cobweb-launcher 1.1.13__tar.gz → 1.1.15__tar.gz
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 cobweb-launcher might be problematic. Click here for more details.
- {cobweb-launcher-1.1.13/cobweb_launcher.egg-info → cobweb-launcher-1.1.15}/PKG-INFO +1 -1
- cobweb-launcher-1.1.15/cobweb/crawlers/base_crawler.py +127 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/launchers/launcher.py +1 -1
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/launchers/launcher_pro.py +1 -1
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15/cobweb_launcher.egg-info}/PKG-INFO +1 -1
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/setup.py +1 -1
- cobweb-launcher-1.1.13/cobweb/crawlers/base_crawler.py +0 -122
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/LICENSE +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/README.md +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/common_queue.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/decorators.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/item.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/log.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/request.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/response.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/base/seed.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/constant.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/crawlers/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/crawlers/file_crawler.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/db/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/db/redis_db.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/exceptions/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/exceptions/oss_db_exception.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/launchers/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/pipelines/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/pipelines/base_pipeline.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/pipelines/loghub_pipeline.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/setting.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/utils/__init__.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/utils/oss.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb/utils/tools.py +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb_launcher.egg-info/SOURCES.txt +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb_launcher.egg-info/dependency_links.txt +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb_launcher.egg-info/requires.txt +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb_launcher.egg-info/top_level.txt +0 -0
- {cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/setup.cfg +0 -0
|
@@ -0,0 +1,127 @@
|
|
|
1
|
+
import threading
|
|
2
|
+
|
|
3
|
+
from inspect import isgenerator
|
|
4
|
+
from typing import Union, Callable, Mapping
|
|
5
|
+
|
|
6
|
+
from cobweb.base import Queue, Seed, BaseItem, Request, Response, logger
|
|
7
|
+
from cobweb.constant import DealModel, LogTemplate
|
|
8
|
+
from cobweb.utils import download_log_info
|
|
9
|
+
from cobweb import setting
|
|
10
|
+
|
|
11
|
+
|
|
12
|
+
class Crawler(threading.Thread):
|
|
13
|
+
|
|
14
|
+
def __init__(
|
|
15
|
+
self,
|
|
16
|
+
upload_queue: Queue,
|
|
17
|
+
custom_func: Union[Mapping[str, Callable]],
|
|
18
|
+
launcher_queue: Union[Mapping[str, Queue]],
|
|
19
|
+
):
|
|
20
|
+
super().__init__()
|
|
21
|
+
|
|
22
|
+
self.upload_queue = upload_queue
|
|
23
|
+
for func_name, _callable in custom_func.items():
|
|
24
|
+
if isinstance(_callable, Callable):
|
|
25
|
+
self.__setattr__(func_name, _callable)
|
|
26
|
+
|
|
27
|
+
self.launcher_queue = launcher_queue
|
|
28
|
+
|
|
29
|
+
self.spider_thread_num = setting.SPIDER_THREAD_NUM
|
|
30
|
+
self.max_retries = setting.SPIDER_MAX_RETRIES
|
|
31
|
+
|
|
32
|
+
@staticmethod
|
|
33
|
+
def request(seed: Seed) -> Union[Request, BaseItem]:
|
|
34
|
+
stream = True if setting.DOWNLOAD_MODEL else False
|
|
35
|
+
yield Request(seed.url, seed, stream=stream, timeout=5)
|
|
36
|
+
|
|
37
|
+
@staticmethod
|
|
38
|
+
def download(item: Request) -> Union[Seed, BaseItem, Response, str]:
|
|
39
|
+
response = item.download()
|
|
40
|
+
yield Response(item.seed, response, **item.to_dict)
|
|
41
|
+
|
|
42
|
+
@staticmethod
|
|
43
|
+
def parse(item: Response) -> BaseItem:
|
|
44
|
+
pass
|
|
45
|
+
|
|
46
|
+
def get_seed(self) -> Seed:
|
|
47
|
+
return self.launcher_queue['todo'].pop()
|
|
48
|
+
|
|
49
|
+
def distribute(self, item, seed):
|
|
50
|
+
if isinstance(item, BaseItem):
|
|
51
|
+
self.upload_queue.push(item)
|
|
52
|
+
elif isinstance(item, Seed):
|
|
53
|
+
self.launcher_queue['new'].push(item)
|
|
54
|
+
elif isinstance(item, str) and item == DealModel.poll:
|
|
55
|
+
self.launcher_queue['todo'].push(seed)
|
|
56
|
+
elif isinstance(item, str) and item == DealModel.done:
|
|
57
|
+
self.launcher_queue['done'].push(seed)
|
|
58
|
+
elif isinstance(item, str) and item == DealModel.fail:
|
|
59
|
+
seed.identifier = DealModel.fail
|
|
60
|
+
self.launcher_queue['done'].push(seed)
|
|
61
|
+
else:
|
|
62
|
+
raise TypeError("yield value type error!")
|
|
63
|
+
|
|
64
|
+
def spider(self):
|
|
65
|
+
while True:
|
|
66
|
+
seed = self.get_seed()
|
|
67
|
+
|
|
68
|
+
if not seed:
|
|
69
|
+
continue
|
|
70
|
+
|
|
71
|
+
elif seed.params.retry >= self.max_retries:
|
|
72
|
+
seed.params.identifier = DealModel.fail
|
|
73
|
+
self.launcher_queue['done'].push(seed)
|
|
74
|
+
continue
|
|
75
|
+
|
|
76
|
+
seed_detail_log_info = download_log_info(seed.to_dict)
|
|
77
|
+
|
|
78
|
+
try:
|
|
79
|
+
request_iterators = self.request(seed)
|
|
80
|
+
|
|
81
|
+
if not isgenerator(request_iterators):
|
|
82
|
+
raise TypeError("request function isn't a generator!")
|
|
83
|
+
|
|
84
|
+
for request_item in request_iterators:
|
|
85
|
+
|
|
86
|
+
if isinstance(request_item, BaseItem):
|
|
87
|
+
self.upload_queue.push(request_item)
|
|
88
|
+
|
|
89
|
+
elif isinstance(request_item, Request):
|
|
90
|
+
download_iterators = self.download(request_item)
|
|
91
|
+
if not isgenerator(download_iterators):
|
|
92
|
+
raise TypeError("download function isn't a generator")
|
|
93
|
+
for download_item in download_iterators:
|
|
94
|
+
if isinstance(download_item, Response):
|
|
95
|
+
logger.info(LogTemplate.download_info.format(
|
|
96
|
+
detail=seed_detail_log_info,
|
|
97
|
+
retry=seed.params.retry,
|
|
98
|
+
priority=seed.params.priority,
|
|
99
|
+
seed_version=seed.params.seed_version,
|
|
100
|
+
identifier=seed.identifier or "",
|
|
101
|
+
status=download_item.response,
|
|
102
|
+
response=download_log_info(download_item.to_dict)
|
|
103
|
+
))
|
|
104
|
+
parse_iterators = self.parse(download_item)
|
|
105
|
+
if not isgenerator(parse_iterators):
|
|
106
|
+
raise TypeError("parse function isn't a generator")
|
|
107
|
+
for parse_item in parse_iterators:
|
|
108
|
+
if isinstance(parse_item, Response):
|
|
109
|
+
raise TypeError("upload_item can't be a Response instance")
|
|
110
|
+
self.distribute(parse_item, seed)
|
|
111
|
+
else:
|
|
112
|
+
self.distribute(download_item, seed)
|
|
113
|
+
except Exception as e:
|
|
114
|
+
logger.info(LogTemplate.download_exception.format(
|
|
115
|
+
detail=seed_detail_log_info,
|
|
116
|
+
retry=seed.params.retry,
|
|
117
|
+
priority=seed.params.priority,
|
|
118
|
+
seed_version=seed.params.seed_version,
|
|
119
|
+
identifier=seed.identifier or "", exception=e
|
|
120
|
+
))
|
|
121
|
+
seed.params.retry += 1
|
|
122
|
+
self.launcher_queue['todo'].push(seed)
|
|
123
|
+
|
|
124
|
+
def run(self):
|
|
125
|
+
for index in range(self.spider_thread_num):
|
|
126
|
+
threading.Thread(name=f"spider_{index}", target=self.spider).start()
|
|
127
|
+
|
|
@@ -117,7 +117,7 @@ class LauncherPro(Launcher):
|
|
|
117
117
|
seed = self.__LAUNCHER_QUEUE__['done'].pop()
|
|
118
118
|
if not seed:
|
|
119
119
|
break
|
|
120
|
-
if seed.
|
|
120
|
+
if seed.identifier == DealModel.fail:
|
|
121
121
|
f_seeds.append(seed.to_string)
|
|
122
122
|
elif self._done_model == 1:
|
|
123
123
|
s_seeds.append(seed.to_string)
|
|
@@ -1,122 +0,0 @@
|
|
|
1
|
-
import threading
|
|
2
|
-
|
|
3
|
-
from inspect import isgenerator
|
|
4
|
-
from typing import Union, Callable, Mapping
|
|
5
|
-
|
|
6
|
-
from cobweb.base import Queue, Seed, BaseItem, Request, Response, logger
|
|
7
|
-
from cobweb.constant import DealModel, LogTemplate
|
|
8
|
-
from cobweb.utils import download_log_info
|
|
9
|
-
from cobweb import setting
|
|
10
|
-
|
|
11
|
-
|
|
12
|
-
class Crawler(threading.Thread):
|
|
13
|
-
|
|
14
|
-
def __init__(
|
|
15
|
-
self,
|
|
16
|
-
upload_queue: Queue,
|
|
17
|
-
custom_func: Union[Mapping[str, Callable]],
|
|
18
|
-
launcher_queue: Union[Mapping[str, Queue]],
|
|
19
|
-
):
|
|
20
|
-
super().__init__()
|
|
21
|
-
|
|
22
|
-
self.upload_queue = upload_queue
|
|
23
|
-
for func_name, _callable in custom_func.items():
|
|
24
|
-
if isinstance(_callable, Callable):
|
|
25
|
-
self.__setattr__(func_name, _callable)
|
|
26
|
-
|
|
27
|
-
self.launcher_queue = launcher_queue
|
|
28
|
-
|
|
29
|
-
self.spider_thread_num = setting.SPIDER_THREAD_NUM
|
|
30
|
-
self.max_retries = setting.SPIDER_MAX_RETRIES
|
|
31
|
-
|
|
32
|
-
@staticmethod
|
|
33
|
-
def request(seed: Seed) -> Union[Request, BaseItem]:
|
|
34
|
-
stream = True if setting.DOWNLOAD_MODEL else False
|
|
35
|
-
return Request(seed.url, seed, stream=stream, timeout=5)
|
|
36
|
-
|
|
37
|
-
@staticmethod
|
|
38
|
-
def download(item: Request) -> Union[Seed, BaseItem, Response, str]:
|
|
39
|
-
response = item.download()
|
|
40
|
-
yield Response(item.seed, response, **item.to_dict)
|
|
41
|
-
|
|
42
|
-
@staticmethod
|
|
43
|
-
def parse(item: Response) -> BaseItem:
|
|
44
|
-
pass
|
|
45
|
-
|
|
46
|
-
def get(self) -> Seed:
|
|
47
|
-
return self.launcher_queue['todo'].pop()
|
|
48
|
-
|
|
49
|
-
def spider(self):
|
|
50
|
-
while True:
|
|
51
|
-
seed = self.get()
|
|
52
|
-
|
|
53
|
-
if not seed:
|
|
54
|
-
continue
|
|
55
|
-
|
|
56
|
-
elif seed.params.retry >= self.max_retries:
|
|
57
|
-
seed.params.identifier = DealModel.fail
|
|
58
|
-
self.launcher_queue['done'].push(seed)
|
|
59
|
-
continue
|
|
60
|
-
|
|
61
|
-
item = self.request(seed)
|
|
62
|
-
|
|
63
|
-
if isinstance(item, Request):
|
|
64
|
-
|
|
65
|
-
download_iterators = self.download(item)
|
|
66
|
-
|
|
67
|
-
if not isgenerator(download_iterators):
|
|
68
|
-
raise TypeError("download function isn't a generator")
|
|
69
|
-
|
|
70
|
-
seed_detail_log_info = download_log_info(seed.to_dict)
|
|
71
|
-
|
|
72
|
-
try:
|
|
73
|
-
for it in download_iterators:
|
|
74
|
-
if isinstance(it, Response):
|
|
75
|
-
response_detail_log_info = download_log_info(it.to_dict)
|
|
76
|
-
logger.info(LogTemplate.download_info.format(
|
|
77
|
-
detail=seed_detail_log_info, retry=item.seed.params.retry,
|
|
78
|
-
priority=item.seed.params.priority,
|
|
79
|
-
seed_version=item.seed.params.seed_version,
|
|
80
|
-
identifier=item.seed.identifier or "",
|
|
81
|
-
status=it.response, response=response_detail_log_info
|
|
82
|
-
))
|
|
83
|
-
parse_iterators = self.parse(it)
|
|
84
|
-
if not isgenerator(parse_iterators):
|
|
85
|
-
raise TypeError("parse function isn't a generator")
|
|
86
|
-
for upload_item in parse_iterators:
|
|
87
|
-
if not isinstance(upload_item, BaseItem):
|
|
88
|
-
raise TypeError("upload_item isn't BaseItem subclass")
|
|
89
|
-
self.upload_queue.push(upload_item)
|
|
90
|
-
elif isinstance(it, BaseItem):
|
|
91
|
-
self.upload_queue.push(it)
|
|
92
|
-
elif isinstance(it, Seed):
|
|
93
|
-
self.launcher_queue['new'].push(it)
|
|
94
|
-
elif isinstance(it, str) and it == DealModel.poll:
|
|
95
|
-
self.launcher_queue['todo'].push(item)
|
|
96
|
-
break
|
|
97
|
-
elif isinstance(it, str) and it == DealModel.done:
|
|
98
|
-
self.launcher_queue['done'].push(seed)
|
|
99
|
-
break
|
|
100
|
-
elif isinstance(it, str) and it == DealModel.fail:
|
|
101
|
-
seed.params.identifier = DealModel.fail
|
|
102
|
-
self.launcher_queue['done'].push(seed)
|
|
103
|
-
break
|
|
104
|
-
else:
|
|
105
|
-
raise TypeError("yield value type error!")
|
|
106
|
-
|
|
107
|
-
except Exception as e:
|
|
108
|
-
logger.info(LogTemplate.download_exception.format(
|
|
109
|
-
detail=seed_detail_log_info, retry=seed.params.retry,
|
|
110
|
-
priority=seed.params.priority, seed_version=seed.params.seed_version,
|
|
111
|
-
identifier=seed.identifier or "", exception=e
|
|
112
|
-
))
|
|
113
|
-
seed.params.retry += 1
|
|
114
|
-
self.launcher_queue['todo'].push(seed)
|
|
115
|
-
|
|
116
|
-
elif isinstance(item, BaseItem):
|
|
117
|
-
self.upload_queue.push(item)
|
|
118
|
-
|
|
119
|
-
def run(self):
|
|
120
|
-
for index in range(self.spider_thread_num):
|
|
121
|
-
threading.Thread(name=f"spider_{index}", target=self.spider).start()
|
|
122
|
-
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{cobweb-launcher-1.1.13 → cobweb-launcher-1.1.15}/cobweb_launcher.egg-info/dependency_links.txt
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|