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 ;
10 TUPLE: spider base count max-count sleep max-depth initial-links
11 filters spidered todo nonmatching quiet ;
13 TUPLE: spider-result url depth headers fetch-time parsed-html
14 links processing-time timestamp ;
16 TUPLE: todo-url url depth ;
18 : <todo-url> ( url depth -- todo-url )
23 TUPLE: unique-deque assoc deque ;
25 : <unique-deque> ( -- unique-deque )
26 H{ } clone <dlist> unique-deque boa ;
28 : push-url ( url depth unique-deque -- )
30 [ [ [ t ] dip url>> ] [ assoc>> ] bi* set-at ]
31 [ deque>> push-back ] 2bi ;
33 : pop-url ( unique-deque -- todo-url ) deque>> pop-front ;
35 : peek-url ( unique-deque -- todo-url ) deque>> peek-front ;
37 : <spider> ( base -- spider )
41 swap 0 <unique-deque> [ push-url ] keep >>todo
42 <unique-deque> >>nonmatching
46 H{ } clone >>spidered ;
50 : apply-filters ( links spider -- links' )
51 filters>> [ '[ [ _ 1&& ] filter ] call( seq -- seq' ) ] when* ;
53 : push-links ( links level unique-deque -- )
54 '[ _ _ push-url ] each ;
56 : add-todo ( links level spider -- )
59 : add-nonmatching ( links level spider -- )
60 nonmatching>> push-links ;
62 : filter-base-links ( spider spider-result -- base-links nonmatching-links )
63 [ base>> host>> ] [ links>> prune ] bi*
64 [ host>> = ] with partition ;
66 : add-spidered ( spider spider-result -- )
67 [ [ 1+ ] change-count ] dip
68 2dup [ spidered>> ] [ dup url>> ] bi* rot set-at
69 [ filter-base-links ] 2keep
72 [ tuck [ apply-filters ] 2dip add-todo ] 2bi ;
74 : normalize-hrefs ( links spider -- links' )
76 base>> swap [ derive-url ] with map ;
78 : print-spidering ( url depth -- )
79 "depth: " write number>string write
80 ", spidering: " write . yield ;
82 :: new-spidered-result ( spider url depth -- spider-result )
83 f url spider spidered>> set-at
84 [ url http-get ] benchmark :> fetch-time :> html :> headers
86 html parse-html [ ] [ find-hrefs spider normalize-hrefs ] bi
87 ] benchmark :> processing-time :> links :> parsed-html
88 url depth headers fetch-time parsed-html links processing-time
89 now spider-result boa ;
91 :: spider-page ( spider url depth -- )
92 spider quiet>> [ url depth print-spidering ] unless
93 spider url depth new-spidered-result :> spidered-result
94 spider quiet>> [ spidered-result describe ] unless
95 spider spidered-result add-spidered ;
97 \ spider-page ERROR add-error-logging
99 : spider-sleep ( spider -- )
100 sleep>> [ sleep ] when* ;
102 :: queue-initial-links ( spider -- spider )
103 spider initial-links>> spider normalize-hrefs 0 spider add-todo spider ;
105 : spider-page? ( spider -- ? )
107 [ todo>> deque>> deque-empty? not ]
108 [ [ todo>> peek-url depth>> ] [ max-depth>> ] bi < ]
109 [ [ count>> ] [ max-count>> ] bi < ]
112 : setup-next-url ( spider -- spider url depth )
113 dup todo>> pop-url [ url>> ] [ depth>> ] bi ;
115 : spider-next-page ( spider -- )
116 setup-next-url spider-page ;
120 : run-spider-loop ( spider -- )
122 [ spider-next-page ] [ run-spider-loop ] bi
127 : run-spider ( spider -- spider )
129 queue-initial-links [ run-spider-loop ] keep