crawlo 1.1.6__py3-none-any.whl → 1.1.9__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 (190) 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 +65 -40
  8. crawlo/commands/__init__.py +14 -13
  9. crawlo/commands/check.py +594 -594
  10. crawlo/commands/genspider.py +151 -151
  11. crawlo/commands/help.py +133 -0
  12. crawlo/commands/list.py +155 -155
  13. crawlo/commands/run.py +292 -292
  14. crawlo/commands/startproject.py +418 -418
  15. crawlo/commands/stats.py +188 -188
  16. crawlo/commands/utils.py +186 -186
  17. crawlo/config.py +312 -312
  18. crawlo/config_validator.py +252 -252
  19. crawlo/core/__init__.py +2 -2
  20. crawlo/core/engine.py +345 -345
  21. crawlo/core/processor.py +40 -40
  22. crawlo/core/scheduler.py +136 -136
  23. crawlo/crawler.py +1027 -1027
  24. crawlo/downloader/__init__.py +266 -266
  25. crawlo/downloader/aiohttp_downloader.py +220 -220
  26. crawlo/downloader/cffi_downloader.py +256 -256
  27. crawlo/downloader/httpx_downloader.py +259 -259
  28. crawlo/downloader/hybrid_downloader.py +213 -213
  29. crawlo/downloader/playwright_downloader.py +402 -402
  30. crawlo/downloader/selenium_downloader.py +472 -472
  31. crawlo/event.py +11 -11
  32. crawlo/exceptions.py +81 -81
  33. crawlo/extension/__init__.py +37 -37
  34. crawlo/extension/health_check.py +141 -141
  35. crawlo/extension/log_interval.py +57 -57
  36. crawlo/extension/log_stats.py +81 -81
  37. crawlo/extension/logging_extension.py +43 -43
  38. crawlo/extension/memory_monitor.py +104 -104
  39. crawlo/extension/performance_profiler.py +133 -133
  40. crawlo/extension/request_recorder.py +107 -107
  41. crawlo/filters/__init__.py +154 -154
  42. crawlo/filters/aioredis_filter.py +280 -280
  43. crawlo/filters/memory_filter.py +269 -269
  44. crawlo/items/__init__.py +23 -23
  45. crawlo/items/base.py +21 -21
  46. crawlo/items/fields.py +53 -53
  47. crawlo/items/items.py +104 -104
  48. crawlo/middleware/__init__.py +21 -21
  49. crawlo/middleware/default_header.py +32 -32
  50. crawlo/middleware/download_delay.py +28 -28
  51. crawlo/middleware/middleware_manager.py +135 -135
  52. crawlo/middleware/proxy.py +272 -272
  53. crawlo/middleware/request_ignore.py +30 -30
  54. crawlo/middleware/response_code.py +18 -18
  55. crawlo/middleware/response_filter.py +26 -26
  56. crawlo/middleware/retry.py +124 -124
  57. crawlo/mode_manager.py +211 -211
  58. crawlo/network/__init__.py +21 -21
  59. crawlo/network/request.py +338 -338
  60. crawlo/network/response.py +359 -359
  61. crawlo/pipelines/__init__.py +21 -21
  62. crawlo/pipelines/bloom_dedup_pipeline.py +156 -156
  63. crawlo/pipelines/console_pipeline.py +39 -39
  64. crawlo/pipelines/csv_pipeline.py +316 -316
  65. crawlo/pipelines/database_dedup_pipeline.py +224 -224
  66. crawlo/pipelines/json_pipeline.py +218 -218
  67. crawlo/pipelines/memory_dedup_pipeline.py +115 -115
  68. crawlo/pipelines/mongo_pipeline.py +131 -131
  69. crawlo/pipelines/mysql_pipeline.py +316 -316
  70. crawlo/pipelines/pipeline_manager.py +61 -61
  71. crawlo/pipelines/redis_dedup_pipeline.py +167 -167
  72. crawlo/project.py +187 -187
  73. crawlo/queue/pqueue.py +37 -37
  74. crawlo/queue/queue_manager.py +334 -334
  75. crawlo/queue/redis_priority_queue.py +298 -298
  76. crawlo/settings/__init__.py +7 -7
  77. crawlo/settings/default_settings.py +219 -219
  78. crawlo/settings/setting_manager.py +122 -122
  79. crawlo/spider/__init__.py +639 -639
  80. crawlo/stats_collector.py +59 -59
  81. crawlo/subscriber.py +130 -130
  82. crawlo/task_manager.py +30 -30
  83. crawlo/templates/crawlo.cfg.tmpl +10 -10
  84. crawlo/templates/project/__init__.py.tmpl +3 -3
  85. crawlo/templates/project/items.py.tmpl +17 -17
  86. crawlo/templates/project/middlewares.py.tmpl +109 -110
  87. crawlo/templates/project/pipelines.py.tmpl +96 -97
  88. crawlo/templates/project/run.py.tmpl +46 -252
  89. crawlo/templates/project/settings.py.tmpl +326 -326
  90. crawlo/templates/project/settings_distributed.py.tmpl +119 -119
  91. crawlo/templates/project/settings_gentle.py.tmpl +94 -94
  92. crawlo/templates/project/settings_high_performance.py.tmpl +151 -151
  93. crawlo/templates/project/settings_simple.py.tmpl +68 -68
  94. crawlo/templates/project/spiders/__init__.py.tmpl +5 -5
  95. crawlo/templates/spider/spider.py.tmpl +141 -141
  96. crawlo/tools/__init__.py +182 -182
  97. crawlo/tools/anti_crawler.py +268 -268
  98. crawlo/tools/authenticated_proxy.py +240 -240
  99. crawlo/tools/data_validator.py +180 -180
  100. crawlo/tools/date_tools.py +35 -35
  101. crawlo/tools/distributed_coordinator.py +386 -386
  102. crawlo/tools/retry_mechanism.py +220 -220
  103. crawlo/tools/scenario_adapter.py +262 -262
  104. crawlo/utils/__init__.py +35 -35
  105. crawlo/utils/batch_processor.py +260 -260
  106. crawlo/utils/controlled_spider_mixin.py +439 -439
  107. crawlo/utils/date_tools.py +290 -290
  108. crawlo/utils/db_helper.py +343 -343
  109. crawlo/utils/enhanced_error_handler.py +359 -359
  110. crawlo/utils/env_config.py +105 -105
  111. crawlo/utils/error_handler.py +125 -125
  112. crawlo/utils/func_tools.py +82 -82
  113. crawlo/utils/large_scale_config.py +286 -286
  114. crawlo/utils/large_scale_helper.py +343 -343
  115. crawlo/utils/log.py +128 -128
  116. crawlo/utils/performance_monitor.py +284 -284
  117. crawlo/utils/queue_helper.py +175 -175
  118. crawlo/utils/redis_connection_pool.py +334 -334
  119. crawlo/utils/redis_key_validator.py +199 -199
  120. crawlo/utils/request.py +267 -267
  121. crawlo/utils/request_serializer.py +219 -219
  122. crawlo/utils/spider_loader.py +62 -62
  123. crawlo/utils/system.py +11 -11
  124. crawlo/utils/tools.py +4 -4
  125. crawlo/utils/url.py +39 -39
  126. {crawlo-1.1.6.dist-info → crawlo-1.1.9.dist-info}/METADATA +626 -401
  127. crawlo-1.1.9.dist-info/RECORD +190 -0
  128. examples/__init__.py +7 -7
  129. tests/DOUBLE_CRAWLO_PREFIX_FIX_REPORT.md +81 -81
  130. tests/__init__.py +7 -7
  131. tests/advanced_tools_example.py +275 -275
  132. tests/authenticated_proxy_example.py +236 -236
  133. tests/cleaners_example.py +160 -160
  134. tests/config_validation_demo.py +102 -102
  135. tests/controlled_spider_example.py +205 -205
  136. tests/date_tools_example.py +180 -180
  137. tests/dynamic_loading_example.py +523 -523
  138. tests/dynamic_loading_test.py +104 -104
  139. tests/env_config_example.py +133 -133
  140. tests/error_handling_example.py +171 -171
  141. tests/redis_key_validation_demo.py +130 -130
  142. tests/response_improvements_example.py +144 -144
  143. tests/test_advanced_tools.py +148 -148
  144. tests/test_all_redis_key_configs.py +145 -145
  145. tests/test_authenticated_proxy.py +141 -141
  146. tests/test_cleaners.py +54 -54
  147. tests/test_comprehensive.py +146 -146
  148. tests/test_config_validator.py +193 -193
  149. tests/test_date_tools.py +123 -123
  150. tests/test_double_crawlo_fix.py +207 -207
  151. tests/test_double_crawlo_fix_simple.py +124 -124
  152. tests/test_dynamic_downloaders_proxy.py +124 -124
  153. tests/test_dynamic_proxy.py +92 -92
  154. tests/test_dynamic_proxy_config.py +146 -146
  155. tests/test_dynamic_proxy_real.py +109 -109
  156. tests/test_edge_cases.py +303 -303
  157. tests/test_enhanced_error_handler.py +270 -270
  158. tests/test_env_config.py +121 -121
  159. tests/test_error_handler_compatibility.py +112 -112
  160. tests/test_final_validation.py +153 -153
  161. tests/test_framework_env_usage.py +103 -103
  162. tests/test_integration.py +356 -356
  163. tests/test_item_dedup_redis_key.py +122 -122
  164. tests/test_parsel.py +29 -29
  165. tests/test_performance.py +327 -327
  166. tests/test_proxy_health_check.py +32 -32
  167. tests/test_proxy_middleware_integration.py +136 -136
  168. tests/test_proxy_providers.py +56 -56
  169. tests/test_proxy_stats.py +19 -19
  170. tests/test_proxy_strategies.py +59 -59
  171. tests/test_queue_manager_double_crawlo.py +230 -230
  172. tests/test_queue_manager_redis_key.py +176 -176
  173. tests/test_redis_config.py +28 -28
  174. tests/test_redis_connection_pool.py +294 -294
  175. tests/test_redis_key_naming.py +181 -181
  176. tests/test_redis_key_validator.py +123 -123
  177. tests/test_redis_queue.py +224 -224
  178. tests/test_request_serialization.py +70 -70
  179. tests/test_response_improvements.py +152 -152
  180. tests/test_scheduler.py +241 -241
  181. tests/test_simple_response.py +61 -61
  182. tests/test_telecom_spider_redis_key.py +205 -205
  183. tests/test_template_content.py +87 -87
  184. tests/test_template_redis_key.py +134 -134
  185. tests/test_tools.py +153 -153
  186. tests/tools_example.py +257 -257
  187. crawlo-1.1.6.dist-info/RECORD +0 -189
  188. {crawlo-1.1.6.dist-info → crawlo-1.1.9.dist-info}/WHEEL +0 -0
  189. {crawlo-1.1.6.dist-info → crawlo-1.1.9.dist-info}/entry_points.txt +0 -0
  190. {crawlo-1.1.6.dist-info → crawlo-1.1.9.dist-info}/top_level.txt +0 -0
@@ -1,181 +1,181 @@
1
- #!/usr/bin/python
2
- # -*- coding: UTF-8 -*-
3
- """
4
- Crawlo框架日期工具使用示例
5
- """
6
- from crawlo.tools import (
7
- TimeUtils,
8
- parse_time,
9
- format_time,
10
- time_diff,
11
- to_timestamp,
12
- to_datetime,
13
- now,
14
- to_timezone,
15
- to_utc,
16
- to_local
17
- )
18
-
19
-
20
- def demo_date_parsing():
21
- """演示日期解析功能"""
22
- print("=== 日期解析演示 ===\n")
23
-
24
- # 解析各种格式的日期字符串
25
- date_strings = [
26
- "2025-09-10 14:30:00",
27
- "September 10, 2025 2:30 PM",
28
- "2025/09/10 14:30:00",
29
- "10-09-2025 14:30:00",
30
- "2025年9月10日 14时30分00秒"
31
- ]
32
-
33
- for date_str in date_strings:
34
- parsed = parse_time(date_str)
35
- print(f"解析 '{date_str}' -> {parsed}")
36
-
37
- print()
38
-
39
-
40
- def demo_date_formatting():
41
- """演示日期格式化功能"""
42
- print("=== 日期格式化演示 ===\n")
43
-
44
- # 获取当前时间
45
- current_time = now()
46
- print(f"当前时间: {current_time}")
47
-
48
- # 使用不同格式进行格式化
49
- formats = [
50
- "%Y-%m-%d",
51
- "%Y年%m月%d日",
52
- "%Y-%m-%d %H:%M:%S",
53
- "%B %d, %Y at %I:%M %p",
54
- "%A, %B %d, %Y"
55
- ]
56
-
57
- for fmt in formats:
58
- formatted = format_time(current_time, fmt)
59
- print(f"格式化为 '{fmt}' -> {formatted}")
60
-
61
- print()
62
-
63
-
64
- def demo_time_calculations():
65
- """演示时间计算功能"""
66
- print("=== 时间计算演示 ===\n")
67
-
68
- # 计算时间差
69
- start_time = "2025-09-10 10:00:00"
70
- end_time = "2025-09-10 14:30:45"
71
-
72
- diff_seconds = time_diff(start_time, end_time, "seconds")
73
- diff_minutes = time_diff(start_time, end_time, "minutes")
74
- diff_hours = time_diff(start_time, end_time, "hours")
75
-
76
- print(f"起始时间: {start_time}")
77
- print(f"结束时间: {end_time}")
78
- print(f"时间差 - 秒: {diff_seconds}")
79
- print(f"时间差 - 分钟: {diff_minutes}")
80
- print(f"时间差 - 小时: {diff_hours}")
81
-
82
- print()
83
-
84
- # 日期加减
85
- base_date = "2025-09-10"
86
- plus_30_days = TimeUtils.add_days(base_date, 30)
87
- minus_15_days = TimeUtils.add_days(base_date, -15)
88
- plus_3_months = TimeUtils.add_months(base_date, 3)
89
-
90
- print(f"基础日期: {base_date}")
91
- print(f"加30天: {plus_30_days}")
92
- print(f"减15天: {minus_15_days}")
93
- print(f"加3个月: {plus_3_months}")
94
-
95
- print()
96
-
97
-
98
- def demo_timezone_conversion():
99
- """演示时区转换功能"""
100
- print("=== 时区转换演示 ===\n")
101
-
102
- # 创建一个时间
103
- dt = parse_time("2025-09-10 14:30:00")
104
- print(f"原始时间: {dt}")
105
-
106
- # 转换为UTC时区
107
- utc_time = to_utc(dt)
108
- print(f"UTC时间: {utc_time}")
109
-
110
- # 转换为本地时区
111
- local_time = to_local(dt)
112
- print(f"本地时间: {local_time}")
113
-
114
- # 转换为纽约时区
115
- ny_time = to_timezone(dt, "America/New_York")
116
- print(f"纽约时间: {ny_time}")
117
-
118
- # 转换为伦敦时区
119
- london_time = to_timezone(dt, "Europe/London")
120
- print(f"伦敦时间: {london_time}")
121
-
122
- print()
123
-
124
-
125
- def demo_timestamp_conversion():
126
- """演示时间戳转换功能"""
127
- print("=== 时间戳转换演示 ===\n")
128
-
129
- # 获取当前时间戳
130
- current_timestamp = to_timestamp(now())
131
- print(f"当前时间戳: {current_timestamp}")
132
-
133
- # 从时间戳转换为datetime
134
- dt_from_timestamp = to_datetime(current_timestamp)
135
- print(f"从时间戳转换: {dt_from_timestamp}")
136
-
137
- # 从时间戳创建带时区的datetime
138
- dt_with_tz = to_timezone(to_datetime(current_timestamp), "Asia/Shanghai")
139
- print(f"带时区的时间: {dt_with_tz}")
140
-
141
- print()
142
-
143
-
144
- def demo_in_spider():
145
- """演示在爬虫中使用日期工具"""
146
- print("=== 在爬虫中使用日期工具 ===\n")
147
- print("在爬虫项目中,您可以这样使用日期工具:")
148
- print("""
149
- from crawlo import Spider
150
- from crawlo.tools import parse_time, format_time
151
-
152
- class NewsSpider(Spider):
153
- def parse(self, response):
154
- # 提取文章发布时间
155
- publish_time_str = response.css('.publish-time::text').get()
156
-
157
- # 解析发布时间
158
- publish_time = parse_time(publish_time_str)
159
-
160
- # 格式化时间为标准格式
161
- formatted_time = format_time(publish_time, "%Y-%m-%d %H:%M:%S")
162
-
163
- # 计算文章发布多久了
164
- current_time = self.tools.now()
165
- hours_since_publish = self.tools.time_diff(publish_time, current_time, "hours")
166
-
167
- # 根据发布时间决定是否继续处理
168
- if hours_since_publish < 24: # 只处理24小时内的文章
169
- # 处理文章...
170
- pass
171
- """)
172
-
173
-
174
- if __name__ == '__main__':
175
- # 运行演示
176
- demo_date_parsing()
177
- demo_date_formatting()
178
- demo_time_calculations()
179
- demo_timezone_conversion()
180
- demo_timestamp_conversion()
1
+ #!/usr/bin/python
2
+ # -*- coding: UTF-8 -*-
3
+ """
4
+ Crawlo框架日期工具使用示例
5
+ """
6
+ from crawlo.tools import (
7
+ TimeUtils,
8
+ parse_time,
9
+ format_time,
10
+ time_diff,
11
+ to_timestamp,
12
+ to_datetime,
13
+ now,
14
+ to_timezone,
15
+ to_utc,
16
+ to_local
17
+ )
18
+
19
+
20
+ def demo_date_parsing():
21
+ """演示日期解析功能"""
22
+ print("=== 日期解析演示 ===\n")
23
+
24
+ # 解析各种格式的日期字符串
25
+ date_strings = [
26
+ "2025-09-10 14:30:00",
27
+ "September 10, 2025 2:30 PM",
28
+ "2025/09/10 14:30:00",
29
+ "10-09-2025 14:30:00",
30
+ "2025年9月10日 14时30分00秒"
31
+ ]
32
+
33
+ for date_str in date_strings:
34
+ parsed = parse_time(date_str)
35
+ print(f"解析 '{date_str}' -> {parsed}")
36
+
37
+ print()
38
+
39
+
40
+ def demo_date_formatting():
41
+ """演示日期格式化功能"""
42
+ print("=== 日期格式化演示 ===\n")
43
+
44
+ # 获取当前时间
45
+ current_time = now()
46
+ print(f"当前时间: {current_time}")
47
+
48
+ # 使用不同格式进行格式化
49
+ formats = [
50
+ "%Y-%m-%d",
51
+ "%Y年%m月%d日",
52
+ "%Y-%m-%d %H:%M:%S",
53
+ "%B %d, %Y at %I:%M %p",
54
+ "%A, %B %d, %Y"
55
+ ]
56
+
57
+ for fmt in formats:
58
+ formatted = format_time(current_time, fmt)
59
+ print(f"格式化为 '{fmt}' -> {formatted}")
60
+
61
+ print()
62
+
63
+
64
+ def demo_time_calculations():
65
+ """演示时间计算功能"""
66
+ print("=== 时间计算演示 ===\n")
67
+
68
+ # 计算时间差
69
+ start_time = "2025-09-10 10:00:00"
70
+ end_time = "2025-09-10 14:30:45"
71
+
72
+ diff_seconds = time_diff(start_time, end_time, "seconds")
73
+ diff_minutes = time_diff(start_time, end_time, "minutes")
74
+ diff_hours = time_diff(start_time, end_time, "hours")
75
+
76
+ print(f"起始时间: {start_time}")
77
+ print(f"结束时间: {end_time}")
78
+ print(f"时间差 - 秒: {diff_seconds}")
79
+ print(f"时间差 - 分钟: {diff_minutes}")
80
+ print(f"时间差 - 小时: {diff_hours}")
81
+
82
+ print()
83
+
84
+ # 日期加减
85
+ base_date = "2025-09-10"
86
+ plus_30_days = TimeUtils.add_days(base_date, 30)
87
+ minus_15_days = TimeUtils.add_days(base_date, -15)
88
+ plus_3_months = TimeUtils.add_months(base_date, 3)
89
+
90
+ print(f"基础日期: {base_date}")
91
+ print(f"加30天: {plus_30_days}")
92
+ print(f"减15天: {minus_15_days}")
93
+ print(f"加3个月: {plus_3_months}")
94
+
95
+ print()
96
+
97
+
98
+ def demo_timezone_conversion():
99
+ """演示时区转换功能"""
100
+ print("=== 时区转换演示 ===\n")
101
+
102
+ # 创建一个时间
103
+ dt = parse_time("2025-09-10 14:30:00")
104
+ print(f"原始时间: {dt}")
105
+
106
+ # 转换为UTC时区
107
+ utc_time = to_utc(dt)
108
+ print(f"UTC时间: {utc_time}")
109
+
110
+ # 转换为本地时区
111
+ local_time = to_local(dt)
112
+ print(f"本地时间: {local_time}")
113
+
114
+ # 转换为纽约时区
115
+ ny_time = to_timezone(dt, "America/New_York")
116
+ print(f"纽约时间: {ny_time}")
117
+
118
+ # 转换为伦敦时区
119
+ london_time = to_timezone(dt, "Europe/London")
120
+ print(f"伦敦时间: {london_time}")
121
+
122
+ print()
123
+
124
+
125
+ def demo_timestamp_conversion():
126
+ """演示时间戳转换功能"""
127
+ print("=== 时间戳转换演示 ===\n")
128
+
129
+ # 获取当前时间戳
130
+ current_timestamp = to_timestamp(now())
131
+ print(f"当前时间戳: {current_timestamp}")
132
+
133
+ # 从时间戳转换为datetime
134
+ dt_from_timestamp = to_datetime(current_timestamp)
135
+ print(f"从时间戳转换: {dt_from_timestamp}")
136
+
137
+ # 从时间戳创建带时区的datetime
138
+ dt_with_tz = to_timezone(to_datetime(current_timestamp), "Asia/Shanghai")
139
+ print(f"带时区的时间: {dt_with_tz}")
140
+
141
+ print()
142
+
143
+
144
+ def demo_in_spider():
145
+ """演示在爬虫中使用日期工具"""
146
+ print("=== 在爬虫中使用日期工具 ===\n")
147
+ print("在爬虫项目中,您可以这样使用日期工具:")
148
+ print("""
149
+ from crawlo import Spider
150
+ from crawlo.tools import parse_time, format_time
151
+
152
+ class NewsSpider(Spider):
153
+ def parse(self, response):
154
+ # 提取文章发布时间
155
+ publish_time_str = response.css('.publish-time::text').get()
156
+
157
+ # 解析发布时间
158
+ publish_time = parse_time(publish_time_str)
159
+
160
+ # 格式化时间为标准格式
161
+ formatted_time = format_time(publish_time, "%Y-%m-%d %H:%M:%S")
162
+
163
+ # 计算文章发布多久了
164
+ current_time = self.tools.now()
165
+ hours_since_publish = self.tools.time_diff(publish_time, current_time, "hours")
166
+
167
+ # 根据发布时间决定是否继续处理
168
+ if hours_since_publish < 24: # 只处理24小时内的文章
169
+ # 处理文章...
170
+ pass
171
+ """)
172
+
173
+
174
+ if __name__ == '__main__':
175
+ # 运行演示
176
+ demo_date_parsing()
177
+ demo_date_formatting()
178
+ demo_time_calculations()
179
+ demo_timezone_conversion()
180
+ demo_timestamp_conversion()
181
181
  demo_in_spider()