crawlo 1.2.3__py3-none-any.whl → 1.2.4__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 (222) hide show
  1. crawlo/__init__.py +61 -61
  2. crawlo/__version__.py +1 -1
  3. crawlo/cleaners/__init__.py +60 -60
  4. crawlo/cleaners/data_formatter.py +225 -225
  5. crawlo/cleaners/encoding_converter.py +125 -125
  6. crawlo/cleaners/text_cleaner.py +232 -232
  7. crawlo/cli.py +81 -81
  8. crawlo/commands/__init__.py +14 -14
  9. crawlo/commands/check.py +594 -594
  10. crawlo/commands/genspider.py +151 -151
  11. crawlo/commands/help.py +144 -142
  12. crawlo/commands/list.py +155 -155
  13. crawlo/commands/run.py +323 -292
  14. crawlo/commands/startproject.py +420 -417
  15. crawlo/commands/stats.py +187 -187
  16. crawlo/commands/utils.py +186 -186
  17. crawlo/config.py +312 -312
  18. crawlo/config_validator.py +251 -251
  19. crawlo/core/__init__.py +2 -2
  20. crawlo/core/engine.py +354 -354
  21. crawlo/core/processor.py +40 -40
  22. crawlo/core/scheduler.py +143 -143
  23. crawlo/crawler.py +1110 -1027
  24. crawlo/data/__init__.py +5 -5
  25. crawlo/data/user_agents.py +107 -107
  26. crawlo/downloader/__init__.py +266 -266
  27. crawlo/downloader/aiohttp_downloader.py +220 -220
  28. crawlo/downloader/cffi_downloader.py +256 -256
  29. crawlo/downloader/httpx_downloader.py +259 -259
  30. crawlo/downloader/hybrid_downloader.py +212 -212
  31. crawlo/downloader/playwright_downloader.py +402 -402
  32. crawlo/downloader/selenium_downloader.py +472 -472
  33. crawlo/event.py +11 -11
  34. crawlo/exceptions.py +81 -81
  35. crawlo/extension/__init__.py +37 -37
  36. crawlo/extension/health_check.py +141 -141
  37. crawlo/extension/log_interval.py +57 -57
  38. crawlo/extension/log_stats.py +81 -81
  39. crawlo/extension/logging_extension.py +43 -43
  40. crawlo/extension/memory_monitor.py +104 -104
  41. crawlo/extension/performance_profiler.py +133 -133
  42. crawlo/extension/request_recorder.py +107 -107
  43. crawlo/filters/__init__.py +154 -154
  44. crawlo/filters/aioredis_filter.py +280 -280
  45. crawlo/filters/memory_filter.py +269 -269
  46. crawlo/items/__init__.py +23 -23
  47. crawlo/items/base.py +21 -21
  48. crawlo/items/fields.py +52 -52
  49. crawlo/items/items.py +104 -104
  50. crawlo/middleware/__init__.py +21 -21
  51. crawlo/middleware/default_header.py +131 -131
  52. crawlo/middleware/download_delay.py +104 -104
  53. crawlo/middleware/middleware_manager.py +135 -135
  54. crawlo/middleware/offsite.py +114 -114
  55. crawlo/middleware/proxy.py +367 -367
  56. crawlo/middleware/request_ignore.py +86 -86
  57. crawlo/middleware/response_code.py +163 -163
  58. crawlo/middleware/response_filter.py +136 -136
  59. crawlo/middleware/retry.py +124 -124
  60. crawlo/mode_manager.py +211 -211
  61. crawlo/network/__init__.py +21 -21
  62. crawlo/network/request.py +338 -338
  63. crawlo/network/response.py +359 -359
  64. crawlo/pipelines/__init__.py +21 -21
  65. crawlo/pipelines/bloom_dedup_pipeline.py +156 -156
  66. crawlo/pipelines/console_pipeline.py +39 -39
  67. crawlo/pipelines/csv_pipeline.py +316 -316
  68. crawlo/pipelines/database_dedup_pipeline.py +222 -222
  69. crawlo/pipelines/json_pipeline.py +218 -218
  70. crawlo/pipelines/memory_dedup_pipeline.py +115 -115
  71. crawlo/pipelines/mongo_pipeline.py +131 -131
  72. crawlo/pipelines/mysql_pipeline.py +317 -317
  73. crawlo/pipelines/pipeline_manager.py +61 -61
  74. crawlo/pipelines/redis_dedup_pipeline.py +165 -165
  75. crawlo/project.py +279 -187
  76. crawlo/queue/pqueue.py +37 -37
  77. crawlo/queue/queue_manager.py +337 -337
  78. crawlo/queue/redis_priority_queue.py +298 -298
  79. crawlo/settings/__init__.py +7 -7
  80. crawlo/settings/default_settings.py +217 -226
  81. crawlo/settings/setting_manager.py +122 -122
  82. crawlo/spider/__init__.py +639 -639
  83. crawlo/stats_collector.py +59 -59
  84. crawlo/subscriber.py +129 -129
  85. crawlo/task_manager.py +30 -30
  86. crawlo/templates/crawlo.cfg.tmpl +10 -10
  87. crawlo/templates/project/__init__.py.tmpl +3 -3
  88. crawlo/templates/project/items.py.tmpl +17 -17
  89. crawlo/templates/project/middlewares.py.tmpl +118 -118
  90. crawlo/templates/project/pipelines.py.tmpl +96 -96
  91. crawlo/templates/project/run.py.tmpl +47 -45
  92. crawlo/templates/project/settings.py.tmpl +350 -325
  93. crawlo/templates/project/settings_distributed.py.tmpl +160 -121
  94. crawlo/templates/project/settings_gentle.py.tmpl +133 -94
  95. crawlo/templates/project/settings_high_performance.py.tmpl +155 -151
  96. crawlo/templates/project/settings_simple.py.tmpl +108 -68
  97. crawlo/templates/project/spiders/__init__.py.tmpl +5 -5
  98. crawlo/templates/spider/spider.py.tmpl +143 -143
  99. crawlo/tools/__init__.py +182 -182
  100. crawlo/tools/anti_crawler.py +268 -268
  101. crawlo/tools/authenticated_proxy.py +240 -240
  102. crawlo/tools/data_validator.py +180 -180
  103. crawlo/tools/date_tools.py +35 -35
  104. crawlo/tools/distributed_coordinator.py +386 -386
  105. crawlo/tools/retry_mechanism.py +220 -220
  106. crawlo/tools/scenario_adapter.py +262 -262
  107. crawlo/utils/__init__.py +35 -35
  108. crawlo/utils/batch_processor.py +259 -259
  109. crawlo/utils/controlled_spider_mixin.py +439 -439
  110. crawlo/utils/date_tools.py +290 -290
  111. crawlo/utils/db_helper.py +343 -343
  112. crawlo/utils/enhanced_error_handler.py +356 -356
  113. crawlo/utils/env_config.py +105 -105
  114. crawlo/utils/error_handler.py +123 -123
  115. crawlo/utils/func_tools.py +82 -82
  116. crawlo/utils/large_scale_config.py +286 -286
  117. crawlo/utils/large_scale_helper.py +344 -344
  118. crawlo/utils/log.py +128 -128
  119. crawlo/utils/performance_monitor.py +285 -285
  120. crawlo/utils/queue_helper.py +175 -175
  121. crawlo/utils/redis_connection_pool.py +334 -334
  122. crawlo/utils/redis_key_validator.py +198 -198
  123. crawlo/utils/request.py +267 -267
  124. crawlo/utils/request_serializer.py +218 -218
  125. crawlo/utils/spider_loader.py +61 -61
  126. crawlo/utils/system.py +11 -11
  127. crawlo/utils/tools.py +4 -4
  128. crawlo/utils/url.py +39 -39
  129. {crawlo-1.2.3.dist-info → crawlo-1.2.4.dist-info}/METADATA +764 -692
  130. crawlo-1.2.4.dist-info/RECORD +206 -0
  131. examples/__init__.py +7 -7
  132. tests/DOUBLE_CRAWLO_PREFIX_FIX_REPORT.md +81 -81
  133. tests/__init__.py +7 -7
  134. tests/advanced_tools_example.py +275 -275
  135. tests/authenticated_proxy_example.py +236 -236
  136. tests/cleaners_example.py +160 -160
  137. tests/config_validation_demo.py +102 -102
  138. tests/controlled_spider_example.py +205 -205
  139. tests/date_tools_example.py +180 -180
  140. tests/dynamic_loading_example.py +523 -523
  141. tests/dynamic_loading_test.py +104 -104
  142. tests/env_config_example.py +133 -133
  143. tests/error_handling_example.py +171 -171
  144. tests/redis_key_validation_demo.py +130 -130
  145. tests/response_improvements_example.py +144 -144
  146. tests/test_advanced_tools.py +148 -148
  147. tests/test_all_redis_key_configs.py +145 -145
  148. tests/test_authenticated_proxy.py +141 -141
  149. tests/test_cleaners.py +54 -54
  150. tests/test_comprehensive.py +146 -146
  151. tests/test_config_validator.py +193 -193
  152. tests/test_crawlo_proxy_integration.py +172 -172
  153. tests/test_date_tools.py +123 -123
  154. tests/test_default_header_middleware.py +158 -158
  155. tests/test_double_crawlo_fix.py +207 -207
  156. tests/test_double_crawlo_fix_simple.py +124 -124
  157. tests/test_download_delay_middleware.py +221 -221
  158. tests/test_downloader_proxy_compatibility.py +268 -268
  159. tests/test_dynamic_downloaders_proxy.py +124 -124
  160. tests/test_dynamic_proxy.py +92 -92
  161. tests/test_dynamic_proxy_config.py +146 -146
  162. tests/test_dynamic_proxy_real.py +109 -109
  163. tests/test_edge_cases.py +303 -303
  164. tests/test_enhanced_error_handler.py +270 -270
  165. tests/test_env_config.py +121 -121
  166. tests/test_error_handler_compatibility.py +112 -112
  167. tests/test_final_validation.py +153 -153
  168. tests/test_framework_env_usage.py +103 -103
  169. tests/test_integration.py +356 -356
  170. tests/test_item_dedup_redis_key.py +122 -122
  171. tests/test_offsite_middleware.py +221 -221
  172. tests/test_parsel.py +29 -29
  173. tests/test_performance.py +327 -327
  174. tests/test_proxy_api.py +264 -264
  175. tests/test_proxy_health_check.py +32 -32
  176. tests/test_proxy_middleware.py +121 -121
  177. tests/test_proxy_middleware_enhanced.py +216 -216
  178. tests/test_proxy_middleware_integration.py +136 -136
  179. tests/test_proxy_providers.py +56 -56
  180. tests/test_proxy_stats.py +19 -19
  181. tests/test_proxy_strategies.py +59 -59
  182. tests/test_queue_manager_double_crawlo.py +173 -173
  183. tests/test_queue_manager_redis_key.py +176 -176
  184. tests/test_real_scenario_proxy.py +195 -195
  185. tests/test_redis_config.py +28 -28
  186. tests/test_redis_connection_pool.py +294 -294
  187. tests/test_redis_key_naming.py +181 -181
  188. tests/test_redis_key_validator.py +123 -123
  189. tests/test_redis_queue.py +224 -224
  190. tests/test_request_ignore_middleware.py +182 -182
  191. tests/test_request_serialization.py +70 -70
  192. tests/test_response_code_middleware.py +349 -349
  193. tests/test_response_filter_middleware.py +427 -427
  194. tests/test_response_improvements.py +152 -152
  195. tests/test_retry_middleware.py +241 -241
  196. tests/test_scheduler.py +241 -241
  197. tests/test_simple_response.py +61 -61
  198. tests/test_telecom_spider_redis_key.py +205 -205
  199. tests/test_template_content.py +87 -87
  200. tests/test_template_redis_key.py +134 -134
  201. tests/test_tools.py +153 -153
  202. tests/tools_example.py +257 -257
  203. crawlo-1.2.3.dist-info/RECORD +0 -222
  204. examples/aiohttp_settings.py +0 -42
  205. examples/curl_cffi_settings.py +0 -41
  206. examples/default_header_middleware_example.py +0 -107
  207. examples/default_header_spider_example.py +0 -129
  208. examples/download_delay_middleware_example.py +0 -160
  209. examples/httpx_settings.py +0 -42
  210. examples/multi_downloader_proxy_example.py +0 -81
  211. examples/offsite_middleware_example.py +0 -55
  212. examples/offsite_spider_example.py +0 -107
  213. examples/proxy_spider_example.py +0 -166
  214. examples/request_ignore_middleware_example.py +0 -51
  215. examples/request_ignore_spider_example.py +0 -99
  216. examples/response_code_middleware_example.py +0 -52
  217. examples/response_filter_middleware_example.py +0 -67
  218. examples/tong_hua_shun_settings.py +0 -62
  219. examples/tong_hua_shun_spider.py +0 -170
  220. {crawlo-1.2.3.dist-info → crawlo-1.2.4.dist-info}/WHEEL +0 -0
  221. {crawlo-1.2.3.dist-info → crawlo-1.2.4.dist-info}/entry_points.txt +0 -0
  222. {crawlo-1.2.3.dist-info → crawlo-1.2.4.dist-info}/top_level.txt +0 -0
@@ -1,123 +1,123 @@
1
- #!/usr/bin/python
2
- # -*- coding: UTF-8 -*-
3
- import json
4
- from copy import deepcopy
5
- from importlib import import_module
6
- from collections.abc import MutableMapping
7
-
8
- from crawlo.settings import default_settings
9
-
10
-
11
- class SettingManager(MutableMapping):
12
-
13
- def __init__(self, values=None):
14
- self.attributes = {}
15
- self.set_settings(default_settings)
16
- self.update_attributes(values)
17
-
18
- def get(self, key, default=None):
19
- """安全获取值,不触发递归"""
20
- value = self.attributes.get(key, default)
21
- return value if value is not None else default
22
-
23
- def get_int(self, key, default=0):
24
- return int(self.get(key, default=default))
25
-
26
- def get_float(self, key, default=0.0):
27
- return float(self.get(key, default=default))
28
-
29
- def get_bool(self, key, default=False):
30
- got = self.get(key, default=default)
31
- if isinstance(got, bool):
32
- return got
33
- if isinstance(got, (int, float)):
34
- return bool(got)
35
- got_lower = str(got).strip().lower()
36
- if got_lower in ('1', 'true'):
37
- return True
38
- if got_lower in ('0', 'false'):
39
- return False
40
- raise ValueError(
41
- f"Unsupported value for boolean setting: {got}. "
42
- "Supported values are: 0/1, True/False, '0'/'1', 'True'/'False' (case-insensitive)."
43
- )
44
-
45
- def get_list(self, key, default=None):
46
- values = self.get(key, default or [])
47
- if isinstance(values, str):
48
- return [v.strip() for v in values.split(',') if v.strip()]
49
- try:
50
- return list(values)
51
- except TypeError:
52
- return [values]
53
-
54
- def get_dict(self, key, default=None):
55
- value = self.get(key, default or {})
56
- if isinstance(value, str):
57
- value = json.loads(value)
58
- try:
59
- return dict(value)
60
- except TypeError:
61
- return value
62
-
63
- def set(self, key, value):
64
- self.attributes[key] = value
65
-
66
- def set_settings(self, module):
67
- if isinstance(module, str):
68
- module = import_module(module)
69
- for key in dir(module):
70
- if key.isupper():
71
- self.set(key, getattr(module, key))
72
-
73
- # 实现 MutableMapping 必须的方法
74
- def __getitem__(self, item):
75
- return self.attributes[item]
76
-
77
- def __setitem__(self, key, value):
78
- self.set(key, value)
79
-
80
- def __delitem__(self, key):
81
- del self.attributes[key]
82
-
83
- def __iter__(self):
84
- return iter(self.attributes)
85
-
86
- def __len__(self):
87
- return len(self.attributes)
88
-
89
- def __str__(self):
90
- return f'<Settings: {self.attributes}>'
91
-
92
- __repr__ = __str__
93
-
94
- def update_attributes(self, attributes):
95
- if attributes is not None:
96
- for key, value in attributes.items():
97
- self.set(key, value)
98
-
99
- def copy(self):
100
- return deepcopy(self)
101
-
102
- def __deepcopy__(self, memo):
103
- """
104
- 自定义深度复制方法,避免复制logger等不可pickle的对象
105
- """
106
- # 创建一个新的实例
107
- cls = self.__class__
108
- new_instance = cls.__new__(cls)
109
-
110
- # 复制attributes字典,但排除不可pickle的对象
111
- new_attributes = {}
112
- for key, value in self.attributes.items():
113
- try:
114
- # 尝试深度复制值
115
- new_attributes[key] = deepcopy(value, memo)
116
- except Exception:
117
- # 如果复制失败,保留原始引用(对于logger等对象)
118
- new_attributes[key] = value
119
-
120
- # 设置新实例的attributes
121
- new_instance.attributes = new_attributes
122
-
1
+ #!/usr/bin/python
2
+ # -*- coding: UTF-8 -*-
3
+ import json
4
+ from copy import deepcopy
5
+ from importlib import import_module
6
+ from collections.abc import MutableMapping
7
+
8
+ from crawlo.settings import default_settings
9
+
10
+
11
+ class SettingManager(MutableMapping):
12
+
13
+ def __init__(self, values=None):
14
+ self.attributes = {}
15
+ self.set_settings(default_settings)
16
+ self.update_attributes(values)
17
+
18
+ def get(self, key, default=None):
19
+ """安全获取值,不触发递归"""
20
+ value = self.attributes.get(key, default)
21
+ return value if value is not None else default
22
+
23
+ def get_int(self, key, default=0):
24
+ return int(self.get(key, default=default))
25
+
26
+ def get_float(self, key, default=0.0):
27
+ return float(self.get(key, default=default))
28
+
29
+ def get_bool(self, key, default=False):
30
+ got = self.get(key, default=default)
31
+ if isinstance(got, bool):
32
+ return got
33
+ if isinstance(got, (int, float)):
34
+ return bool(got)
35
+ got_lower = str(got).strip().lower()
36
+ if got_lower in ('1', 'true'):
37
+ return True
38
+ if got_lower in ('0', 'false'):
39
+ return False
40
+ raise ValueError(
41
+ f"Unsupported value for boolean setting: {got}. "
42
+ "Supported values are: 0/1, True/False, '0'/'1', 'True'/'False' (case-insensitive)."
43
+ )
44
+
45
+ def get_list(self, key, default=None):
46
+ values = self.get(key, default or [])
47
+ if isinstance(values, str):
48
+ return [v.strip() for v in values.split(',') if v.strip()]
49
+ try:
50
+ return list(values)
51
+ except TypeError:
52
+ return [values]
53
+
54
+ def get_dict(self, key, default=None):
55
+ value = self.get(key, default or {})
56
+ if isinstance(value, str):
57
+ value = json.loads(value)
58
+ try:
59
+ return dict(value)
60
+ except TypeError:
61
+ return value
62
+
63
+ def set(self, key, value):
64
+ self.attributes[key] = value
65
+
66
+ def set_settings(self, module):
67
+ if isinstance(module, str):
68
+ module = import_module(module)
69
+ for key in dir(module):
70
+ if key.isupper():
71
+ self.set(key, getattr(module, key))
72
+
73
+ # 实现 MutableMapping 必须的方法
74
+ def __getitem__(self, item):
75
+ return self.attributes[item]
76
+
77
+ def __setitem__(self, key, value):
78
+ self.set(key, value)
79
+
80
+ def __delitem__(self, key):
81
+ del self.attributes[key]
82
+
83
+ def __iter__(self):
84
+ return iter(self.attributes)
85
+
86
+ def __len__(self):
87
+ return len(self.attributes)
88
+
89
+ def __str__(self):
90
+ return f'<Settings: {self.attributes}>'
91
+
92
+ __repr__ = __str__
93
+
94
+ def update_attributes(self, attributes):
95
+ if attributes is not None:
96
+ for key, value in attributes.items():
97
+ self.set(key, value)
98
+
99
+ def copy(self):
100
+ return deepcopy(self)
101
+
102
+ def __deepcopy__(self, memo):
103
+ """
104
+ 自定义深度复制方法,避免复制logger等不可pickle的对象
105
+ """
106
+ # 创建一个新的实例
107
+ cls = self.__class__
108
+ new_instance = cls.__new__(cls)
109
+
110
+ # 复制attributes字典,但排除不可pickle的对象
111
+ new_attributes = {}
112
+ for key, value in self.attributes.items():
113
+ try:
114
+ # 尝试深度复制值
115
+ new_attributes[key] = deepcopy(value, memo)
116
+ except Exception:
117
+ # 如果复制失败,保留原始引用(对于logger等对象)
118
+ new_attributes[key] = value
119
+
120
+ # 设置新实例的attributes
121
+ new_instance.attributes = new_attributes
122
+
123
123
  return new_instance