1 ! Copyright (C) 2008 Doug Coleman.
2 ! See http://factorcode.org/license.txt for BSD license.
3 USING: accessors assocs combinators combinators.short-circuit
4 fry html.parser http.client io kernel locals math sequences
5 sets splitting unicode.case unicode.categories urls
6 urls.encoding shuffle ;
7 IN: html.parser.analyzer
9 : scrape-html ( url -- headers vector )
12 : attribute ( tag string -- obj/f )
13 swap attributes>> [ at ] [ drop f ] if* ;
15 : attribute* ( tag string -- obj ? )
16 swap attributes>> [ at* ] [ drop f f ] if* ;
18 : attribute? ( tag string -- obj )
19 swap attributes>> [ key? ] [ drop f ] if* ;
21 : find-all ( seq quot -- alist )
22 [ <enum> >alist ] [ '[ second @ ] ] bi* filter ; inline
24 : loopn-index ( n quot -- )
25 [ iota ] [ '[ @ not ] ] bi* find 2drop ; inline
28 [ drop ] prepose loopn-index ; inline
30 ERROR: undefined-find-nth m n seq quot ;
32 : check-trivial-find ( m n seq quot -- m n seq quot )
33 pick 0 = [ undefined-find-nth ] when ; inline
35 : find-nth-from ( m n seq quot -- i/f elt/f )
36 check-trivial-find [ f ] 3dip '[
37 drop _ _ find-from [ dup [ 1 + ] when ] dip over
38 ] loopn [ dup [ 1 - ] when ] dip ; inline
40 : find-nth ( n seq quot -- i/f elt/f )
41 [ 0 ] 3dip find-nth-from ; inline
43 : find-last-nth-from ( m n seq quot -- i/f elt/f )
44 check-trivial-find [ f ] 3dip '[
45 drop _ _ find-last-from [ dup [ 1 - ] when ] dip over
46 ] loopn [ dup [ 1 + ] when ] dip ; inline
48 : find-last-nth ( n seq quot -- i/f elt/f )
49 [ [ nip length 1 - ] [ ] 2bi ] dip find-last-nth-from ; inline
51 : find-first-name ( vector string -- i/f tag/f )
52 >lower '[ name>> _ = ] find ; inline
54 : find-matching-close ( vector string -- i/f tag/f )
56 '[ [ name>> _ = ] [ closing?>> ] bi and ] find ; inline
58 : find-between* ( vector i/f tag/f -- vector )
60 [ tail-slice ] [ name>> ] bi*
61 dupd find-matching-close drop [ 1 + ] [ 1 ] if*
67 : find-between ( vector i/f tag/f -- vector )
68 find-between* dup length 3 >= [
69 [ rest-slice but-last-slice ] keep like
72 : find-between-first ( vector string -- vector' )
73 dupd find-first-name find-between ; inline
75 : find-between-all ( vector quot -- seq )
77 '[ _ [ closing?>> not ] bi and ] find-all
78 [ first2 find-between* ] with map ; inline
80 : remove-blank-text ( vector -- vector' )
83 [ text>> [ blank? ] all? not ] [ drop t ] if
86 : trim-text ( vector -- vector' )
89 [ [ [ blank? ] trim ] change-text ] when
92 : find-by-id ( vector id -- vector' elt/f )
93 '[ "id" attribute _ = ] find ;
95 : find-by-class ( vector id -- vector' elt/f )
96 '[ "class" attribute _ = ] find ;
98 : find-by-name ( vector string -- vector elt/f )
99 >lower '[ name>> _ = ] find ;
101 : find-by-id-between ( vector string -- vector' )
103 '[ "id" attribute _ = ] find find-between* ;
105 : find-by-class-between ( vector string -- vector' )
107 '[ "class" attribute _ = ] find find-between* ;
109 : find-by-class-id-between ( vector class id -- vector' )
112 [ "class" attribute _ = ]
113 [ "id" attribute _ = ] bi and
119 : find-by-attribute-key ( vector key -- vector' elt/? )
121 [ attributes>> at _ = ] filter sift ;
123 : find-by-attribute-key-value ( vector value key -- vector' )
125 [ attributes>> at over = ] with filter nip sift ;
127 : find-first-attribute-key-value ( vector value key -- i/f tag/f )
129 [ attributes>> at over = ] with find rot drop ;
131 : tag-link ( tag -- link/f ) "href" attribute ;
133 : find-links ( vector -- vector' )
134 [ { [ name>> "a" = ] [ "href" attribute ] } 1&& ]
137 : find-images ( vector -- vector' )
144 values [ "src" attribute ] map ;
146 : find-by-text ( seq quot -- tag )
147 [ dup name>> text = ] prepose find drop ; inline
149 : find-opening-tags-by-name ( name seq -- seq )
150 [ { [ name>> = ] [ closing?>> not ] } 1&& ] with find-all ;
152 : href-contains? ( str tag -- ? )
153 "href" attribute* [ subseq? ] [ 2drop f ] if ;
155 : find-hrefs ( vector -- vector' )
157 [ [ { [ name>> "a" = ] [ "href" attribute? ] } 1&& ] filter ] map sift
158 [ [ "href" attribute ] map ] map concat [ >url ] map ;
160 : find-frame-links ( vector -- vector' )
161 [ name>> "frame" = ] find-between-all
162 [ [ "src" attribute ] map sift ] map concat sift
165 : find-all-links ( vector -- vector' )
166 [ find-hrefs ] [ find-frame-links ] bi union ;
168 : find-forms ( vector -- vector' )
169 "form" over find-opening-tags-by-name
170 swap [ [ first2 ] dip find-between* ] curry map
171 [ [ name>> { "form" "input" } member? ] filter ] map ;
173 : find-html-objects ( vector string -- vector' )
174 dupd find-opening-tags-by-name
175 [ first2 find-between* ] curry map ;
177 : form-action ( vector -- string )
178 [ name>> "form" = ] find nip "action" attribute ;
180 : hidden-form-values ( vector -- strings )
181 [ "type" attribute "hidden" = ] filter ;
186 [ bl bl bl bl [ write "=" write ] [ write bl ] bi* nl ] assoc-each ;
188 : form. ( vector -- )
189 [ closing?>> not ] filter
192 { [ dup name>> "form" = ]
193 [ "form action: " write "action" attribute print ] }
194 { [ dup name>> "input" = ] [ input. ] }
199 : query>assoc* ( str -- hash )
200 "?" split1 nip query>assoc ;
202 : html-class? ( tag string -- ? )
203 swap "class" attribute = ;
205 : html-id? ( tag string -- ? )
206 swap "id" attribute = ;
208 : opening-tag? ( tag -- ? )
211 TUPLE: link attributes clickable ;
213 : <link> ( vector -- link )
214 [ first attributes>> ]
215 [ [ name>> { text "img" } member? ] filter ] bi
218 : link. ( vector -- )
219 [ "href" attribute write nl ]
220 [ clickable>> [ bl bl text>> print ] each nl ] bi ;