1
|
|
|
<?php |
2
|
|
|
declare(strict_types = 1); |
3
|
|
|
namespace hexydec\agentzero; |
4
|
|
|
|
5
|
|
|
class crawlers { |
6
|
|
|
|
7
|
|
|
public static function getApp(string $value, array $data = []) : array { |
8
|
|
|
if (!\str_contains($value, '://')) { // bot will be in the URL |
9
|
|
|
$parts = \explode('/', $value, 2); |
10
|
|
|
|
11
|
|
|
// process version |
12
|
|
|
if (!empty($parts[1])) { |
13
|
|
|
$parts[1] = \ltrim($parts[1], 'v'); |
14
|
|
|
$parts[1] = \substr($parts[1], 0, \strspn($parts[1], '0123456789.')); |
15
|
|
|
} |
16
|
|
|
return \array_merge([ |
17
|
|
|
'type' => 'robot', |
18
|
|
|
'app' => $parts[0], |
19
|
|
|
'appversion' => empty($parts[1]) ? null : $parts[1] |
20
|
|
|
], $data); |
21
|
|
|
} |
22
|
|
|
return []; |
23
|
|
|
} |
24
|
|
|
|
25
|
|
|
public static function get() { |
26
|
|
|
$fn = [ |
27
|
|
|
'search' => fn (string $value) : array => self::getApp($value, ['category' => 'search']), |
28
|
|
|
'ads' => fn (string $value) : array => self::getApp($value, ['category' => 'ads']), |
29
|
|
|
'validator' => fn (string $value) : array => self::getApp($value, ['category' => 'validator']), |
30
|
|
|
'feed' => fn (string $value) : array => self::getApp($value, ['category' => 'feed']), |
31
|
|
|
'crawler' => fn (string $value) : array => self::getApp($value, ['category' => 'crawler']), |
32
|
|
|
'monitor' => fn (string $value) : array => self::getApp($value, ['category' => 'monitor']), |
33
|
|
|
'scraper' => fn (string $value) : array => self::getApp($value, ['category' => 'scraper']), |
34
|
|
|
'map' => function (string $value, int $i, array $tokens) : ?array { |
35
|
|
|
if (!\str_contains($value, '://')) { // bot will be in the URL |
36
|
|
|
$parts = \explode('/', $value, 2); |
37
|
|
|
|
38
|
|
|
// special case |
39
|
|
|
if (\strcasecmp($parts[0], 'Spider') === 0) { |
40
|
|
|
$ua = \implode(' ', $tokens); |
41
|
|
|
if (\mb_stripos($ua, 'Screaming Frog SEO Spider') === 0) { |
42
|
|
|
$parts[0] = 'Screaming Frog SEO Spider'; |
43
|
|
|
} elseif (\mb_stripos($ua, 'Sogou web spider') === 0) { |
44
|
|
|
$parts[0] = 'Sogou web spider'; |
45
|
|
|
} |
46
|
|
|
} |
47
|
|
|
$category = [ |
48
|
|
|
'yacybot' => 'search', |
49
|
|
|
'Googlebot' => 'search', |
50
|
|
|
'Googlebot-Mobile' => 'search', |
51
|
|
|
'Googlebot-Image' => 'search', |
52
|
|
|
'Googlebot-Video' => 'search', |
53
|
|
|
'Googlebot-News' => 'search', |
54
|
|
|
'Storebot-Google' => 'search', |
55
|
|
|
'AdsBot-Google' => 'ads', |
56
|
|
|
'AdsBot-Google-Mobile' => 'ads', |
57
|
|
|
'Bingbot' => 'search', |
58
|
|
|
'bingbot' => 'search', |
59
|
|
|
'adidxbot' => 'ads', |
60
|
|
|
'DuckDuckBot' => 'search', |
61
|
|
|
'DuckDuckGo-Favicons-Bot' => 'search', |
62
|
|
|
'coccocbot-image' => 'search', |
63
|
|
|
'coccocbot-web' => 'search', |
64
|
|
|
'Baiduspider' => 'search', |
65
|
|
|
'Applebot' => 'search', |
66
|
|
|
'YandexBot' => 'search', |
67
|
|
|
'MJ12bot' => 'search', |
68
|
|
|
'Mail.RU_Bot' => 'search', |
69
|
|
|
'HaosouSpider' => 'search', |
70
|
|
|
'360Spider' => 'search', |
71
|
|
|
'Exabot' => 'search', |
72
|
|
|
'Sogou web spider' => 'search', |
73
|
|
|
'UptimeRobot' => 'monitor', |
74
|
|
|
'PetalBot' => 'search', |
75
|
|
|
'Screaming Frog SEO Spider' => 'crawler', |
76
|
|
|
'Twitterbot' => 'feed', |
77
|
|
|
'Xbot' => 'feed', |
78
|
|
|
'Discordbot' => 'feed', |
79
|
|
|
'PRTGCloudBot' => 'monitor', |
80
|
|
|
'Bytespider' => 'search', |
81
|
|
|
'LinkedInBot' => 'feed', |
82
|
|
|
'PaperLiBot' => 'feed', |
83
|
|
|
'bitlybot' => 'feed', |
84
|
|
|
'TinEye-bot' => 'search', |
85
|
|
|
'Pinterestbot' => 'feed', |
86
|
|
|
'WebCrawler' => 'crawler', |
87
|
|
|
'webprosbot' => 'crawler', |
88
|
|
|
'GuzzleHttp' => 'scraper', |
89
|
|
|
'TelegramBot' => 'feed', |
90
|
|
|
'Ruby' => 'scraper', |
91
|
|
|
'SEMrushBot' => 'crawler', |
92
|
|
|
'Mediatoolkitbot' => 'crawler' |
93
|
|
|
]; |
94
|
|
|
return self::getApp($value, [ |
95
|
|
|
'category' => $category[$parts[0]] ?? null, |
96
|
|
|
'app' => $parts[0] |
97
|
|
|
]); |
98
|
|
|
} |
99
|
|
|
return null; |
100
|
|
|
} |
101
|
|
|
]; |
102
|
|
|
return [ |
103
|
|
|
'Yahoo! Slurp' => [ |
104
|
|
|
'match' => 'exact', |
105
|
|
|
'categories' => $fn['search'] |
106
|
|
|
], |
107
|
|
|
'facebookexternalhit/' => [ |
108
|
|
|
'match' => 'start', |
109
|
|
|
'categories' => $fn['feed'] |
110
|
|
|
], |
111
|
|
|
'Google-Site-Verification/' => [ |
112
|
|
|
'match' => 'start', |
113
|
|
|
'categories' => $fn['validator'] |
114
|
|
|
], |
115
|
|
|
'Google-InspectionTool/' => [ |
116
|
|
|
'match' => 'start', |
117
|
|
|
'categories' => $fn['search'] |
118
|
|
|
], |
119
|
|
|
'Mediapartners-Google' => [ |
120
|
|
|
'match' => 'start', |
121
|
|
|
'categories' => $fn['search'] |
122
|
|
|
], |
123
|
|
|
'FeedFetcher-Google' => [ |
124
|
|
|
'match' => 'exact', |
125
|
|
|
'categories' => $fn['feed'] |
126
|
|
|
], |
127
|
|
|
'GoogleProducer' => [ |
128
|
|
|
'match' => 'exact', |
129
|
|
|
'categories' => $fn['feed'] |
130
|
|
|
], |
131
|
|
|
'CFNetwork/' => [ |
132
|
|
|
'match' => 'start', |
133
|
|
|
'categories' => $fn['feed'] |
134
|
|
|
], |
135
|
|
|
'Siteimprove.com' => [ |
136
|
|
|
'match' => 'any', |
137
|
|
|
'categories' => $fn['crawler'] |
138
|
|
|
], |
139
|
|
|
'Google Page Speed Insights' => [ |
140
|
|
|
'match' => 'exact', |
141
|
|
|
'categories' => $fn['validator'] |
142
|
|
|
], |
143
|
|
|
'Qwantify' => [ |
144
|
|
|
'match' => 'start', |
145
|
|
|
'categories' => $fn['search'] |
146
|
|
|
], |
147
|
|
|
'okhttp' => [ |
148
|
|
|
'match' => 'start', |
149
|
|
|
'categories' => $fn['scraper'] |
150
|
|
|
], |
151
|
|
|
'python' => [ |
152
|
|
|
'match' => 'start', |
153
|
|
|
'categories' => $fn['scraper'] |
154
|
|
|
], |
155
|
|
|
'Nessus' => [ |
156
|
|
|
'match' => 'start', |
157
|
|
|
'categories' => $fn['monitor'] |
158
|
|
|
], |
159
|
|
|
'Chrome-Lighthouse' => [ |
160
|
|
|
'match' => 'start', |
161
|
|
|
'categories' => $fn['validator'] |
162
|
|
|
], |
163
|
|
|
'PingdomTMS/' => [ |
164
|
|
|
'match' => 'start', |
165
|
|
|
'categories' => $fn['monitor'] |
166
|
|
|
], |
167
|
|
|
'Pingdom.com' => [ |
168
|
|
|
'match' => 'start', |
169
|
|
|
'categories' => function (string $value) : array { |
170
|
|
|
$version = \explode('_', \trim($value, '_')); |
171
|
|
|
return [ |
172
|
|
|
'type' => 'robot', |
173
|
|
|
'category' => 'monitor', |
174
|
|
|
'app' => 'Pingdom.com Bot', |
175
|
|
|
'appversion' => \end($version) |
176
|
|
|
]; |
177
|
|
|
} |
178
|
|
|
], |
179
|
|
|
'proximic' => [ |
180
|
|
|
'match' => 'exact', |
181
|
|
|
'categories' => $fn['ads'] |
182
|
|
|
], |
183
|
|
|
'WordPress' => [ |
184
|
|
|
'match' => 'start', |
185
|
|
|
'categories' => $fn['monitor'] |
186
|
|
|
], |
187
|
|
|
'PRTG Network Monitor' => [ |
188
|
|
|
'match' => 'exact', |
189
|
|
|
'categories' => $fn['monitor'] |
190
|
|
|
], |
191
|
|
|
'Site24x7' => [ |
192
|
|
|
'match' => 'exact', |
193
|
|
|
'categories' => $fn['monitor'] |
194
|
|
|
], |
195
|
|
|
'adbeat.com' => [ |
196
|
|
|
'match' => 'start', |
197
|
|
|
'categories' => fn (string $value) : array => [ |
198
|
|
|
'type' => 'robot', |
199
|
|
|
'category' => 'ads', |
200
|
|
|
'app' => 'Adbeat', |
201
|
|
|
'url' => $value |
202
|
|
|
] |
203
|
|
|
], |
204
|
|
|
'MicrosoftPreview/' => [ |
205
|
|
|
'match' => 'start', |
206
|
|
|
'categories' => $fn['feed'] |
207
|
|
|
], |
208
|
|
|
'Let\'s Encrypt validation server' => [ |
209
|
|
|
'match' => 'exact', |
210
|
|
|
'categories' => $fn['validator'] |
211
|
|
|
], |
212
|
|
|
'Expanse' => [ |
213
|
|
|
'match' => 'start', |
214
|
|
|
'categories' => $fn['crawler'] |
215
|
|
|
], |
216
|
|
|
'WhatsApp/' => [ |
217
|
|
|
'match' => 'start', |
218
|
|
|
'categories' => $fn['feed'] |
219
|
|
|
], |
220
|
|
|
'Apache-HttpClient/' => [ |
221
|
|
|
'match' => 'start', |
222
|
|
|
'categories' => $fn['scraper'] |
223
|
|
|
], |
224
|
|
|
'eCairn-Grabber/' => [ |
225
|
|
|
'match' => 'start', |
226
|
|
|
'categories' => $fn['scraper'] |
227
|
|
|
], |
228
|
|
|
'SEOkicks' => [ |
229
|
|
|
'match' => 'exact', |
230
|
|
|
'categories' => $fn['crawler'] |
231
|
|
|
], |
232
|
|
|
'PostmanRuntime/' => [ |
233
|
|
|
'match' => 'start', |
234
|
|
|
'categories' => $fn['scraper'] |
235
|
|
|
], |
236
|
|
|
'axios/' => [ |
237
|
|
|
'match' => 'start', |
238
|
|
|
'categories' => $fn['scraper'] |
239
|
|
|
], |
240
|
|
|
'Rogerbot/' => [ |
241
|
|
|
'match' => 'start', |
242
|
|
|
'categories' => $fn['crawler'] |
243
|
|
|
], |
244
|
|
|
'Go-http-client/' => [ |
245
|
|
|
'match' => 'start', |
246
|
|
|
'categories' => $fn['scraper'] |
247
|
|
|
], |
248
|
|
|
'DashLinkPreviews/' => [ |
249
|
|
|
'match' => 'start', |
250
|
|
|
'categories' => $fn['feed'] |
251
|
|
|
], |
252
|
|
|
'Microsoft Office' => [ |
253
|
|
|
'match' => 'start', |
254
|
|
|
'categories' => function (string $value, int $i, array $tokens) : array { |
255
|
|
|
$data = [ |
256
|
|
|
'type' => 'robot', |
257
|
|
|
'category' => 'feed' |
258
|
|
|
]; |
259
|
|
|
if (\str_contains($value, '/')) { |
260
|
|
|
foreach (\array_slice($tokens, $i + 1) AS $item) { |
261
|
|
|
if (\str_starts_with($item, 'Microsoft ')) { |
262
|
|
|
$parts = \explode(' ', $item); |
263
|
|
|
$data['app'] = $parts[0].' '.$parts[1]; |
264
|
|
|
if (isset($parts[2])) { |
265
|
|
|
$data['appversion'] = $parts[2]; |
266
|
|
|
} |
267
|
|
|
break; |
268
|
|
|
} |
269
|
|
|
} |
270
|
|
|
if (!isset($data['app'])) { |
271
|
|
|
$parts = \explode('/', $value, 2); |
272
|
|
|
$data['app'] = $parts[0]; |
273
|
|
|
if (!isset($data['appversion'])) { |
274
|
|
|
$data['appversion'] = $parts[1]; |
275
|
|
|
} |
276
|
|
|
} |
277
|
|
|
} else { |
278
|
|
|
$parts = \explode(' ', $value); |
279
|
|
|
$data['app'] = $parts[0].' '.$parts[1].' '.$parts[2]; |
280
|
|
|
$data['appversion'] = $parts[3] ?? null; |
281
|
|
|
} |
282
|
|
|
return $data; |
283
|
|
|
} |
284
|
|
|
], |
285
|
|
|
'PycURL/' => [ |
286
|
|
|
'match' => 'start', |
287
|
|
|
'categories' => $fn['scraper'] |
288
|
|
|
], |
289
|
|
|
'lua-resty-http/' => [ |
290
|
|
|
'match' => 'start', |
291
|
|
|
'categories' => $fn['scraper'] |
292
|
|
|
], |
293
|
|
|
'Snapchat/' => [ |
294
|
|
|
'match' => 'start', |
295
|
|
|
'categories' => $fn['feed'] |
296
|
|
|
], |
297
|
|
|
'HTTPClient/' => [ |
298
|
|
|
'match' => 'start', |
299
|
|
|
'categories' => $fn['scraper'] |
300
|
|
|
], |
301
|
|
|
'WhatsApp/' => [ |
302
|
|
|
'match' => 'any', |
303
|
|
|
'categories' => $fn['feed'] |
304
|
|
|
], |
305
|
|
|
'Hootsuite-Authoring/' => [ |
306
|
|
|
'match' => 'start', |
307
|
|
|
'categories' => $fn['feed'] |
308
|
|
|
], |
309
|
|
|
'ApacheBench/' => [ |
310
|
|
|
'match' => 'start', |
311
|
|
|
'categories' => $fn['validator'] |
312
|
|
|
], |
313
|
|
|
'Asana/' => [ |
314
|
|
|
'match' => 'start', |
315
|
|
|
'categories' => $fn['feed'] |
316
|
|
|
], |
317
|
|
|
'Java/' => [ |
318
|
|
|
'match' => 'start', |
319
|
|
|
'categories' => $fn['scraper'] |
320
|
|
|
], |
321
|
|
|
'curl/' => [ |
322
|
|
|
'match' => 'start', |
323
|
|
|
'categories' => $fn['scraper'] |
324
|
|
|
], |
325
|
|
|
'feed' => [ |
326
|
|
|
'match' => 'any', |
327
|
|
|
'categories' => $fn['feed'] |
328
|
|
|
], |
329
|
|
|
'spider' => [ |
330
|
|
|
'match' => 'any', |
331
|
|
|
'categories' => $fn['map'] |
332
|
|
|
], |
333
|
|
|
'crawler' => [ |
334
|
|
|
'match' => 'any', |
335
|
|
|
'categories' => $fn['map'] |
336
|
|
|
], |
337
|
|
|
'bot' => [ |
338
|
|
|
'match' => 'any', |
339
|
|
|
'categories' => $fn['map'] |
340
|
|
|
], |
341
|
|
|
]; |
342
|
|
|
} |
343
|
|
|
} |