crawlo 1.3.3__py3-none-any.whl → 1.3.5__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 -63
  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 -323
  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 +46 -2
  16. crawlo/core/engine.py +439 -365
  17. crawlo/core/processor.py +40 -40
  18. crawlo/core/scheduler.py +257 -256
  19. crawlo/crawler.py +639 -1167
  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 -226
  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 -52
  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 +28 -0
  40. crawlo/factories/base.py +69 -0
  41. crawlo/factories/crawler.py +104 -0
  42. crawlo/factories/registry.py +85 -0
  43. crawlo/filters/__init__.py +154 -154
  44. crawlo/filters/aioredis_filter.py +257 -234
  45. crawlo/filters/memory_filter.py +269 -269
  46. crawlo/framework.py +292 -0
  47. crawlo/initialization/__init__.py +40 -0
  48. crawlo/initialization/built_in.py +426 -0
  49. crawlo/initialization/context.py +142 -0
  50. crawlo/initialization/core.py +194 -0
  51. crawlo/initialization/phases.py +149 -0
  52. crawlo/initialization/registry.py +146 -0
  53. crawlo/items/__init__.py +23 -23
  54. crawlo/items/base.py +23 -22
  55. crawlo/items/fields.py +52 -52
  56. crawlo/items/items.py +104 -104
  57. crawlo/logging/__init__.py +38 -0
  58. crawlo/logging/config.py +97 -0
  59. crawlo/logging/factory.py +129 -0
  60. crawlo/logging/manager.py +112 -0
  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 -187
  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 -318
  85. crawlo/pipelines/pipeline_manager.py +76 -75
  86. crawlo/pipelines/redis_dedup_pipeline.py +166 -166
  87. crawlo/project.py +327 -325
  88. crawlo/queue/pqueue.py +43 -37
  89. crawlo/queue/queue_manager.py +503 -379
  90. crawlo/queue/redis_priority_queue.py +326 -306
  91. crawlo/settings/__init__.py +7 -7
  92. crawlo/settings/default_settings.py +321 -225
  93. crawlo/settings/setting_manager.py +214 -198
  94. crawlo/spider/__init__.py +657 -639
  95. crawlo/stats_collector.py +73 -59
  96. crawlo/subscriber.py +129 -129
  97. crawlo/task_manager.py +139 -30
  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 +168 -267
  104. crawlo/templates/project/settings_distributed.py.tmpl +167 -180
  105. crawlo/templates/project/settings_gentle.py.tmpl +167 -61
  106. crawlo/templates/project/settings_high_performance.py.tmpl +168 -131
  107. crawlo/templates/project/settings_minimal.py.tmpl +66 -35
  108. crawlo/templates/project/settings_simple.py.tmpl +165 -102
  109. crawlo/templates/project/spiders/__init__.py.tmpl +10 -6
  110. crawlo/templates/run.py.tmpl +34 -38
  111. crawlo/templates/spider/spider.py.tmpl +143 -143
  112. crawlo/templates/spiders_init.py.tmpl +10 -0
  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 +365 -0
  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 +26 -0
  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 -124
  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 +80 -200
  138. crawlo/utils/performance_monitor.py +285 -285
  139. crawlo/utils/queue_helper.py +175 -175
  140. crawlo/utils/redis_connection_pool.py +388 -351
  141. crawlo/utils/redis_key_validator.py +198 -198
  142. crawlo/utils/request.py +267 -267
  143. crawlo/utils/request_serializer.py +225 -218
  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-1.3.3.dist-info → crawlo-1.3.5.dist-info}/METADATA +1126 -1020
  149. crawlo-1.3.5.dist-info/RECORD +288 -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 +107 -107
  154. tests/baidu_performance_test.py +109 -0
  155. tests/baidu_test.py +60 -0
  156. tests/cleaners_example.py +160 -160
  157. tests/comprehensive_framework_test.py +213 -0
  158. tests/comprehensive_test.py +82 -0
  159. tests/comprehensive_testing_summary.md +187 -0
  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 +70 -0
  164. tests/debug_framework_logger.py +85 -0
  165. tests/debug_log_config.py +127 -0
  166. tests/debug_log_levels.py +64 -0
  167. tests/debug_pipelines.py +66 -66
  168. tests/detailed_log_test.py +234 -0
  169. tests/distributed_test.py +67 -0
  170. tests/distributed_test_debug.py +77 -0
  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_command_test_report.md +0 -0
  176. tests/final_comprehensive_test.py +152 -0
  177. tests/final_log_test.py +261 -0
  178. tests/final_validation_test.py +183 -0
  179. tests/fix_log_test.py +143 -0
  180. tests/framework_performance_test.py +203 -0
  181. tests/log_buffering_test.py +112 -0
  182. tests/log_generation_timing_test.py +154 -0
  183. tests/optimized_performance_test.py +212 -0
  184. tests/performance_comparison.py +246 -0
  185. tests/queue_blocking_test.py +114 -0
  186. tests/queue_test.py +90 -0
  187. tests/redis_key_validation_demo.py +130 -130
  188. tests/request_params_example.py +150 -150
  189. tests/response_improvements_example.py +144 -144
  190. tests/scrapy_comparison/ofweek_scrapy.py +139 -0
  191. tests/scrapy_comparison/scrapy_test.py +134 -0
  192. tests/simple_command_test.py +120 -0
  193. tests/simple_crawlo_test.py +128 -0
  194. tests/simple_log_test.py +58 -0
  195. tests/simple_log_test2.py +138 -0
  196. tests/simple_optimization_test.py +129 -0
  197. tests/simple_spider_test.py +50 -0
  198. tests/simple_test.py +48 -0
  199. tests/spider_log_timing_test.py +178 -0
  200. tests/test_advanced_tools.py +148 -148
  201. tests/test_all_commands.py +231 -0
  202. tests/test_all_redis_key_configs.py +145 -145
  203. tests/test_authenticated_proxy.py +141 -141
  204. tests/test_batch_processor.py +179 -0
  205. tests/test_cleaners.py +54 -54
  206. tests/test_component_factory.py +175 -0
  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 +80 -0
  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 +246 -0
  227. tests/test_env_config.py +121 -121
  228. tests/test_error_handler_compatibility.py +112 -112
  229. tests/test_factories.py +253 -0
  230. tests/test_final_validation.py +153 -153
  231. tests/test_framework_env_usage.py +103 -103
  232. tests/test_framework_logger.py +67 -0
  233. tests/test_framework_startup.py +65 -0
  234. tests/test_get_component_logger.py +84 -0
  235. tests/test_integration.py +169 -169
  236. tests/test_item_dedup_redis_key.py +122 -122
  237. tests/test_large_scale_config.py +113 -0
  238. tests/test_large_scale_helper.py +236 -0
  239. tests/test_logging_system.py +283 -0
  240. tests/test_mode_change.py +73 -0
  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 +116 -0
  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 +42 -0
  256. tests/test_queue_manager_double_crawlo.py +173 -173
  257. tests/test_queue_manager_redis_key.py +176 -176
  258. tests/test_random_user_agent.py +72 -72
  259. tests/test_real_scenario_proxy.py +195 -195
  260. tests/test_redis_config.py +28 -28
  261. tests/test_redis_connection_pool.py +294 -294
  262. tests/test_redis_key_naming.py +181 -181
  263. tests/test_redis_key_validator.py +123 -123
  264. tests/test_redis_queue.py +224 -224
  265. tests/test_request_ignore_middleware.py +182 -182
  266. tests/test_request_params.py +111 -111
  267. tests/test_request_serialization.py +70 -70
  268. tests/test_response_code_middleware.py +349 -349
  269. tests/test_response_filter_middleware.py +427 -427
  270. tests/test_response_improvements.py +152 -152
  271. tests/test_retry_middleware.py +241 -241
  272. tests/test_scheduler.py +252 -252
  273. tests/test_scheduler_config_update.py +133 -133
  274. tests/test_simple_response.py +61 -61
  275. tests/test_telecom_spider_redis_key.py +205 -205
  276. tests/test_template_content.py +87 -87
  277. tests/test_template_redis_key.py +134 -134
  278. tests/test_tools.py +159 -159
  279. tests/test_user_agents.py +96 -96
  280. tests/tools_example.py +260 -260
  281. tests/untested_features_report.md +139 -0
  282. tests/verify_debug.py +52 -0
  283. tests/verify_distributed.py +117 -117
  284. tests/verify_log_fix.py +112 -0
  285. crawlo-1.3.3.dist-info/RECORD +0 -219
  286. tests/DOUBLE_CRAWLO_PREFIX_FIX_REPORT.md +0 -82
  287. {crawlo-1.3.3.dist-info → crawlo-1.3.5.dist-info}/WHEEL +0 -0
  288. {crawlo-1.3.3.dist-info → crawlo-1.3.5.dist-info}/entry_points.txt +0 -0
  289. {crawlo-1.3.3.dist-info → crawlo-1.3.5.dist-info}/top_level.txt +0 -0
@@ -0,0 +1,113 @@
1
+ #!/usr/bin/env python3
2
+ # -*- coding: utf-8 -*-
3
+ """
4
+ 大规模配置工具测试
5
+ 测试 LargeScaleConfig, apply_large_scale_config
6
+ """
7
+ import sys
8
+ import os
9
+ import unittest
10
+ from unittest.mock import Mock, patch, MagicMock
11
+
12
+ # 添加项目根目录到 Python 路径
13
+ sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
14
+
15
+ from crawlo.utils.large_scale_config import LargeScaleConfig, apply_large_scale_config
16
+
17
+
18
+ class TestLargeScaleConfig(unittest.TestCase):
19
+ """大规模配置工具测试"""
20
+
21
+ def test_conservative_config(self):
22
+ """测试保守配置"""
23
+ config = LargeScaleConfig.conservative_config(concurrency=8)
24
+
25
+ # 验证基本配置
26
+ self.assertEqual(config['CONCURRENCY'], 8)
27
+ self.assertEqual(config['SCHEDULER_MAX_QUEUE_SIZE'], 80) # 8 * 10
28
+ self.assertEqual(config['DOWNLOAD_DELAY'], 0.2)
29
+ self.assertEqual(config['MAX_RUNNING_SPIDERS'], 1)
30
+
31
+ # 验证连接池配置
32
+ self.assertEqual(config['CONNECTION_POOL_LIMIT'], 16) # 8 * 2
33
+
34
+ # 验证重试配置
35
+ self.assertEqual(config['MAX_RETRY_TIMES'], 2)
36
+
37
+ def test_balanced_config(self):
38
+ """测试平衡配置"""
39
+ config = LargeScaleConfig.balanced_config(concurrency=16)
40
+
41
+ # 验证基本配置
42
+ self.assertEqual(config['CONCURRENCY'], 16)
43
+ self.assertEqual(config['SCHEDULER_MAX_QUEUE_SIZE'], 240) # 16 * 15
44
+ self.assertEqual(config['DOWNLOAD_DELAY'], 0.1)
45
+ self.assertEqual(config['MAX_RUNNING_SPIDERS'], 2)
46
+
47
+ # 验证连接池配置
48
+ self.assertEqual(config['CONNECTION_POOL_LIMIT'], 48) # 16 * 3
49
+
50
+ # 验证重试配置
51
+ self.assertEqual(config['MAX_RETRY_TIMES'], 3)
52
+
53
+ def test_aggressive_config(self):
54
+ """测试激进配置"""
55
+ config = LargeScaleConfig.aggressive_config(concurrency=32)
56
+
57
+ # 验证基本配置
58
+ self.assertEqual(config['CONCURRENCY'], 32)
59
+ self.assertEqual(config['SCHEDULER_MAX_QUEUE_SIZE'], 640) # 32 * 20
60
+ self.assertEqual(config['DOWNLOAD_DELAY'], 0.05)
61
+ self.assertEqual(config['MAX_RUNNING_SPIDERS'], 3)
62
+
63
+ # 验证连接池配置
64
+ self.assertEqual(config['CONNECTION_POOL_LIMIT'], 128) # 32 * 4
65
+
66
+ # 验证重试配置
67
+ self.assertEqual(config['MAX_RETRY_TIMES'], 5)
68
+
69
+ def test_memory_optimized_config(self):
70
+ """测试内存优化配置"""
71
+ config = LargeScaleConfig.memory_optimized_config(concurrency=12)
72
+
73
+ # 验证基本配置
74
+ self.assertEqual(config['CONCURRENCY'], 12)
75
+ self.assertEqual(config['SCHEDULER_MAX_QUEUE_SIZE'], 60) # 12 * 5
76
+ self.assertEqual(config['DOWNLOAD_DELAY'], 0.1)
77
+ self.assertEqual(config['MAX_RUNNING_SPIDERS'], 1)
78
+
79
+ # 验证连接池配置
80
+ self.assertEqual(config['CONNECTION_POOL_LIMIT'], 12) # 12 * 1
81
+
82
+ # 验证内存限制配置
83
+ self.assertEqual(config['DOWNLOAD_MAXSIZE'], 2 * 1024 * 1024) # 2MB
84
+ self.assertEqual(config['DOWNLOAD_WARN_SIZE'], 512 * 1024) # 512KB
85
+
86
+ # 验证重试配置
87
+ self.assertEqual(config['MAX_RETRY_TIMES'], 2)
88
+
89
+ def test_apply_large_scale_config(self):
90
+ """测试应用大规模配置"""
91
+ settings_dict = {}
92
+
93
+ # 应用平衡配置
94
+ apply_large_scale_config(settings_dict, "balanced", 16)
95
+
96
+ # 验证配置已应用
97
+ self.assertEqual(settings_dict['CONCURRENCY'], 16)
98
+ self.assertEqual(settings_dict['SCHEDULER_MAX_QUEUE_SIZE'], 240)
99
+ self.assertEqual(settings_dict['DOWNLOAD_DELAY'], 0.1)
100
+
101
+ def test_apply_large_scale_config_invalid_type(self):
102
+ """测试应用无效的大规模配置类型"""
103
+ settings_dict = {}
104
+
105
+ # 应用无效配置类型应该抛出异常
106
+ with self.assertRaises(ValueError) as context:
107
+ apply_large_scale_config(settings_dict, "invalid_type", 16)
108
+
109
+ self.assertIn("不支持的配置类型", str(context.exception))
110
+
111
+
112
+ if __name__ == '__main__':
113
+ unittest.main()
@@ -0,0 +1,236 @@
1
+ #!/usr/bin/env python3
2
+ # -*- coding: utf-8 -*-
3
+ """
4
+ 大规模爬虫辅助工具测试
5
+ 测试 LargeScaleHelper, ProgressManager, MemoryOptimizer, DataSourceAdapter, LargeScaleSpiderMixin
6
+ """
7
+ import sys
8
+ import os
9
+ import unittest
10
+ from unittest.mock import Mock, patch, MagicMock
11
+ import asyncio
12
+ import json
13
+ import tempfile
14
+
15
+ # 添加项目根目录到 Python 路径
16
+ sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
17
+
18
+ from crawlo.utils.large_scale_helper import (
19
+ LargeScaleHelper,
20
+ ProgressManager,
21
+ MemoryOptimizer,
22
+ DataSourceAdapter,
23
+ LargeScaleSpiderMixin
24
+ )
25
+
26
+
27
+ class TestLargeScaleHelper(unittest.TestCase):
28
+ """大规模爬虫辅助类测试"""
29
+
30
+ def setUp(self):
31
+ """测试前准备"""
32
+ self.helper = LargeScaleHelper(batch_size=100, checkpoint_interval=500)
33
+
34
+ def test_helper_initialization(self):
35
+ """测试辅助类初始化"""
36
+ self.assertEqual(self.helper.batch_size, 100)
37
+ self.assertEqual(self.helper.checkpoint_interval, 500)
38
+
39
+ def test_batch_iterator_with_list(self):
40
+ """测试批次迭代器与列表数据源"""
41
+ data = list(range(250)) # 250个元素
42
+ batches = list(self.helper.batch_iterator(data))
43
+
44
+ # 验证批次数量
45
+ self.assertEqual(len(batches), 3) # 250/100 = 3批次(向上取整)
46
+
47
+ # 验证每个批次的大小
48
+ self.assertEqual(len(batches[0]), 100)
49
+ self.assertEqual(len(batches[1]), 100)
50
+ self.assertEqual(len(batches[2]), 50) # 最后一个批次
51
+
52
+ # 验证数据完整性
53
+ all_data = []
54
+ for batch in batches:
55
+ all_data.extend(batch)
56
+ self.assertEqual(all_data, data)
57
+
58
+ def test_batch_iterator_with_offset(self):
59
+ """测试批次迭代器与偏移量"""
60
+ data = list(range(250)) # 250个元素
61
+ batches = list(self.helper.batch_iterator(data, start_offset=50))
62
+
63
+ # 验证批次数量
64
+ self.assertEqual(len(batches), 2) # 剩余200个元素,2个批次
65
+
66
+ # 验证数据正确性
67
+ all_data = []
68
+ for batch in batches:
69
+ all_data.extend(batch)
70
+ self.assertEqual(all_data, list(range(50, 250)))
71
+
72
+ def test_batch_iterator_invalid_source(self):
73
+ """测试批次迭代器与无效数据源"""
74
+ with self.assertRaises(ValueError) as context:
75
+ list(self.helper.batch_iterator(123)) # 整数不是有效的数据源
76
+ self.assertIn("不支持的数据源类型", str(context.exception))
77
+
78
+
79
+ class TestProgressManager(unittest.TestCase):
80
+ """进度管理器测试"""
81
+
82
+ def setUp(self):
83
+ """测试前准备"""
84
+ # 创建临时文件用于测试
85
+ self.temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.json')
86
+ self.temp_file.close()
87
+ self.progress_manager = ProgressManager(self.temp_file.name)
88
+
89
+ def tearDown(self):
90
+ """测试后清理"""
91
+ try:
92
+ os.unlink(self.temp_file.name)
93
+ except:
94
+ pass
95
+
96
+ def test_progress_manager_initialization(self):
97
+ """测试进度管理器初始化"""
98
+ self.assertEqual(self.progress_manager.progress_file, self.temp_file.name)
99
+
100
+ def test_save_and_load_progress(self):
101
+ """测试保存和加载进度"""
102
+ # 保存进度
103
+ self.progress_manager.save_progress(
104
+ batch_num=10,
105
+ processed_count=1000,
106
+ skipped_count=50
107
+ )
108
+
109
+ # 加载进度
110
+ progress = self.progress_manager.load_progress()
111
+
112
+ # 验证进度数据
113
+ self.assertEqual(progress['batch_num'], 10)
114
+ self.assertEqual(progress['processed_count'], 1000)
115
+ self.assertEqual(progress['skipped_count'], 50)
116
+ self.assertIn('timestamp', progress)
117
+ self.assertIn('formatted_time', progress)
118
+
119
+ def test_load_progress_file_not_found(self):
120
+ """测试加载不存在的进度文件"""
121
+ # 创建一个新的进度管理器与不存在的文件
122
+ non_existent_file = tempfile.gettempdir() + "/non_existent_progress.json"
123
+ pm = ProgressManager(non_existent_file)
124
+
125
+ # 加载进度应该返回默认值
126
+ progress = pm.load_progress()
127
+ self.assertEqual(progress['batch_num'], 0)
128
+ self.assertEqual(progress['processed_count'], 0)
129
+ self.assertEqual(progress['skipped_count'], 0)
130
+
131
+
132
+ class TestMemoryOptimizer(unittest.TestCase):
133
+ """内存优化器测试"""
134
+
135
+ def setUp(self):
136
+ """测试前准备"""
137
+ self.optimizer = MemoryOptimizer(max_memory_mb=100)
138
+
139
+ def test_optimizer_initialization(self):
140
+ """测试内存优化器初始化"""
141
+ self.assertEqual(self.optimizer.max_memory_mb, 100)
142
+
143
+ def test_should_pause_for_memory_without_psutil(self):
144
+ """测试在没有psutil时的内存检查"""
145
+ # 在没有psutil的情况下,should_pause_for_memory应该返回False
146
+ result = self.optimizer.should_pause_for_memory()
147
+ self.assertFalse(result)
148
+
149
+ def test_force_garbage_collection(self):
150
+ """测试强制垃圾回收"""
151
+ # 这个方法应该能正常执行而不抛出异常
152
+ try:
153
+ self.optimizer.force_garbage_collection()
154
+ success = True
155
+ except:
156
+ success = False
157
+ self.assertTrue(success)
158
+
159
+
160
+ class TestDataSourceAdapter(unittest.TestCase):
161
+ """数据源适配器测试"""
162
+
163
+ def test_from_file_adapter(self):
164
+ """测试文件数据源适配器"""
165
+ # 创建临时文件
166
+ with tempfile.NamedTemporaryFile(mode='w', delete=False) as f:
167
+ for i in range(10):
168
+ f.write(f"line {i}\n")
169
+ temp_file_name = f.name
170
+
171
+ try:
172
+ # 创建文件数据源适配器
173
+ adapter = DataSourceAdapter.from_file(temp_file_name, batch_size=5)
174
+
175
+ # 获取第一批数据
176
+ batch = adapter(0, 5)
177
+ self.assertEqual(len(batch), 5)
178
+ self.assertEqual(batch[0], "line 0")
179
+ self.assertEqual(batch[4], "line 4")
180
+
181
+ # 获取第二批数据
182
+ batch = adapter(5, 5)
183
+ self.assertEqual(len(batch), 5)
184
+ self.assertEqual(batch[0], "line 5")
185
+ self.assertEqual(batch[4], "line 9")
186
+
187
+ finally:
188
+ # 清理临时文件
189
+ os.unlink(temp_file_name)
190
+
191
+
192
+ class TestLargeScaleSpiderMixin(unittest.TestCase):
193
+ """大规模爬虫混入类测试"""
194
+
195
+ def test_mixin_initialization(self):
196
+ """测试混入类初始化"""
197
+ # 创建一个模拟的爬虫类
198
+ class MockSpider:
199
+ def __init__(self):
200
+ self.name = "test_spider"
201
+
202
+ class TestSpider(MockSpider, LargeScaleSpiderMixin):
203
+ def __init__(self):
204
+ MockSpider.__init__(self)
205
+ LargeScaleSpiderMixin.__init__(self)
206
+
207
+ spider = TestSpider()
208
+
209
+ # 验证初始化
210
+ self.assertEqual(spider.name, "test_spider")
211
+ self.assertIsNotNone(spider.large_scale_helper)
212
+ self.assertIsNotNone(spider.progress_manager)
213
+ self.assertIsNotNone(spider.memory_optimizer)
214
+
215
+ def test_mixin_attributes(self):
216
+ """测试混入类属性"""
217
+ # 创建一个模拟的爬虫类
218
+ class MockSpider:
219
+ def __init__(self):
220
+ self.name = "test_spider"
221
+
222
+ class TestSpider(MockSpider, LargeScaleSpiderMixin):
223
+ def __init__(self):
224
+ MockSpider.__init__(self)
225
+ LargeScaleSpiderMixin.__init__(self)
226
+
227
+ spider = TestSpider()
228
+
229
+ # 验证属性
230
+ self.assertIsInstance(spider.large_scale_helper, LargeScaleHelper)
231
+ self.assertIsInstance(spider.progress_manager, ProgressManager)
232
+ self.assertIsInstance(spider.memory_optimizer, MemoryOptimizer)
233
+
234
+
235
+ if __name__ == '__main__':
236
+ unittest.main()
@@ -0,0 +1,283 @@
1
+ #!/usr/bin/env python
2
+ # -*- coding: UTF-8 -*-
3
+ """
4
+ 全面测试Crawlo框架日志系统的所有功能
5
+ """
6
+
7
+ import os
8
+ import sys
9
+ import tempfile
10
+ import shutil
11
+ from pathlib import Path
12
+
13
+ # 添加项目根目录到Python路径
14
+ project_root = Path(__file__).parent.parent
15
+ sys.path.insert(0, str(project_root))
16
+
17
+ from crawlo.logging import configure_logging as configure, get_logger, LogManager
18
+ from crawlo.logging.config import LogConfig
19
+
20
+
21
+ def test_basic_logging():
22
+ """测试基本日志功能"""
23
+ print("=== 测试基本日志功能 ===")
24
+
25
+ # 1. 测试默认配置
26
+ print("1. 测试默认配置...")
27
+ configure()
28
+ config = LogManager().config
29
+ print(f" 默认级别: {config.level}")
30
+ print(f" 默认格式: {config.format}")
31
+
32
+ # 2. 测试获取logger
33
+ logger = get_logger('test.basic')
34
+ print(f" Logger名称: {logger.name}")
35
+ print(f" Handlers数量: {len(logger.handlers)}")
36
+
37
+ # 3. 测试日志输出
38
+ logger.info("这是一条INFO级别日志")
39
+ logger.warning("这是一条WARNING级别日志")
40
+ logger.error("这是一条ERROR级别日志")
41
+ print(" 日志输出完成")
42
+
43
+
44
+ def test_file_logging():
45
+ """测试文件日志功能"""
46
+ print("\n=== 测试文件日志功能 ===")
47
+
48
+ # 创建临时目录
49
+ temp_dir = tempfile.mkdtemp()
50
+ log_file = os.path.join(temp_dir, 'test.log')
51
+
52
+ try:
53
+ # 1. 配置文件日志
54
+ print("1. 配置文件日志...")
55
+ configure(
56
+ LOG_LEVEL='DEBUG',
57
+ LOG_FILE=log_file,
58
+ LOG_MAX_BYTES=1024, # 小的轮转大小用于测试
59
+ LOG_BACKUP_COUNT=3
60
+ )
61
+
62
+ config = LogManager().config
63
+ print(f" 日志文件: {config.file_path}")
64
+ print(f" 轮转大小: {config.max_bytes}")
65
+ print(f" 备份数量: {config.backup_count}")
66
+
67
+ # 2. 获取logger并测试输出
68
+ logger = get_logger('test.file')
69
+ print("2. 测试日志输出...")
70
+
71
+ # 输出多条日志以触发轮转
72
+ for i in range(50):
73
+ logger.debug(f"这是第{i+1}条DEBUG日志,用于测试轮转功能")
74
+ logger.info(f"这是第{i+1}条INFO日志,用于测试轮转功能")
75
+
76
+ # 3. 检查日志文件
77
+ print("3. 检查日志文件...")
78
+ if os.path.exists(log_file):
79
+ file_size = os.path.getsize(log_file)
80
+ print(f" 主日志文件大小: {file_size} 字节")
81
+ else:
82
+ print(" 主日志文件不存在!")
83
+
84
+ # 检查备份文件
85
+ for i in range(1, 4):
86
+ backup_file = f"{log_file}.{i}"
87
+ if os.path.exists(backup_file):
88
+ backup_size = os.path.getsize(backup_file)
89
+ print(f" 备份文件 {i} 大小: {backup_size} 字节")
90
+ else:
91
+ print(f" 备份文件 {i} 不存在")
92
+
93
+ finally:
94
+ # 清理临时目录
95
+ shutil.rmtree(temp_dir, ignore_errors=True)
96
+
97
+
98
+ def test_console_logging():
99
+ """测试控制台日志功能"""
100
+ print("\n=== 测试控制台日志功能 ===")
101
+
102
+ # 1. 配置仅控制台日志
103
+ print("1. 配置仅控制台日志...")
104
+ configure(
105
+ LOG_LEVEL='INFO',
106
+ LOG_CONSOLE_ENABLED=True,
107
+ LOG_FILE_ENABLED=False
108
+ )
109
+
110
+ config = LogManager().config
111
+ print(f" 控制台启用: {config.console_enabled}")
112
+ print(f" 文件日志启用: {config.file_enabled}")
113
+
114
+ # 2. 测试日志输出
115
+ logger = get_logger('test.console')
116
+ print("2. 测试日志输出...")
117
+ logger.info("这条日志应该只在控制台显示")
118
+ logger.debug("这条DEBUG日志不应该显示(级别低于INFO)")
119
+
120
+
121
+ def test_module_level_logging():
122
+ """测试模块级别日志功能"""
123
+ print("\n=== 测试模块级别日志功能 ===")
124
+
125
+ # 1. 配置模块级别
126
+ print("1. 配置模块级别...")
127
+ configure(
128
+ LOG_LEVEL='WARNING',
129
+ LOG_LEVELS={
130
+ 'test.module.high': 'DEBUG',
131
+ 'test.module.low': 'ERROR'
132
+ }
133
+ )
134
+
135
+ config = LogManager().config
136
+ print(f" 默认级别: {config.level}")
137
+ print(f" 模块级别配置: {config.module_levels}")
138
+
139
+ # 2. 测试不同模块的日志级别
140
+ print("2. 测试不同模块的日志级别...")
141
+
142
+ # 高级别模块(DEBUG)
143
+ high_logger = get_logger('test.module.high')
144
+ print(f" high_logger 级别: {logging.getLevelName(high_logger.handlers[0].level)}")
145
+ high_logger.debug("这条DEBUG日志应该显示")
146
+ high_logger.info("这条INFO日志应该显示")
147
+
148
+ # 低级别模块(ERROR)
149
+ low_logger = get_logger('test.module.low')
150
+ print(f" low_logger 级别: {logging.getLevelName(low_logger.handlers[0].level)}")
151
+ low_logger.info("这条INFO日志不应该显示")
152
+ low_logger.error("这条ERROR日志应该显示")
153
+
154
+ # 默认级别模块(WARNING)
155
+ default_logger = get_logger('test.module.default')
156
+ print(f" default_logger 级别: {logging.getLevelName(default_logger.handlers[0].level)}")
157
+ default_logger.info("这条INFO日志不应该显示")
158
+ default_logger.warning("这条WARNING日志应该显示")
159
+
160
+
161
+ def test_log_config_from_settings():
162
+ """测试从settings对象创建配置"""
163
+ print("\n=== 测试从settings对象创建配置 ===")
164
+
165
+ # 1. 创建模拟settings对象
166
+ class MockSettings:
167
+ def __init__(self):
168
+ self.LOG_LEVEL = 'DEBUG'
169
+ self.LOG_FILE = 'test_settings.log'
170
+ self.LOG_MAX_BYTES = 5 * 1024 * 1024
171
+ self.LOG_BACKUP_COUNT = 2
172
+ self.LOG_ENCODING = 'utf-8'
173
+ self.LOG_CONSOLE_ENABLED = True
174
+ self.LOG_FILE_ENABLED = True
175
+ self.LOG_LEVELS = {'test.custom': 'INFO'}
176
+
177
+ # 2. 从settings创建配置
178
+ print("2. 从settings创建配置...")
179
+ settings = MockSettings()
180
+ config = LogConfig.from_settings(settings)
181
+
182
+ print(f" 级别: {config.level}")
183
+ print(f" 文件: {config.file_path}")
184
+ print(f" 轮转大小: {config.max_bytes}")
185
+ print(f" 备份数量: {config.backup_count}")
186
+ print(f" 编码: {config.encoding}")
187
+ print(f" 控制台启用: {config.console_enabled}")
188
+ print(f" 文件启用: {config.file_enabled}")
189
+ print(f" 模块级别: {config.module_levels}")
190
+
191
+
192
+ def test_log_config_validation():
193
+ """测试日志配置验证功能"""
194
+ print("\n=== 测试日志配置验证功能 ===")
195
+
196
+ # 1. 测试有效配置
197
+ print("1. 测试有效配置...")
198
+ valid_config = LogConfig(
199
+ level='INFO',
200
+ file_path='test_valid.log'
201
+ )
202
+ is_valid = valid_config.validate()
203
+ print(f" 有效配置验证结果: {is_valid}")
204
+
205
+ # 2. 测试无效级别
206
+ print("2. 测试无效级别...")
207
+ invalid_config = LogConfig(
208
+ level='INVALID',
209
+ file_path='test_invalid.log'
210
+ )
211
+ is_valid = invalid_config.validate()
212
+ print(f" 无效级别验证结果: {is_valid}")
213
+
214
+ # 3. 测试目录创建
215
+ print("3. 测试目录创建...")
216
+ temp_dir = tempfile.mkdtemp()
217
+ nested_log_file = os.path.join(temp_dir, 'subdir', 'test.log')
218
+
219
+ config = LogConfig(
220
+ level='INFO',
221
+ file_path=nested_log_file
222
+ )
223
+ is_valid = config.validate()
224
+ print(f" 目录创建验证结果: {is_valid}")
225
+
226
+ # 清理
227
+ shutil.rmtree(temp_dir, ignore_errors=True)
228
+
229
+
230
+ def test_logger_caching():
231
+ """测试Logger缓存功能"""
232
+ print("\n=== 测试Logger缓存功能 ===")
233
+
234
+ # 重置配置以确保测试环境干净
235
+ LogManager().reset()
236
+
237
+ # 1. 获取同一个logger多次
238
+ print("1. 测试缓存...")
239
+ logger1 = get_logger('test.cache')
240
+ logger2 = get_logger('test.cache')
241
+
242
+ print(f" 第一次获取: {id(logger1)}")
243
+ print(f" 第二次获取: {id(logger2)}")
244
+ print(f" 是否为同一对象: {logger1 is logger2}")
245
+
246
+ # 2. 测试配置更新后缓存刷新
247
+ print("2. 测试配置更新...")
248
+ configure(LOG_LEVEL='ERROR')
249
+ logger3 = get_logger('test.cache')
250
+ print(f" 更新配置后获取: {id(logger3)}")
251
+ print(f" 是否与之前相同: {logger1 is logger3}")
252
+
253
+
254
+ def main():
255
+ """主测试函数"""
256
+ print("开始全面测试Crawlo框架日志系统...")
257
+
258
+ try:
259
+ # 运行所有测试
260
+ test_basic_logging()
261
+ test_file_logging()
262
+ test_console_logging()
263
+ test_module_level_logging()
264
+ test_log_config_from_settings()
265
+ test_log_config_validation()
266
+ test_logger_caching()
267
+
268
+ print("\n=== 所有测试完成 ===")
269
+
270
+ except Exception as e:
271
+ print(f"\n测试过程中出现错误: {e}")
272
+ import traceback
273
+ traceback.print_exc()
274
+ return 1
275
+
276
+ return 0
277
+
278
+
279
+ # 导入logging模块用于测试
280
+ import logging
281
+
282
+ if __name__ == '__main__':
283
+ sys.exit(main())
@@ -0,0 +1,73 @@
1
+ #!/usr/bin/env python
2
+ # -*- coding: utf-8 -*-
3
+ """
4
+ 测试运行模式日志级别修改
5
+ """
6
+ import sys
7
+ import os
8
+ sys.path.insert(0, '/')
9
+
10
+ def test_mode_log_level():
11
+ print("=== 测试运行模式日志级别修改 ===")
12
+
13
+ # 删除旧的日志文件
14
+ test_log_file = '/Users/oscar/projects/Crawlo/test_mode_change.log'
15
+ if os.path.exists(test_log_file):
16
+ os.remove(test_log_file)
17
+
18
+ # 准备测试设置
19
+ test_settings = {
20
+ 'PROJECT_NAME': 'test_mode_change',
21
+ 'LOG_LEVEL': 'INFO',
22
+ 'LOG_FILE': test_log_file,
23
+ 'RUN_MODE': 'standalone'
24
+ }
25
+
26
+ try:
27
+ # 初始化框架
28
+ from crawlo.core.framework_initializer import initialize_framework
29
+ settings = initialize_framework(test_settings)
30
+
31
+ print(f"设置初始化完成: {settings.get('PROJECT_NAME')}")
32
+
33
+ # 检查日志文件是否包含运行模式信息
34
+ if os.path.exists(test_log_file):
35
+ with open(test_log_file, 'r', encoding='utf-8') as f:
36
+ content = f.read()
37
+ print(f"日志文件内容长度: {len(content)} 字符")
38
+
39
+ # 检查是否还有INFO级别的运行模式信息
40
+ info_lines = [line for line in content.split('\n') if 'INFO' in line and '使用单机模式' in line]
41
+ debug_lines = [line for line in content.split('\n') if 'DEBUG' in line and '使用单机模式' in line]
42
+
43
+ if info_lines:
44
+ print("❌ 仍然发现INFO级别的运行模式信息:")
45
+ for line in info_lines:
46
+ print(f" {line}")
47
+ else:
48
+ print("✅ 没有发现INFO级别的运行模式信息")
49
+
50
+ if debug_lines:
51
+ print("✅ 发现DEBUG级别的运行模式信息:")
52
+ for line in debug_lines:
53
+ print(f" {line}")
54
+ else:
55
+ print("❌ 没有发现DEBUG级别的运行模式信息")
56
+
57
+ print("\n所有日志内容:")
58
+ lines = content.split('\n')
59
+ for i, line in enumerate(lines, 1):
60
+ if line.strip():
61
+ print(f"{i:3d}: {line}")
62
+ else:
63
+ print("❌ 日志文件未创建")
64
+
65
+ except Exception as e:
66
+ print(f"错误: {e}")
67
+ import traceback
68
+ traceback.print_exc()
69
+
70
+ print("=== 测试完成 ===")
71
+
72
+ if __name__ == "__main__":
73
+ test_mode_log_level()