1 ! Copyright (C) 2008 Doug Coleman.
2 ! See http://factorcode.org/license.txt for BSD license.
3 USING: accessors fry html.parser html.parser.analyzer
4 http.client kernel tools.time sets assocs sequences
5 concurrency.combinators io threads namespaces math multiline
6 math.parser inspector urls logging combinators.short-circuit
7 continuations calendar prettyprint dlists deques locals
8 spider.unique-deque combinators concurrency.semaphores ;
11 TUPLE: spider base count max-count sleep max-depth initial-links
12 filters spidered todo nonmatching quiet currently-spidering
13 #threads semaphore follow-robots? robots ;
15 TUPLE: spider-result url depth headers
16 fetched-in parsed-html links processed-in fetched-at ;
18 : <spider> ( base -- spider )
22 over >>currently-spidering
23 swap 0 <unique-deque> [ push-url ] keep >>todo
24 <unique-deque> >>nonmatching
29 1 [ >>#threads ] [ <semaphore> >>semaphore ] bi ;
31 : <spider-result> ( url depth -- spider-result )
38 : apply-filters ( links spider -- links' )
40 '[ [ _ 1&& ] filter ] call( seq -- seq' )
43 : push-links ( links level unique-deque -- )
44 '[ _ _ push-url ] each ;
46 : add-todo ( links level spider -- )
49 : add-nonmatching ( links level spider -- )
50 nonmatching>> push-links ;
52 : filter-base-links ( spider spider-result -- base-links nonmatching-links )
53 [ base>> host>> ] [ links>> members ] bi*
54 [ host>> = ] with partition ;
56 :: add-spidered ( spider spider-result -- )
57 spider [ 1 + ] change-count drop
59 spider-result dup url>>
60 spider spidered>> set-at
62 spider spider-result filter-base-links :> ( matching nonmatching )
63 spider-result depth>> 1 + :> depth
65 nonmatching depth spider add-nonmatching
67 matching spider apply-filters depth spider add-todo ;
69 : normalize-hrefs ( base links -- links' )
70 [ derive-url ] with map ;
72 : print-spidering ( spider-result -- )
73 [ url>> ] [ depth>> ] bi
74 "depth: " write number>string write
75 ", spidering: " write . yield ;
77 :: fill-spidered-result ( spider spider-result -- )
78 f spider-result url>> spider spidered>> set-at
79 [ spider-result url>> http-get ] benchmark :> ( headers html fetched-in )
82 spider currently-spidering>>
83 over find-all-links normalize-hrefs
84 ] benchmark :> ( parsed-html links processed-in )
87 fetched-in >>fetched-in
88 parsed-html >>parsed-html
90 processed-in >>processed-in
91 now >>fetched-at drop ;
93 :: spider-page ( spider spider-result -- )
94 spider quiet>> [ spider-result print-spidering ] unless
95 spider spider-result fill-spidered-result
96 spider quiet>> [ spider-result describe ] unless
97 spider spider-result add-spidered ;
99 \ spider-page ERROR add-error-logging
101 : spider-sleep ( spider -- ) sleep>> [ sleep ] when* ;
103 : queue-initial-links ( spider -- spider )
104 [ [ currently-spidering>> ] [ initial-links>> ] bi normalize-hrefs 0 ]
108 : spider-page? ( spider -- ? )
110 [ todo>> deque>> deque-empty? not ]
111 [ [ todo>> peek-url depth>> ] [ max-depth>> ] bi < ]
112 [ [ count>> ] [ max-count>> ] bi < ]
115 : setup-next-url ( spider -- spider spider-result )
116 dup todo>> peek-url url>> >>currently-spidering
117 dup todo>> pop-url [ url>> ] [ depth>> ] bi <spider-result> ;
119 : spider-next-page ( spider -- )
120 setup-next-url spider-page ;
124 : run-spider-loop ( spider -- )
126 [ spider-next-page ] [ spider-sleep ] [ run-spider-loop ] tri
131 : run-spider ( spider -- spider )
134 [ run-spider-loop ] keep