2008-09-25 03:06:49 -04:00
|
|
|
! Copyright (C) 2008 Doug Coleman.
|
|
|
|
! See http://factorcode.org/license.txt for BSD license.
|
2010-08-29 14:22:11 -04:00
|
|
|
USING: accessors assocs combinators combinators.short-circuit
|
|
|
|
fry html.parser http.client io kernel locals math sequences
|
|
|
|
sets splitting unicode.case unicode.categories urls
|
2010-08-29 16:24:55 -04:00
|
|
|
urls.encoding shuffle ;
|
2007-12-04 15:14:33 -05:00
|
|
|
IN: html.parser.analyzer
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-09-29 21:21:12 -04:00
|
|
|
: scrape-html ( url -- headers vector )
|
|
|
|
http-get parse-html ;
|
2008-05-19 19:58:56 -04:00
|
|
|
|
2010-08-29 14:22:11 -04:00
|
|
|
: attribute ( tag string -- obj/f )
|
|
|
|
swap attributes>> [ at ] [ drop f ] if* ;
|
|
|
|
|
|
|
|
: attribute* ( tag string -- obj ? )
|
|
|
|
swap attributes>> [ at* ] [ drop f f ] if* ;
|
|
|
|
|
|
|
|
: attribute? ( tag string -- obj )
|
|
|
|
swap attributes>> [ key? ] [ drop f ] if* ;
|
|
|
|
|
2008-09-25 03:02:47 -04:00
|
|
|
: find-all ( seq quot -- alist )
|
|
|
|
[ <enum> >alist ] [ '[ second @ ] ] bi* filter ; inline
|
2008-03-06 18:09:48 -05:00
|
|
|
|
2010-08-29 16:24:55 -04:00
|
|
|
: loopn-index ( n quot -- )
|
|
|
|
[ iota ] [ '[ @ not ] ] bi* find 2drop ; inline
|
|
|
|
|
|
|
|
: loopn ( n quot -- )
|
|
|
|
[ drop ] prepose loopn-index ; inline
|
|
|
|
|
|
|
|
ERROR: undefined-find-nth m n seq quot ;
|
|
|
|
|
|
|
|
: check-trivial-find ( m n seq quot -- m n seq quot )
|
|
|
|
pick 0 = [ undefined-find-nth ] when ; inline
|
|
|
|
|
|
|
|
: find-nth-from ( m n seq quot -- i/f elt/f )
|
|
|
|
check-trivial-find [ f ] 3dip '[
|
|
|
|
drop _ _ find-from [ dup [ 1 + ] when ] dip over
|
|
|
|
] loopn [ dup [ 1 - ] when ] dip ; inline
|
|
|
|
|
|
|
|
: find-nth ( n seq quot -- i/f elt/f )
|
|
|
|
[ 0 ] 3dip find-nth-from ; inline
|
|
|
|
|
|
|
|
: find-last-nth-from ( m n seq quot -- i/f elt/f )
|
|
|
|
check-trivial-find [ f ] 3dip '[
|
|
|
|
drop _ _ find-last-from [ dup [ 1 - ] when ] dip over
|
|
|
|
] loopn [ dup [ 1 + ] when ] dip ; inline
|
|
|
|
|
|
|
|
: find-last-nth ( n seq quot -- i/f elt/f )
|
|
|
|
[ [ nip length 1 - ] [ ] 2bi ] dip find-last-nth-from ; inline
|
2008-03-06 18:09:48 -05:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-first-name ( vector string -- i/f tag/f )
|
|
|
|
>lower '[ name>> _ = ] find ; inline
|
2008-09-25 03:02:47 -04:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-matching-close ( vector string -- i/f tag/f )
|
|
|
|
>lower
|
|
|
|
'[ [ name>> _ = ] [ closing?>> ] bi and ] find ; inline
|
2008-09-25 03:02:47 -04:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-between* ( vector i/f tag/f -- vector )
|
|
|
|
over integer? [
|
|
|
|
[ tail-slice ] [ name>> ] bi*
|
2010-08-29 14:22:11 -04:00
|
|
|
dupd find-matching-close drop [ 1 + ] [ 1 ] if*
|
|
|
|
head
|
2008-03-06 18:09:48 -05:00
|
|
|
] [
|
2008-09-25 03:02:47 -04:00
|
|
|
3drop V{ } clone
|
2008-05-24 00:34:10 -04:00
|
|
|
] if ; inline
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: find-between ( vector i/f tag/f -- vector )
|
2008-09-25 03:02:47 -04:00
|
|
|
find-between* dup length 3 >= [
|
|
|
|
[ rest-slice but-last-slice ] keep like
|
|
|
|
] when ; inline
|
2008-03-06 18:09:48 -05:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-between-first ( vector string -- vector' )
|
|
|
|
dupd find-first-name find-between ; inline
|
2008-09-25 03:02:47 -04:00
|
|
|
|
|
|
|
: find-between-all ( vector quot -- seq )
|
2008-10-19 16:52:59 -04:00
|
|
|
dupd
|
|
|
|
'[ _ [ closing?>> not ] bi and ] find-all
|
2009-03-22 20:25:55 -04:00
|
|
|
[ first2 find-between* ] with map ; inline
|
2008-03-06 18:09:48 -05:00
|
|
|
|
2007-11-29 12:18:46 -05:00
|
|
|
: remove-blank-text ( vector -- vector' )
|
2007-09-20 18:09:08 -04:00
|
|
|
[
|
2008-09-25 03:02:47 -04:00
|
|
|
dup name>> text =
|
|
|
|
[ text>> [ blank? ] all? not ] [ drop t ] if
|
2008-04-26 00:17:08 -04:00
|
|
|
] filter ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2007-11-29 12:18:46 -05:00
|
|
|
: trim-text ( vector -- vector' )
|
|
|
|
[
|
2008-09-25 03:02:47 -04:00
|
|
|
dup name>> text =
|
|
|
|
[ [ [ blank? ] trim ] change-text ] when
|
2007-11-29 12:18:46 -05:00
|
|
|
] map ;
|
|
|
|
|
2008-11-16 17:59:25 -05:00
|
|
|
: find-by-id ( vector id -- vector' elt/f )
|
2010-08-29 14:22:11 -04:00
|
|
|
'[ "id" attribute _ = ] find ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
2008-11-16 17:59:25 -05:00
|
|
|
: find-by-class ( vector id -- vector' elt/f )
|
2010-08-29 14:22:11 -04:00
|
|
|
'[ "class" attribute _ = ] find ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-11-16 17:59:25 -05:00
|
|
|
: find-by-name ( vector string -- vector elt/f )
|
2008-10-19 16:52:59 -04:00
|
|
|
>lower '[ name>> _ = ] find ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-by-id-between ( vector string -- vector' )
|
|
|
|
dupd
|
2010-08-29 14:22:11 -04:00
|
|
|
'[ "id" attribute _ = ] find find-between* ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: find-by-class-between ( vector string -- vector' )
|
|
|
|
dupd
|
2010-08-29 14:22:11 -04:00
|
|
|
'[ "class" attribute _ = ] find find-between* ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: find-by-class-id-between ( vector class id -- vector' )
|
2010-08-29 14:22:11 -04:00
|
|
|
[
|
|
|
|
'[
|
|
|
|
[ "class" attribute _ = ]
|
|
|
|
[ "id" attribute _ = ] bi and
|
|
|
|
] find
|
|
|
|
] [
|
|
|
|
2drop find-between*
|
|
|
|
] 3bi ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
2008-11-16 17:59:25 -05:00
|
|
|
: find-by-attribute-key ( vector key -- vector' elt/? )
|
2008-10-19 16:52:59 -04:00
|
|
|
>lower
|
|
|
|
[ attributes>> at _ = ] filter sift ;
|
|
|
|
|
|
|
|
: find-by-attribute-key-value ( vector value key -- vector' )
|
|
|
|
>lower
|
2010-08-29 14:22:11 -04:00
|
|
|
[ attributes>> at over = ] with filter nip sift ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-first-attribute-key-value ( vector value key -- i/f tag/f )
|
|
|
|
>lower
|
2008-05-19 19:58:56 -04:00
|
|
|
[ attributes>> at over = ] with find rot drop ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2010-08-29 14:22:11 -04:00
|
|
|
: tag-link ( tag -- link/f ) "href" attribute ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-05-19 19:58:56 -04:00
|
|
|
: find-links ( vector -- vector' )
|
2010-08-29 14:22:11 -04:00
|
|
|
[ { [ name>> "a" = ] [ "href" attribute ] } 1&& ]
|
2008-05-19 19:58:56 -04:00
|
|
|
find-between-all ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2009-08-26 15:28:06 -04:00
|
|
|
: find-images ( vector -- vector' )
|
|
|
|
[
|
|
|
|
{
|
|
|
|
[ name>> "img" = ]
|
2010-08-29 14:22:11 -04:00
|
|
|
[ "src" attribute ]
|
2009-08-26 15:28:06 -04:00
|
|
|
} 1&&
|
|
|
|
] find-all
|
2010-08-29 14:22:11 -04:00
|
|
|
values [ "src" attribute ] map ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-03-06 18:09:48 -05:00
|
|
|
: find-by-text ( seq quot -- tag )
|
2009-03-22 20:25:55 -04:00
|
|
|
[ dup name>> text = ] prepose find drop ; inline
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-02-11 17:27:40 -05:00
|
|
|
: find-opening-tags-by-name ( name seq -- seq )
|
2010-08-29 14:22:11 -04:00
|
|
|
[ { [ name>> = ] [ closing?>> not ] } 1&& ] with find-all ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-02-11 17:27:40 -05:00
|
|
|
: href-contains? ( str tag -- ? )
|
2010-08-29 14:22:11 -04:00
|
|
|
"href" attribute* [ subseq? ] [ 2drop f ] if ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-08-14 00:09:43 -04:00
|
|
|
: find-hrefs ( vector -- vector' )
|
|
|
|
find-links
|
2010-08-29 14:22:11 -04:00
|
|
|
[ [ { [ name>> "a" = ] [ "href" attribute? ] } 1&& ] filter ] map sift
|
|
|
|
[ [ "href" attribute ] map ] map concat [ >url ] map ;
|
2009-04-01 03:33:38 -04:00
|
|
|
|
|
|
|
: find-frame-links ( vector -- vector' )
|
|
|
|
[ name>> "frame" = ] find-between-all
|
2010-08-29 14:22:11 -04:00
|
|
|
[ [ "src" attribute ] map sift ] map concat sift
|
2009-04-03 22:15:52 -04:00
|
|
|
[ >url ] map ;
|
2009-04-01 03:33:38 -04:00
|
|
|
|
|
|
|
: find-all-links ( vector -- vector' )
|
2010-02-27 14:52:24 -05:00
|
|
|
[ find-hrefs ] [ find-frame-links ] bi union ;
|
2008-02-11 18:19:27 -05:00
|
|
|
|
2008-05-19 19:58:56 -04:00
|
|
|
: find-forms ( vector -- vector' )
|
|
|
|
"form" over find-opening-tags-by-name
|
2008-12-14 01:38:37 -05:00
|
|
|
swap [ [ first2 ] dip find-between* ] curry map
|
2008-05-19 19:58:56 -04:00
|
|
|
[ [ name>> { "form" "input" } member? ] filter ] map ;
|
|
|
|
|
2008-10-19 16:52:59 -04:00
|
|
|
: find-html-objects ( vector string -- vector' )
|
2010-10-25 13:50:43 -04:00
|
|
|
over find-opening-tags-by-name
|
|
|
|
[ first2 find-between* ] with map ;
|
2008-05-19 19:58:56 -04:00
|
|
|
|
|
|
|
: form-action ( vector -- string )
|
2010-08-29 14:22:11 -04:00
|
|
|
[ name>> "form" = ] find nip "action" attribute ;
|
2007-09-20 18:09:08 -04:00
|
|
|
|
2008-05-19 19:58:56 -04:00
|
|
|
: hidden-form-values ( vector -- strings )
|
2010-08-29 14:22:11 -04:00
|
|
|
[ "type" attribute "hidden" = ] filter ;
|
2008-05-19 19:58:56 -04:00
|
|
|
|
|
|
|
: input. ( tag -- )
|
|
|
|
dup name>> print
|
|
|
|
attributes>>
|
|
|
|
[ bl bl bl bl [ write "=" write ] [ write bl ] bi* nl ] assoc-each ;
|
|
|
|
|
|
|
|
: form. ( vector -- )
|
|
|
|
[ closing?>> not ] filter
|
|
|
|
[
|
|
|
|
{
|
|
|
|
{ [ dup name>> "form" = ]
|
2010-08-29 14:22:11 -04:00
|
|
|
[ "form action: " write "action" attribute print ] }
|
2008-05-19 19:58:56 -04:00
|
|
|
{ [ dup name>> "input" = ] [ input. ] }
|
|
|
|
[ drop ]
|
|
|
|
} cond
|
|
|
|
] each ;
|
|
|
|
|
|
|
|
: query>assoc* ( str -- hash )
|
|
|
|
"?" split1 nip query>assoc ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: html-class? ( tag string -- ? )
|
2010-08-29 14:22:11 -04:00
|
|
|
swap "class" attribute = ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: html-id? ( tag string -- ? )
|
2010-08-29 14:22:11 -04:00
|
|
|
swap "id" attribute = ;
|
2008-10-19 16:52:59 -04:00
|
|
|
|
|
|
|
: opening-tag? ( tag -- ? )
|
|
|
|
closing?>> not ;
|
2010-08-29 14:22:11 -04:00
|
|
|
|
|
|
|
TUPLE: link attributes clickable ;
|
|
|
|
|
|
|
|
: <link> ( vector -- link )
|
|
|
|
[ first attributes>> ]
|
|
|
|
[ [ name>> { text "img" } member? ] filter ] bi
|
|
|
|
link boa ;
|
|
|
|
|
|
|
|
: link. ( vector -- )
|
|
|
|
[ "href" attribute write nl ]
|
|
|
|
[ clickable>> [ bl bl text>> print ] each nl ] bi ;
|