cobweb-launcher 1.3.6__py3-none-any.whl → 1.3.8__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.
- cobweb/base/__init__.py +9 -9
- cobweb/base/dotting.py +1 -1
- cobweb/utils/oss.py +7 -7
- {cobweb_launcher-1.3.6.dist-info → cobweb_launcher-1.3.8.dist-info}/METADATA +1 -1
- cobweb_launcher-1.3.8.dist-info/RECORD +40 -0
- cobweb/base/decorators.py +0 -40
- cobweb/crawlers/base_crawler.py +0 -144
- cobweb/crawlers/file_crawler.py +0 -98
- cobweb/pipelines/base_pipeline.py +0 -54
- cobweb/pipelines/loghub_pipeline.py +0 -34
- cobweb/utils/dotting.py +0 -32
- cobweb_/__init__.py +0 -2
- cobweb_/base/__init__.py +0 -9
- cobweb_/base/common_queue.py +0 -30
- cobweb_/base/decorators.py +0 -40
- cobweb_/base/item.py +0 -46
- cobweb_/base/log.py +0 -94
- cobweb_/base/request.py +0 -82
- cobweb_/base/response.py +0 -23
- cobweb_/base/seed.py +0 -114
- cobweb_/constant.py +0 -94
- cobweb_/crawlers/__init__.py +0 -1
- cobweb_/crawlers/crawler.py +0 -184
- cobweb_/db/__init__.py +0 -2
- cobweb_/db/api_db.py +0 -82
- cobweb_/db/redis_db.py +0 -130
- cobweb_/exceptions/__init__.py +0 -1
- cobweb_/exceptions/oss_db_exception.py +0 -28
- cobweb_/launchers/__init__.py +0 -3
- cobweb_/launchers/launcher.py +0 -235
- cobweb_/launchers/launcher_air.py +0 -88
- cobweb_/launchers/launcher_api.py +0 -221
- cobweb_/launchers/launcher_pro.py +0 -222
- cobweb_/pipelines/__init__.py +0 -3
- cobweb_/pipelines/pipeline.py +0 -69
- cobweb_/pipelines/pipeline_console.py +0 -22
- cobweb_/pipelines/pipeline_loghub.py +0 -34
- cobweb_/setting.py +0 -74
- cobweb_/utils/__init__.py +0 -5
- cobweb_/utils/bloom.py +0 -58
- cobweb_/utils/dotting.py +0 -32
- cobweb_/utils/oss.py +0 -94
- cobweb_/utils/tools.py +0 -42
- cobweb_launcher-1.3.6.dist-info/RECORD +0 -111
- cobweb_new/__init__.py +0 -2
- cobweb_new/base/__init__.py +0 -72
- cobweb_new/base/common_queue.py +0 -53
- cobweb_new/base/decorators.py +0 -72
- cobweb_new/base/item.py +0 -46
- cobweb_new/base/log.py +0 -94
- cobweb_new/base/request.py +0 -82
- cobweb_new/base/response.py +0 -23
- cobweb_new/base/seed.py +0 -118
- cobweb_new/constant.py +0 -105
- cobweb_new/crawlers/__init__.py +0 -1
- cobweb_new/crawlers/crawler-new.py +0 -85
- cobweb_new/crawlers/crawler.py +0 -170
- cobweb_new/db/__init__.py +0 -2
- cobweb_new/db/api_db.py +0 -82
- cobweb_new/db/redis_db.py +0 -158
- cobweb_new/exceptions/__init__.py +0 -1
- cobweb_new/exceptions/oss_db_exception.py +0 -28
- cobweb_new/launchers/__init__.py +0 -3
- cobweb_new/launchers/launcher.py +0 -237
- cobweb_new/launchers/launcher_air.py +0 -88
- cobweb_new/launchers/launcher_api.py +0 -161
- cobweb_new/launchers/launcher_pro.py +0 -96
- cobweb_new/launchers/tesss.py +0 -47
- cobweb_new/pipelines/__init__.py +0 -3
- cobweb_new/pipelines/pipeline.py +0 -68
- cobweb_new/pipelines/pipeline_console.py +0 -22
- cobweb_new/pipelines/pipeline_loghub.py +0 -34
- cobweb_new/setting.py +0 -95
- cobweb_new/utils/__init__.py +0 -5
- cobweb_new/utils/bloom.py +0 -58
- cobweb_new/utils/oss.py +0 -94
- cobweb_new/utils/tools.py +0 -42
- {cobweb_launcher-1.3.6.dist-info → cobweb_launcher-1.3.8.dist-info}/LICENSE +0 -0
- {cobweb_launcher-1.3.6.dist-info → cobweb_launcher-1.3.8.dist-info}/WHEEL +0 -0
- {cobweb_launcher-1.3.6.dist-info → cobweb_launcher-1.3.8.dist-info}/top_level.txt +0 -0
cobweb_/base/item.py
DELETED
@@ -1,46 +0,0 @@
|
|
1
|
-
from .seed import Seed
|
2
|
-
from collections import namedtuple
|
3
|
-
|
4
|
-
|
5
|
-
class Item(type):
|
6
|
-
|
7
|
-
def __new__(cls, name, bases, dct):
|
8
|
-
new_class_instance = type.__new__(cls, name, bases, dct)
|
9
|
-
if name != "BaseItem":
|
10
|
-
table = getattr(new_class_instance, "__TABLE__")
|
11
|
-
fields = getattr(new_class_instance, "__FIELDS__")
|
12
|
-
new_class_instance.Data = namedtuple(table, fields)
|
13
|
-
return new_class_instance
|
14
|
-
|
15
|
-
|
16
|
-
class BaseItem(metaclass=Item):
|
17
|
-
|
18
|
-
__TABLE__ = ""
|
19
|
-
__FIELDS__ = ""
|
20
|
-
|
21
|
-
def __init__(self, seed: Seed, **kwargs):
|
22
|
-
self.seed = seed
|
23
|
-
|
24
|
-
data = {}
|
25
|
-
for key, value in kwargs.items():
|
26
|
-
if key not in self.__FIELDS__:
|
27
|
-
self.__setattr__(key, value)
|
28
|
-
else:
|
29
|
-
data[key] = value
|
30
|
-
|
31
|
-
self.data = self.Data(**data)
|
32
|
-
|
33
|
-
@property
|
34
|
-
def to_dict(self):
|
35
|
-
return self.data._asdict()
|
36
|
-
|
37
|
-
@property
|
38
|
-
def table(self):
|
39
|
-
return self.Data.__name__
|
40
|
-
|
41
|
-
|
42
|
-
class ConsoleItem(BaseItem):
|
43
|
-
|
44
|
-
__TABLE__ = "console"
|
45
|
-
__FIELDS__ = "data"
|
46
|
-
|
cobweb_/base/log.py
DELETED
@@ -1,94 +0,0 @@
|
|
1
|
-
import logging
|
2
|
-
|
3
|
-
|
4
|
-
class ColorCodes:
|
5
|
-
# Text Reset
|
6
|
-
RESET = "\033[0m"
|
7
|
-
|
8
|
-
# Regular Colors
|
9
|
-
RED = "\033[31m"
|
10
|
-
GREEN = "\033[32m"
|
11
|
-
YELLOW = "\033[33m"
|
12
|
-
BLUE = "\033[34m"
|
13
|
-
PURPLE = "\033[35m"
|
14
|
-
CYAN = "\033[36m"
|
15
|
-
WHITE = "\033[37m"
|
16
|
-
|
17
|
-
# Bright Colors
|
18
|
-
BRIGHT_RED = "\033[91m"
|
19
|
-
BRIGHT_GREEN = "\033[92m"
|
20
|
-
BRIGHT_YELLOW = "\033[93m"
|
21
|
-
BRIGHT_BLUE = "\033[94m"
|
22
|
-
BRIGHT_PURPLE = "\033[95m"
|
23
|
-
BRIGHT_CYAN = "\033[96m"
|
24
|
-
BRIGHT_WHITE = "\033[97m"
|
25
|
-
|
26
|
-
# Background Colors
|
27
|
-
BG_RED = "\033[41m"
|
28
|
-
BG_GREEN = "\033[42m"
|
29
|
-
BG_YELLOW = "\033[43m"
|
30
|
-
BG_BLUE = "\033[44m"
|
31
|
-
BG_PURPLE = "\033[45m"
|
32
|
-
BG_CYAN = "\033[46m"
|
33
|
-
BG_WHITE = "\033[47m"
|
34
|
-
|
35
|
-
# Bright Background Colors
|
36
|
-
BG_BRIGHT_RED = "\033[101m"
|
37
|
-
BG_BRIGHT_GREEN = "\033[102m"
|
38
|
-
BG_BRIGHT_YELLOW = "\033[103m"
|
39
|
-
BG_BRIGHT_BLUE = "\033[104m"
|
40
|
-
BG_BRIGHT_PURPLE = "\033[105m"
|
41
|
-
BG_BRIGHT_CYAN = "\033[106m"
|
42
|
-
BG_BRIGHT_WHITE = "\033[107m"
|
43
|
-
|
44
|
-
# Text Styles
|
45
|
-
BOLD = "\033[1m"
|
46
|
-
DIM = "\033[2m"
|
47
|
-
ITALIC = "\033[3m"
|
48
|
-
UNDERLINE = "\033[4m"
|
49
|
-
BLINK = "\033[5m"
|
50
|
-
REVERSE = "\033[7m"
|
51
|
-
HIDDEN = "\033[8m"
|
52
|
-
|
53
|
-
|
54
|
-
class Log:
|
55
|
-
logging.getLogger('oss2.api').setLevel(logging.WARNING)
|
56
|
-
logging.basicConfig(
|
57
|
-
level=logging.INFO,
|
58
|
-
format=f'%(asctime)s %(name)s [%(filename)s:%(lineno)d %(funcName)s]'
|
59
|
-
f' %(levelname)s -> %(message)s'
|
60
|
-
)
|
61
|
-
log = logging.getLogger()
|
62
|
-
|
63
|
-
def set_log_name(self, name):
|
64
|
-
self.__class__.log = logging.getLogger(name)
|
65
|
-
|
66
|
-
@property
|
67
|
-
def debug(self):
|
68
|
-
return self.__class__.log.debug
|
69
|
-
|
70
|
-
@property
|
71
|
-
def info(self):
|
72
|
-
return self.__class__.log.info
|
73
|
-
|
74
|
-
@property
|
75
|
-
def warning(self):
|
76
|
-
return self.__class__.log.warning
|
77
|
-
|
78
|
-
@property
|
79
|
-
def exception(self):
|
80
|
-
return self.__class__.log.exception
|
81
|
-
|
82
|
-
@property
|
83
|
-
def error(self):
|
84
|
-
return self.__class__.log.error
|
85
|
-
|
86
|
-
@property
|
87
|
-
def critical(self):
|
88
|
-
return self.__class__.log.critical
|
89
|
-
|
90
|
-
|
91
|
-
logger = Log()
|
92
|
-
|
93
|
-
|
94
|
-
|
cobweb_/base/request.py
DELETED
@@ -1,82 +0,0 @@
|
|
1
|
-
import random
|
2
|
-
import requests
|
3
|
-
|
4
|
-
|
5
|
-
class Request:
|
6
|
-
|
7
|
-
__REQUEST_ATTRS__ = {
|
8
|
-
"params",
|
9
|
-
"headers",
|
10
|
-
"cookies",
|
11
|
-
"data",
|
12
|
-
"json",
|
13
|
-
"files",
|
14
|
-
"auth",
|
15
|
-
"timeout",
|
16
|
-
"proxies",
|
17
|
-
"hooks",
|
18
|
-
"stream",
|
19
|
-
"verify",
|
20
|
-
"cert",
|
21
|
-
"allow_redirects",
|
22
|
-
}
|
23
|
-
|
24
|
-
def __init__(
|
25
|
-
self,
|
26
|
-
url,
|
27
|
-
seed,
|
28
|
-
random_ua=True,
|
29
|
-
check_status_code=True,
|
30
|
-
**kwargs
|
31
|
-
):
|
32
|
-
self.url = url
|
33
|
-
self.seed = seed
|
34
|
-
self.check_status_code = check_status_code
|
35
|
-
self.request_setting = {}
|
36
|
-
|
37
|
-
for k, v in kwargs.items():
|
38
|
-
if k in self.__class__.__REQUEST_ATTRS__:
|
39
|
-
self.request_setting[k] = v
|
40
|
-
continue
|
41
|
-
self.__setattr__(k, v)
|
42
|
-
|
43
|
-
if not getattr(self, "method", None):
|
44
|
-
self.method = "POST" if self.request_setting.get("data") or self.request_setting.get("json") else "GET"
|
45
|
-
|
46
|
-
if random_ua:
|
47
|
-
self._build_header()
|
48
|
-
|
49
|
-
@property
|
50
|
-
def _random_ua(self) -> str:
|
51
|
-
v1 = random.randint(4, 15)
|
52
|
-
v2 = random.randint(3, 11)
|
53
|
-
v3 = random.randint(1, 16)
|
54
|
-
v4 = random.randint(533, 605)
|
55
|
-
v5 = random.randint(1000, 6000)
|
56
|
-
v6 = random.randint(10, 80)
|
57
|
-
user_agent = (f"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_{v1}_{v2}) AppleWebKit/{v4}.{v3} "
|
58
|
-
f"(KHTML, like Gecko) Chrome/105.0.0.0 Safari/{v4}.{v3} Edg/105.0.{v5}.{v6}")
|
59
|
-
return user_agent
|
60
|
-
|
61
|
-
def _build_header(self) -> dict:
|
62
|
-
if not self.request_setting.get("headers"):
|
63
|
-
self.request_setting["headers"] = {"accept": "*/*", "user-agent": self._random_ua}
|
64
|
-
elif "user-agent" not in [key.lower() for key in self.request_setting["headers"].keys()]:
|
65
|
-
self.request_setting["headers"]["user-agent"] = self._random_ua
|
66
|
-
|
67
|
-
def download(self) -> requests.Response:
|
68
|
-
response = requests.request(self.method, self.url, **self.request_setting)
|
69
|
-
if self.check_status_code:
|
70
|
-
response.raise_for_status()
|
71
|
-
return response
|
72
|
-
|
73
|
-
@property
|
74
|
-
def to_dict(self):
|
75
|
-
_dict = self.__dict__.copy()
|
76
|
-
_dict.pop('url')
|
77
|
-
_dict.pop('seed')
|
78
|
-
_dict.pop('check_status_code')
|
79
|
-
_dict.pop('request_setting')
|
80
|
-
return _dict
|
81
|
-
|
82
|
-
|
cobweb_/base/response.py
DELETED
@@ -1,23 +0,0 @@
|
|
1
|
-
|
2
|
-
|
3
|
-
class Response:
|
4
|
-
|
5
|
-
def __init__(
|
6
|
-
self,
|
7
|
-
seed,
|
8
|
-
response,
|
9
|
-
**kwargs
|
10
|
-
):
|
11
|
-
self.seed = seed
|
12
|
-
self.response = response
|
13
|
-
|
14
|
-
for k, v in kwargs.items():
|
15
|
-
self.__setattr__(k, v)
|
16
|
-
|
17
|
-
@property
|
18
|
-
def to_dict(self):
|
19
|
-
_dict = self.__dict__.copy()
|
20
|
-
_dict.pop('seed')
|
21
|
-
_dict.pop('response')
|
22
|
-
return _dict
|
23
|
-
|
cobweb_/base/seed.py
DELETED
@@ -1,114 +0,0 @@
|
|
1
|
-
import json
|
2
|
-
import time
|
3
|
-
import hashlib
|
4
|
-
|
5
|
-
|
6
|
-
class SeedParams:
|
7
|
-
|
8
|
-
def __init__(self, retry, priority, seed_version, seed_status=None):
|
9
|
-
self.retry = retry or 0
|
10
|
-
self.priority = priority or 300
|
11
|
-
self.seed_version = seed_version or int(time.time())
|
12
|
-
self.seed_status = seed_status
|
13
|
-
|
14
|
-
|
15
|
-
class Seed:
|
16
|
-
|
17
|
-
__SEED_PARAMS__ = [
|
18
|
-
"retry",
|
19
|
-
"priority",
|
20
|
-
"seed_version",
|
21
|
-
"seed_status"
|
22
|
-
]
|
23
|
-
|
24
|
-
def __init__(
|
25
|
-
self,
|
26
|
-
seed,
|
27
|
-
sid=None,
|
28
|
-
retry=None,
|
29
|
-
priority=None,
|
30
|
-
seed_version=None,
|
31
|
-
seed_status=None,
|
32
|
-
**kwargs
|
33
|
-
):
|
34
|
-
if any(isinstance(seed, t) for t in (str, bytes)):
|
35
|
-
try:
|
36
|
-
item = json.loads(seed)
|
37
|
-
self._init_seed(item)
|
38
|
-
except json.JSONDecodeError:
|
39
|
-
self.__setattr__("url", seed)
|
40
|
-
elif isinstance(seed, dict):
|
41
|
-
self._init_seed(seed)
|
42
|
-
else:
|
43
|
-
raise TypeError(Exception(
|
44
|
-
f"seed type error, "
|
45
|
-
f"must be str or dict! "
|
46
|
-
f"seed: {seed}"
|
47
|
-
))
|
48
|
-
|
49
|
-
seed_params = {
|
50
|
-
"retry": retry,
|
51
|
-
"priority": priority,
|
52
|
-
"seed_version": seed_version,
|
53
|
-
"seed_status": seed_status,
|
54
|
-
}
|
55
|
-
|
56
|
-
if kwargs:
|
57
|
-
self._init_seed(kwargs)
|
58
|
-
seed_params.update({
|
59
|
-
k:v for k, v in kwargs.items()
|
60
|
-
if k in self.__SEED_PARAMS__
|
61
|
-
})
|
62
|
-
if sid or not getattr(self, "sid", None):
|
63
|
-
self._init_id(sid)
|
64
|
-
self.params = SeedParams(**seed_params)
|
65
|
-
|
66
|
-
def __getattr__(self, name):
|
67
|
-
return None
|
68
|
-
|
69
|
-
def __setitem__(self, key, value):
|
70
|
-
setattr(self, key, value)
|
71
|
-
|
72
|
-
def __getitem__(self, item):
|
73
|
-
return getattr(self, item)
|
74
|
-
|
75
|
-
def __str__(self):
|
76
|
-
return json.dumps(self.__dict__, ensure_ascii=False)
|
77
|
-
|
78
|
-
def __repr__(self):
|
79
|
-
chars = [f"{k}={v}" for k, v in self.__dict__.items()]
|
80
|
-
return f'{self.__class__.__name__}({", ".join(chars)})'
|
81
|
-
|
82
|
-
def _init_seed(self, seed_info:dict):
|
83
|
-
for k, v in seed_info.items():
|
84
|
-
if k not in self.__SEED_PARAMS__:
|
85
|
-
self.__setattr__(k, v)
|
86
|
-
|
87
|
-
def _init_id(self, sid):
|
88
|
-
if not sid:
|
89
|
-
sid = hashlib.md5(self.to_string.encode()).hexdigest()
|
90
|
-
self.__setattr__("sid", sid)
|
91
|
-
|
92
|
-
@property
|
93
|
-
def to_dict(self) -> dict:
|
94
|
-
seed = self.__dict__.copy()
|
95
|
-
if seed.get("params"):
|
96
|
-
del seed["params"]
|
97
|
-
return seed
|
98
|
-
|
99
|
-
@property
|
100
|
-
def to_string(self) -> str:
|
101
|
-
return json.dumps(
|
102
|
-
self.to_dict,
|
103
|
-
ensure_ascii=False,
|
104
|
-
separators=(",", ":")
|
105
|
-
)
|
106
|
-
|
107
|
-
@property
|
108
|
-
def get_all(self):
|
109
|
-
return json.dumps(
|
110
|
-
self.__dict__,
|
111
|
-
ensure_ascii=False,
|
112
|
-
separators=(",", ":")
|
113
|
-
)
|
114
|
-
|
cobweb_/constant.py
DELETED
@@ -1,94 +0,0 @@
|
|
1
|
-
|
2
|
-
class CrawlerModel:
|
3
|
-
|
4
|
-
default = "cobweb.crawlers.Crawler"
|
5
|
-
file_air = "cobweb.crawlers.FileCrawlerAir"
|
6
|
-
file_pro = "cobweb.crawlers.FileCrawlerPro"
|
7
|
-
|
8
|
-
|
9
|
-
class LauncherModel:
|
10
|
-
task = "launcher model: task"
|
11
|
-
resident = "launcher model: resident"
|
12
|
-
|
13
|
-
|
14
|
-
class DownloadModel:
|
15
|
-
common = "download model: common"
|
16
|
-
file = "download model: file"
|
17
|
-
|
18
|
-
|
19
|
-
class LogModel:
|
20
|
-
simple = "log model: simple"
|
21
|
-
common = "log model: common"
|
22
|
-
detailed = "log model: detailed"
|
23
|
-
|
24
|
-
|
25
|
-
class DealModel:
|
26
|
-
fail = "deal model: fail"
|
27
|
-
done = "deal model: done"
|
28
|
-
poll = "deal model: poll"
|
29
|
-
|
30
|
-
|
31
|
-
class LogTemplate:
|
32
|
-
|
33
|
-
console_item = """
|
34
|
-
----------------------- start - console pipeline -----------------
|
35
|
-
种子详情 \n{seed_detail}
|
36
|
-
解析详情 \n{parse_detail}
|
37
|
-
----------------------- end - console pipeline ------------------
|
38
|
-
"""
|
39
|
-
|
40
|
-
launcher_air_polling = """
|
41
|
-
----------------------- start - 轮训日志: {task} -----------------
|
42
|
-
内存队列
|
43
|
-
种子数: {doing_len}
|
44
|
-
待消费: {todo_len}
|
45
|
-
已消费: {done_len}
|
46
|
-
存储队列
|
47
|
-
待上传: {upload_len}
|
48
|
-
----------------------- end - 轮训日志: {task} ------------------
|
49
|
-
"""
|
50
|
-
|
51
|
-
launcher_pro_polling = """
|
52
|
-
----------------------- start - 轮训日志: {task} -----------------
|
53
|
-
内存队列
|
54
|
-
种子数: {doing_len}
|
55
|
-
待消费: {todo_len}
|
56
|
-
已消费: {done_len}
|
57
|
-
redis队列
|
58
|
-
种子数: {redis_seed_count}
|
59
|
-
待消费: {redis_todo_len}
|
60
|
-
消费中: {redis_doing_len}
|
61
|
-
存储队列
|
62
|
-
待上传: {upload_len}
|
63
|
-
----------------------- end - 轮训日志: {task} ------------------
|
64
|
-
"""
|
65
|
-
|
66
|
-
download_exception = """
|
67
|
-
----------------------- download exception -----------------------
|
68
|
-
种子详情 \n{detail}
|
69
|
-
种子参数
|
70
|
-
retry : {retry}
|
71
|
-
priority : {priority}
|
72
|
-
seed_version : {seed_version}
|
73
|
-
identifier : {identifier}
|
74
|
-
exception
|
75
|
-
msg : {exception}
|
76
|
-
------------------------------------------------------------------
|
77
|
-
"""
|
78
|
-
|
79
|
-
download_info = """
|
80
|
-
------------------------ download info ---------------------------
|
81
|
-
种子详情 \n{detail}
|
82
|
-
种子参数
|
83
|
-
retry : {retry}
|
84
|
-
priority : {priority}
|
85
|
-
seed_version : {seed_version}
|
86
|
-
identifier : {identifier}
|
87
|
-
response
|
88
|
-
status : {status} \n{response}
|
89
|
-
------------------------------------------------------------------
|
90
|
-
"""
|
91
|
-
|
92
|
-
@staticmethod
|
93
|
-
def log_info(item: dict) -> str:
|
94
|
-
return "\n".join([" " * 12 + f"{str(k).ljust(14)}: {str(v)}" for k, v in item.items()])
|
cobweb_/crawlers/__init__.py
DELETED
@@ -1 +0,0 @@
|
|
1
|
-
from .crawler import Crawler
|
cobweb_/crawlers/crawler.py
DELETED
@@ -1,184 +0,0 @@
|
|
1
|
-
import json
|
2
|
-
import threading
|
3
|
-
import time
|
4
|
-
import traceback
|
5
|
-
from inspect import isgenerator
|
6
|
-
from typing import Union, Callable, Mapping
|
7
|
-
|
8
|
-
from cobweb.constant import DealModel, LogTemplate
|
9
|
-
from cobweb.base import (
|
10
|
-
Queue,
|
11
|
-
Seed,
|
12
|
-
BaseItem,
|
13
|
-
Request,
|
14
|
-
Response,
|
15
|
-
ConsoleItem,
|
16
|
-
logger
|
17
|
-
)
|
18
|
-
from cobweb.utils import LoghubDot
|
19
|
-
|
20
|
-
|
21
|
-
class Crawler(threading.Thread):
|
22
|
-
|
23
|
-
def __init__(
|
24
|
-
self,
|
25
|
-
task: str,
|
26
|
-
project: str,
|
27
|
-
stop: threading.Event,
|
28
|
-
pause: threading.Event,
|
29
|
-
# launcher_queue: Union[Mapping[str, Queue]],
|
30
|
-
get_seed: Callable,
|
31
|
-
set_seed: Callable,
|
32
|
-
add_seed: Callable,
|
33
|
-
delete_seed: Callable,
|
34
|
-
upload_data: Callable,
|
35
|
-
custom_func: Union[Mapping[str, Callable]],
|
36
|
-
thread_num: int,
|
37
|
-
max_retries: int,
|
38
|
-
time_sleep: int,
|
39
|
-
):
|
40
|
-
super().__init__()
|
41
|
-
self.task = task
|
42
|
-
self.project = project
|
43
|
-
self._stop = stop
|
44
|
-
self._pause = pause
|
45
|
-
self._get_seed = get_seed
|
46
|
-
self._set_seed = set_seed
|
47
|
-
self._add_seed = add_seed
|
48
|
-
self._delete_seed = delete_seed
|
49
|
-
self._upload_data = upload_data
|
50
|
-
|
51
|
-
for func_name, _callable in custom_func.items():
|
52
|
-
if isinstance(_callable, Callable):
|
53
|
-
self.__setattr__(func_name, _callable)
|
54
|
-
|
55
|
-
self.thread_num = thread_num
|
56
|
-
self.time_sleep = time_sleep
|
57
|
-
self.max_retries = max_retries
|
58
|
-
|
59
|
-
self.loghub_dot = LoghubDot()
|
60
|
-
|
61
|
-
@staticmethod
|
62
|
-
def request(seed: Seed) -> Union[Request, BaseItem]:
|
63
|
-
yield Request(seed.url, seed, timeout=5)
|
64
|
-
|
65
|
-
@staticmethod
|
66
|
-
def download(item: Request) -> Union[Seed, BaseItem, Response, str]:
|
67
|
-
response = item.download()
|
68
|
-
yield Response(item.seed, response, **item.to_dict)
|
69
|
-
|
70
|
-
@staticmethod
|
71
|
-
def parse(item: Response) -> BaseItem:
|
72
|
-
upload_item = item.to_dict
|
73
|
-
upload_item["text"] = item.response.text
|
74
|
-
yield ConsoleItem(item.seed, data=json.dumps(upload_item, ensure_ascii=False))
|
75
|
-
|
76
|
-
# def get_seed(self) -> Seed:
|
77
|
-
# return self._todo.pop()
|
78
|
-
|
79
|
-
def distribute(self, item, seed):
|
80
|
-
if isinstance(item, BaseItem):
|
81
|
-
self._upload_data(item)
|
82
|
-
elif isinstance(item, Seed):
|
83
|
-
self._add_seed(item)
|
84
|
-
elif isinstance(item, str) and item == DealModel.poll:
|
85
|
-
self._set_seed(seed)
|
86
|
-
elif isinstance(item, str) and item == DealModel.done:
|
87
|
-
self._delete_seed(seed)
|
88
|
-
elif isinstance(item, str) and item == DealModel.fail:
|
89
|
-
seed.params.seed_status = DealModel.fail
|
90
|
-
self._delete_seed(seed)
|
91
|
-
else:
|
92
|
-
raise TypeError("yield value type error!")
|
93
|
-
|
94
|
-
def spider(self):
|
95
|
-
while not self._stop.is_set():
|
96
|
-
|
97
|
-
seed = self._get_seed()
|
98
|
-
|
99
|
-
if not seed:
|
100
|
-
time.sleep(1)
|
101
|
-
continue
|
102
|
-
|
103
|
-
elif seed.params.retry > self.max_retries:
|
104
|
-
seed.params.seed_status = DealModel.fail
|
105
|
-
self._delete_seed(seed)
|
106
|
-
continue
|
107
|
-
|
108
|
-
seed_detail_log_info = LogTemplate.log_info(seed.to_dict)
|
109
|
-
|
110
|
-
try:
|
111
|
-
request_iterators = self.request(seed)
|
112
|
-
|
113
|
-
if not isgenerator(request_iterators):
|
114
|
-
raise TypeError("request function isn't a generator!")
|
115
|
-
|
116
|
-
iterator_status = False
|
117
|
-
|
118
|
-
for request_item in request_iterators:
|
119
|
-
|
120
|
-
iterator_status = True
|
121
|
-
|
122
|
-
if isinstance(request_item, Request):
|
123
|
-
iterator_status = False
|
124
|
-
start_time = time.time()
|
125
|
-
download_iterators = self.download(request_item)
|
126
|
-
if not isgenerator(download_iterators):
|
127
|
-
raise TypeError("download function isn't a generator")
|
128
|
-
|
129
|
-
for download_item in download_iterators:
|
130
|
-
iterator_status = True
|
131
|
-
if isinstance(download_item, Response):
|
132
|
-
iterator_status = False
|
133
|
-
logger.info(LogTemplate.download_info.format(
|
134
|
-
detail=seed_detail_log_info,
|
135
|
-
retry=seed.params.retry,
|
136
|
-
priority=seed.params.priority,
|
137
|
-
seed_version=seed.params.seed_version,
|
138
|
-
identifier=seed.identifier or "",
|
139
|
-
status=download_item.response,
|
140
|
-
response=LogTemplate.log_info(download_item.to_dict)
|
141
|
-
))
|
142
|
-
if isinstance(download_item, Response):
|
143
|
-
end_time = time.time()
|
144
|
-
self.loghub_dot.build(
|
145
|
-
topic=f"{self.project}:{self.task}",
|
146
|
-
cost_time=end_time - start_time,
|
147
|
-
**download_item.to_dict
|
148
|
-
)
|
149
|
-
parse_iterators = self.parse(download_item)
|
150
|
-
if not isgenerator(parse_iterators):
|
151
|
-
raise TypeError("parse function isn't a generator")
|
152
|
-
for parse_item in parse_iterators:
|
153
|
-
iterator_status = True
|
154
|
-
if isinstance(parse_item, Response):
|
155
|
-
raise TypeError("upload_item can't be a Response instance")
|
156
|
-
self.distribute(parse_item, seed)
|
157
|
-
else:
|
158
|
-
self.distribute(download_item, seed)
|
159
|
-
else:
|
160
|
-
self.distribute(request_item, seed)
|
161
|
-
|
162
|
-
if not iterator_status:
|
163
|
-
raise ValueError("request/download/parse function yield value error!")
|
164
|
-
except Exception as e:
|
165
|
-
logger.info(LogTemplate.download_exception.format(
|
166
|
-
detail=seed_detail_log_info,
|
167
|
-
retry=seed.params.retry,
|
168
|
-
priority=seed.params.priority,
|
169
|
-
seed_version=seed.params.seed_version,
|
170
|
-
identifier=seed.identifier or "",
|
171
|
-
exception=''.join(traceback.format_exception(type(e), e, e.__traceback__))
|
172
|
-
))
|
173
|
-
seed.params.retry += 1
|
174
|
-
# self._todo.push(seed)
|
175
|
-
self._set_seed(seed)
|
176
|
-
time.sleep(self.time_sleep * seed.params.retry)
|
177
|
-
finally:
|
178
|
-
time.sleep(0.1)
|
179
|
-
logger.info("spider thread close")
|
180
|
-
|
181
|
-
def run(self):
|
182
|
-
for index in range(self.thread_num):
|
183
|
-
threading.Thread(name=f"spider_{index}", target=self.spider).start()
|
184
|
-
|
cobweb_/db/__init__.py
DELETED