requests' 格式")
162
-
163
- def _validate_middleware_settings(self, config: Dict[str, Any]):
164
- """验证中间件设置"""
165
- # 验证 MIDDLEWARES
166
- middlewares = config.get('MIDDLEWARES', [])
167
- if not isinstance(middlewares, list):
168
- self.errors.append("MIDDLEWARES 必须是列表")
169
- else:
170
- for i, middleware in enumerate(middlewares):
171
- if not isinstance(middleware, str):
172
- self.errors.append(f"MIDDLEWARES[{i}] 必须是字符串")
173
-
174
- def _validate_pipeline_settings(self, config: Dict[str, Any]):
175
- """验证管道设置"""
176
- # 验证 PIPELINES
177
- pipelines = config.get('PIPELINES', [])
178
- if not isinstance(pipelines, list):
179
- self.errors.append("PIPELINES 必须是列表")
180
- else:
181
- for i, pipeline in enumerate(pipelines):
182
- if not isinstance(pipeline, str):
183
- self.errors.append(f"PIPELINES[{i}] 必须是字符串")
184
-
185
- def _validate_extension_settings(self, config: Dict[str, Any]):
186
- """验证扩展设置"""
187
- # 验证 EXTENSIONS
188
- extensions = config.get('EXTENSIONS', [])
189
- if not isinstance(extensions, list):
190
- self.errors.append("EXTENSIONS 必须是列表")
191
- else:
192
- for i, extension in enumerate(extensions):
193
- if not isinstance(extension, str):
194
- self.errors.append(f"EXTENSIONS[{i}] 必须是字符串")
195
-
196
- def _validate_logging_settings(self, config: Dict[str, Any]):
197
- """验证日志设置"""
198
- log_level = config.get('LOG_LEVEL', 'INFO')
199
- valid_log_levels = ['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL']
200
- if log_level not in valid_log_levels:
201
- self.errors.append(f"LOG_LEVEL 必须是以下值之一: {valid_log_levels}")
202
-
203
- log_file = config.get('LOG_FILE')
204
- if log_file is not None and not isinstance(log_file, str):
205
- self.errors.append("LOG_FILE 必须是字符串")
206
-
207
- def _is_valid_redis_key(self, key: str) -> bool:
208
- """检查Redis key是否符合命名规范"""
209
- # 检查是否以 crawlo: 开头
210
- if not key.startswith('crawlo:'):
211
- return False
212
-
213
- # 检查是否包含必要的部分
214
- parts = key.split(':')
215
- if len(parts) < 3:
216
- return False
217
-
218
- # 检查是否包含 queue 部分
219
- return 'queue' in parts
220
-
221
- def get_validation_report(self, config: Dict[str, Any]) -> str:
222
- """
223
- 获取验证报告
224
-
225
- Args:
226
- config: 配置字典
227
-
228
- Returns:
229
- str: 验证报告
230
- """
231
- is_valid, errors, warnings = self.validate(config)
232
-
233
- report = []
234
- report.append("=" * 50)
235
- report.append("配置验证报告")
236
- report.append("=" * 50)
237
-
238
- if is_valid:
239
- report.append("配置验证通过")
240
- else:
241
- report.append("配置验证失败")
242
- report.append("错误:")
243
- for error in errors:
244
- report.append(f" - {error}")
245
-
246
- if warnings:
247
- report.append("警告:")
248
- for warning in warnings:
249
- report.append(f" - {warning}")
250
-
251
- report.append("=" * 50)
252
- return "\n".join(report)
253
-
254
-
255
- # 便利函数
256
- def validate_config(config: Dict[str, Any]) -> Tuple[bool, List[str], List[str]]:
257
- """
258
- 验证配置
259
-
260
- Args:
261
- config: 配置字典
262
-
263
- Returns:
264
- Tuple[bool, List[str], List[str]]: (是否有效, 错误列表, 警告列表)
265
- """
266
- validator = ConfigValidator()
267
- return validator.validate(config)
268
-
269
-
270
- def print_validation_report(config: Dict[str, Any]):
271
- """
272
- 打印验证报告
273
-
274
- Args:
275
- config: 配置字典
276
- """
277
- validator = ConfigValidator()
1
+ #!/usr/bin/python
2
+ # -*- coding: UTF-8 -*-
3
+ """
4
+ 配置验证器
5
+ ==========
6
+ 提供配置项的验证和默认值设置功能,确保配置的合理性和一致性。
7
+ """
8
+ from typing import Dict, Any, List, Tuple
9
+
10
+ from crawlo.logging import get_logger
11
+
12
+
13
+ class ConfigValidator:
14
+ """配置验证器"""
15
+
16
+ def __init__(self):
17
+ self.logger = get_logger(self.__class__.__name__)
18
+ self.errors = []
19
+ self.warnings = []
20
+
21
+ def validate(self, config: Dict[str, Any]) -> Tuple[bool, List[str], List[str]]:
22
+ """
23
+ 验证配置
24
+
25
+ Args:
26
+ config: 配置字典
27
+
28
+ Returns:
29
+ Tuple[bool, List[str], List[str]]: (是否有效, 错误列表, 警告列表)
30
+ """
31
+ self.errors = []
32
+ self.warnings = []
33
+
34
+ # 验证各个配置项
35
+ self._validate_basic_settings(config)
36
+ self._validate_network_settings(config)
37
+ self._validate_concurrency_settings(config)
38
+ self._validate_queue_settings(config)
39
+ self._validate_storage_settings(config)
40
+ self._validate_redis_settings(config)
41
+ self._validate_middleware_settings(config)
42
+ self._validate_pipeline_settings(config)
43
+ self._validate_extension_settings(config)
44
+ self._validate_logging_settings(config)
45
+
46
+ is_valid = len(self.errors) == 0
47
+ return is_valid, self.errors, self.warnings
48
+
49
+ def _validate_basic_settings(self, config: Dict[str, Any]):
50
+ """验证基本设置"""
51
+ project_name = config.get('PROJECT_NAME', 'crawlo')
52
+ if not isinstance(project_name, str) or not project_name.strip():
53
+ self.errors.append("PROJECT_NAME 必须是非空字符串")
54
+
55
+ version = config.get('VERSION', '1.0')
56
+ if not isinstance(version, str):
57
+ self.errors.append("VERSION 必须是字符串")
58
+
59
+ def _validate_network_settings(self, config: Dict[str, Any]):
60
+ """验证网络设置"""
61
+ # 下载器验证
62
+ downloader = config.get('DOWNLOADER', 'crawlo.downloader.aiohttp_downloader.AioHttpDownloader')
63
+ if not isinstance(downloader, str):
64
+ self.errors.append("DOWNLOADER 必须是字符串")
65
+
66
+ # 超时设置验证
67
+ timeout = config.get('DOWNLOAD_TIMEOUT', 30)
68
+ if not isinstance(timeout, (int, float)) or timeout <= 0:
69
+ self.errors.append("DOWNLOAD_TIMEOUT 必须是正数")
70
+
71
+ # 延迟设置验证
72
+ delay = config.get('DOWNLOAD_DELAY', 1.0)
73
+ if not isinstance(delay, (int, float)) or delay < 0:
74
+ self.errors.append("DOWNLOAD_DELAY 必须是非负数")
75
+
76
+ # 重试次数验证
77
+ max_retries = config.get('MAX_RETRY_TIMES', 3)
78
+ if not isinstance(max_retries, int) or max_retries < 0:
79
+ self.errors.append("MAX_RETRY_TIMES 必须是非负整数")
80
+
81
+ # 连接池限制验证
82
+ pool_limit = config.get('CONNECTION_POOL_LIMIT', 50)
83
+ if not isinstance(pool_limit, int) or pool_limit <= 0:
84
+ self.errors.append("CONNECTION_POOL_LIMIT 必须是正整数")
85
+
86
+ def _validate_concurrency_settings(self, config: Dict[str, Any]):
87
+ """验证并发设置"""
88
+ concurrency = config.get('CONCURRENCY', 8)
89
+ if not isinstance(concurrency, int) or concurrency <= 0:
90
+ self.errors.append("CONCURRENCY 必须是正整数")
91
+
92
+ max_running_spiders = config.get('MAX_RUNNING_SPIDERS', 3)
93
+ if not isinstance(max_running_spiders, int) or max_running_spiders <= 0:
94
+ self.errors.append("MAX_RUNNING_SPIDERS 必须是正整数")
95
+
96
+ def _validate_queue_settings(self, config: Dict[str, Any]):
97
+ """验证队列设置"""
98
+ queue_type = config.get('QUEUE_TYPE', 'memory')
99
+ valid_queue_types = ['memory', 'redis', 'auto']
100
+ if queue_type not in valid_queue_types:
101
+ self.errors.append(f"QUEUE_TYPE 必须是以下值之一: {valid_queue_types}")
102
+
103
+ # 队列大小验证
104
+ max_queue_size = config.get('SCHEDULER_MAX_QUEUE_SIZE', 2000)
105
+ if not isinstance(max_queue_size, int) or max_queue_size <= 0:
106
+ self.errors.append("SCHEDULER_MAX_QUEUE_SIZE 必须是正整数")
107
+
108
+ # 队列名称验证(如果是Redis队列)
109
+ if queue_type == 'redis':
110
+ queue_name = config.get('SCHEDULER_QUEUE_NAME', '')
111
+ if not queue_name:
112
+ self.errors.append("使用Redis队列时,SCHEDULER_QUEUE_NAME 不能为空")
113
+ elif not self._is_valid_redis_key(queue_name):
114
+ self.warnings.append(f"Redis队列名称 '{queue_name}' 不符合命名规范,建议使用 'crawlo:{config.get('PROJECT_NAME', 'project')}:queue:requests' 格式")
115
+
116
+ def _validate_storage_settings(self, config: Dict[str, Any]):
117
+ """验证存储设置"""
118
+ # MySQL设置验证
119
+ mysql_host = config.get('MYSQL_HOST')
120
+ if mysql_host is not None and not isinstance(mysql_host, str):
121
+ self.errors.append("MYSQL_HOST 必须是字符串")
122
+
123
+ mysql_port = config.get('MYSQL_PORT')
124
+ if mysql_port is not None and (not isinstance(mysql_port, int) or mysql_port <= 0 or mysql_port > 65535):
125
+ self.errors.append("MYSQL_PORT 必须是1-65535之间的整数")
126
+
127
+ # MongoDB设置验证
128
+ mongo_uri = config.get('MONGO_URI')
129
+ if mongo_uri is not None and not isinstance(mongo_uri, str):
130
+ self.errors.append("MONGO_URI 必须是字符串")
131
+
132
+ def _validate_redis_settings(self, config: Dict[str, Any]):
133
+ """验证Redis设置"""
134
+ queue_type = config.get('QUEUE_TYPE', 'memory')
135
+ if queue_type == 'redis':
136
+ # Redis主机验证
137
+ redis_host = config.get('REDIS_HOST', '127.0.0.1')
138
+ if not isinstance(redis_host, str) or not redis_host.strip():
139
+ self.errors.append("REDIS_HOST 必须是非空字符串")
140
+
141
+ # Redis端口验证
142
+ redis_port = config.get('REDIS_PORT', 6379)
143
+ if not isinstance(redis_port, int) or redis_port <= 0 or redis_port > 65535:
144
+ self.errors.append("REDIS_PORT 必须是1-65535之间的整数")
145
+
146
+ # Redis URL验证
147
+ redis_url = config.get('REDIS_URL')
148
+ if redis_url is not None and not isinstance(redis_url, str):
149
+ self.errors.append("REDIS_URL 必须是字符串")
150
+
151
+ # Redis队列名称验证(提供默认值)
152
+ scheduler_queue_name = config.get('SCHEDULER_QUEUE_NAME')
153
+ project_name = config.get('PROJECT_NAME', 'crawlo')
154
+ if scheduler_queue_name is None:
155
+ # 如果没有设置,使用默认值
156
+ scheduler_queue_name = f'crawlo:{project_name}:queue:requests'
157
+
158
+ if not scheduler_queue_name:
159
+ self.errors.append("使用Redis队列时,SCHEDULER_QUEUE_NAME 不能为空")
160
+ elif not self._is_valid_redis_key(scheduler_queue_name):
161
+ self.warnings.append(f"Redis队列名称 '{scheduler_queue_name}' 不符合命名规范,建议使用 'crawlo:{project_name}:queue:requests' 格式")
162
+
163
+ def _validate_middleware_settings(self, config: Dict[str, Any]):
164
+ """验证中间件设置"""
165
+ # 验证 MIDDLEWARES
166
+ middlewares = config.get('MIDDLEWARES', [])
167
+ if not isinstance(middlewares, list):
168
+ self.errors.append("MIDDLEWARES 必须是列表")
169
+ else:
170
+ for i, middleware in enumerate(middlewares):
171
+ if not isinstance(middleware, str):
172
+ self.errors.append(f"MIDDLEWARES[{i}] 必须是字符串")
173
+
174
+ def _validate_pipeline_settings(self, config: Dict[str, Any]):
175
+ """验证管道设置"""
176
+ # 验证 PIPELINES
177
+ pipelines = config.get('PIPELINES', [])
178
+ if not isinstance(pipelines, list):
179
+ self.errors.append("PIPELINES 必须是列表")
180
+ else:
181
+ for i, pipeline in enumerate(pipelines):
182
+ if not isinstance(pipeline, str):
183
+ self.errors.append(f"PIPELINES[{i}] 必须是字符串")
184
+
185
+ def _validate_extension_settings(self, config: Dict[str, Any]):
186
+ """验证扩展设置"""
187
+ # 验证 EXTENSIONS
188
+ extensions = config.get('EXTENSIONS', [])
189
+ if not isinstance(extensions, list):
190
+ self.errors.append("EXTENSIONS 必须是列表")
191
+ else:
192
+ for i, extension in enumerate(extensions):
193
+ if not isinstance(extension, str):
194
+ self.errors.append(f"EXTENSIONS[{i}] 必须是字符串")
195
+
196
+ def _validate_logging_settings(self, config: Dict[str, Any]):
197
+ """验证日志设置"""
198
+ log_level = config.get('LOG_LEVEL', 'INFO')
199
+ valid_log_levels = ['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL']
200
+ if log_level not in valid_log_levels:
201
+ self.errors.append(f"LOG_LEVEL 必须是以下值之一: {valid_log_levels}")
202
+
203
+ log_file = config.get('LOG_FILE')
204
+ if log_file is not None and not isinstance(log_file, str):
205
+ self.errors.append("LOG_FILE 必须是字符串")
206
+
207
+ def _is_valid_redis_key(self, key: str) -> bool:
208
+ """检查Redis key是否符合命名规范"""
209
+ # 检查是否以 crawlo: 开头
210
+ if not key.startswith('crawlo:'):
211
+ return False
212
+
213
+ # 检查是否包含必要的部分
214
+ parts = key.split(':')
215
+ if len(parts) < 3:
216
+ return False
217
+
218
+ # 检查是否包含 queue 部分
219
+ return 'queue' in parts
220
+
221
+ def get_validation_report(self, config: Dict[str, Any]) -> str:
222
+ """
223
+ 获取验证报告
224
+
225
+ Args:
226
+ config: 配置字典
227
+
228
+ Returns:
229
+ str: 验证报告
230
+ """
231
+ is_valid, errors, warnings = self.validate(config)
232
+
233
+ report = []
234
+ report.append("=" * 50)
235
+ report.append("配置验证报告")
236
+ report.append("=" * 50)
237
+
238
+ if is_valid:
239
+ report.append("配置验证通过")
240
+ else:
241
+ report.append("配置验证失败")
242
+ report.append("错误:")
243
+ for error in errors:
244
+ report.append(f" - {error}")
245
+
246
+ if warnings:
247
+ report.append("警告:")
248
+ for warning in warnings:
249
+ report.append(f" - {warning}")
250
+
251
+ report.append("=" * 50)
252
+ return "\n".join(report)
253
+
254
+
255
+ # 便利函数
256
+ def validate_config(config: Dict[str, Any]) -> Tuple[bool, List[str], List[str]]:
257
+ """
258
+ 验证配置
259
+
260
+ Args:
261
+ config: 配置字典
262
+
263
+ Returns:
264
+ Tuple[bool, List[str], List[str]]: (是否有效, 错误列表, 警告列表)
265
+ """
266
+ validator = ConfigValidator()
267
+ return validator.validate(config)
268
+
269
+
270
+ def print_validation_report(config: Dict[str, Any]):
271
+ """
272
+ 打印验证报告
273
+
274
+ Args:
275
+ config: 配置字典
276
+ """
277
+ validator = ConfigValidator()
278
278
  print(validator.get_validation_report(config))
crawlo/core/__init__.py CHANGED
@@ -1,52 +1,52 @@
1
- #!/usr/bin/python
2
- # -*- coding:UTF-8 -*-
3
-
4
- # Crawlo核心模块
5
- # 提供框架的核心组件和初始化功能
6
-
7
- # 使用新的初始化系统
8
- from ..initialization import (
9
- initialize_framework,
10
- is_framework_ready
11
- )
12
-
13
-
14
- # 向后兼容的别名
15
- def async_initialize_framework(*args, **kwargs):
16
- """Async wrapper for framework initialization"""
17
- return initialize_framework(*args, **kwargs)
18
-
19
-
20
- def get_framework_initializer():
21
- """Get framework initializer - compatibility function"""
22
- from ..initialization.core import CoreInitializer
23
- return CoreInitializer()
24
-
25
-
26
- def get_framework_logger(name='crawlo.core'):
27
- """Get framework logger - compatibility function"""
28
- from ..logging import get_logger
29
- return get_logger(name)
30
-
31
-
32
- # 向后兼容
33
- def bootstrap_framework(*args, **kwargs):
34
- """Bootstrap framework - compatibility function"""
35
- return initialize_framework(*args, **kwargs)
36
-
37
-
38
- def get_bootstrap_manager():
39
- """Get bootstrap manager - compatibility function"""
40
- return get_framework_initializer()
41
-
42
-
43
- __all__ = [
44
- 'initialize_framework',
45
- 'async_initialize_framework',
46
- 'get_framework_initializer',
47
- 'is_framework_ready',
48
- 'get_framework_logger',
49
- # 向后兼容
50
- 'bootstrap_framework',
51
- 'get_bootstrap_manager'
52
- ]
1
+ #!/usr/bin/python
2
+ # -*- coding:UTF-8 -*-
3
+
4
+ # Crawlo核心模块
5
+ # 提供框架的核心组件和初始化功能
6
+
7
+ # 使用新的初始化系统
8
+ from ..initialization import (
9
+ initialize_framework,
10
+ is_framework_ready
11
+ )
12
+
13
+
14
+ # 向后兼容的别名
15
+ def async_initialize_framework(*args, **kwargs):
16
+ """Async wrapper for framework initialization"""
17
+ return initialize_framework(*args, **kwargs)
18
+
19
+
20
+ def get_framework_initializer():
21
+ """Get framework initializer - compatibility function"""
22
+ from ..initialization.core import CoreInitializer
23
+ return CoreInitializer()
24
+
25
+
26
+ def get_framework_logger(name='crawlo.core'):
27
+ """Get framework logger - compatibility function"""
28
+ from ..logging import get_logger
29
+ return get_logger(name)
30
+
31
+
32
+ # 向后兼容
33
+ def bootstrap_framework(*args, **kwargs):
34
+ """Bootstrap framework - compatibility function"""
35
+ return initialize_framework(*args, **kwargs)
36
+
37
+
38
+ def get_bootstrap_manager():
39
+ """Get bootstrap manager - compatibility function"""
40
+ return get_framework_initializer()
41
+
42
+
43
+ __all__ = [
44
+ 'initialize_framework',
45
+ 'async_initialize_framework',
46
+ 'get_framework_initializer',
47
+ 'is_framework_ready',
48
+ 'get_framework_logger',
49
+ # 向后兼容
50
+ 'bootstrap_framework',
51
+ 'get_bootstrap_manager'
52
+ ]