#!/usr/bin/tclsh
-# $Id: robot.tcl,v 1.11 2001/01/23 11:26:43 adam Exp $
+# $Id: robot.tcl,v 1.19 2001/06/29 21:47:31 adam Exp $
#
proc RobotFileNext1 {area lead} {
- puts "RobotFileNext1 area=$area lead=$lead"
+ # puts "RobotFileNext1 area=$area lead=$lead"
if {[catch {set ns [glob ${area}/*]}]} {
return {}
}
gets $inf
gets $inf
set distance [string trim [gets $inf]]
- puts "got distance = $distance"
+ # puts "got distance = $distance"
gets $inf
gets $inf
set fromurl [string trim [gets $inf]]
proc RobotFileNext {area} {
global robotSeq global idleTime ns
- puts "RobotFileNext robotSeq=$robotSeq"
+ # puts "RobotFileNext robotSeq=$robotSeq"
if {$robotSeq < 0} {
return {}
}
proc RobotFileExist {area host path} {
- puts "RobotFileExist begin area=$area host=$host path=$path"
+ # puts "RobotFileExist begin area=$area host=$host path=$path"
set lpath [split $path /]
set l [llength $lpath]
incr l -1
set t [lindex $lpath $l]
incr l -1
set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
- puts "RobotFileExist end npath=$npath"
+ # puts "RobotFileExist end npath=$npath"
return [file exists $npath]
}
proc RobotFileUnlink {area host path} {
- puts "RobotFileUnlink begin"
- puts "area=$area host=$host path=$path"
+ # puts "RobotFileUnlink begin"
+ # puts "area=$area host=$host path=$path"
set lpath [split $path /]
set l [llength $lpath]
incr l -1
set t [lindex $lpath $l]
incr l -1
set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
- puts "npath=$npath"
+ # puts "npath=$npath"
set comp [split $npath /]
set l [llength $comp]
incr l -1
if {![catch {glob $path/*}]} return
exec rmdir ./$path
}
- puts "RobotFileUnlink end"
+ # puts "RobotFileUnlink end"
}
proc RobotFileClose {out} {
if {![info exists workdir]} {
return stdout
}
- puts "RobotFileOpen orgPwd=$orgPwd area=$area host=$host path=$path mode=$mode"
+ #puts "RobotFileOpen orgPwd=$orgPwd area=$area host=$host path=$path mode=$mode"
if {[string compare $orgPwd $workdir]} {
puts "ooops. RobotFileOpen failed"
puts "workdir = $workdir"
global URL
global robotsRunning robotsMax idleTime
- puts "RobotStart"
+ # puts "RobotStart"
while {1} {
set url [RobotFileNext unvisited]
if {![string length $url]} {
if {[string first { } $href] >= 0} {
return 0
}
+ if {[string length $href] > 256} {
+ return 0
+ }
+ if {[string first {?} $href] >= 0} {
+ return 0
+ }
if {[string first {?} $url] >= 0 && [string first {?} $href] >= 0} {
return 0
}
if {![string length $surl]} {
set surl /
}
- set ok 0
- foreach domain $domains {
- if {[string match $domain $host]} {
- set ok 1
- break
+ if {[info exist domains]} {
+ set ok 0
+ foreach domain $domains {
+ if {[string match $domain $host]} {
+ set ok 1
+ break
+ }
}
- }
- if {!$ok} {
- return 0
- }
+ if {!$ok} {
+ return 0
+ }
+ }
} else {
regexp {^([^\#]*)} $hpath x surl
set host $URL($url,hostport)
set surl $dpart/$surl
}
}
- set c [split $surl /]
- set i [llength $c]
- incr i -1
- set path [lindex $c $i]
- incr i -1
- while {$i >= 0} {
- switch -- [lindex $c $i] {
+ set surllist [split $surl /]
+ catch {unset path}
+ set pathl 0
+ foreach c $surllist {
+ switch -- $c {
.. {
- incr i -2
- if {$i < 0} {
- set i 0
+ if {$pathl > 0} {
+ incr pathl -1
+ set path [lrange $path 0 $pathl]
}
}
- . {
- incr i -1
- }
- default {
- set path [lindex $c $i]/$path
- incr i -1
+ . {
+
+ }
+ default {
+ incr pathl
+ lappend path $c
}
}
}
+ if {$pathl} {
+ set path [join $path /]
+ } else {
+ set path ""
+ }
regsub -all {~} $path {%7E} path
set href "$method://$host$path"
puts "Ref href = $href"
}
puts $out {></meta>}
} body {
- regsub -all -nocase {<script.*</script>} $body {} abody
+ regsub -all -nocase {<script([^<]|(<!.*>))*</script>} $body {} abody
regsub -all {<[^\>]+>} $abody {} nbody
puts $out "<documentcontent>"
puts $out $nbody
puts $out "</documentcontent>"
- } a {
+ } -nonest a {
if {![info exists parm(href)]} {
puts "no href"
continue
}
}
}
+ } -nonest area {
+ if {![info exists parm(href)]} {
+ puts "no href"
+ continue
+ }
+ if {[expr $distance <= $maxDistance]} {
+ set href [string trim $parm(href)]
+ if {![RobotHref $url href host path]} continue
+
+ puts $out "<cr>"
+ puts $out "<identifier>$href</identifier>"
+ puts $out "<description></description>"
+ puts $out "</cr>"
+
+ if {![RobotFileExist visited $host $path]} {
+ set olddistance 1000
+ if {![RobotFileExist bad $host $path]} {
+ if {[RobotFileExist unvisited $host $path]} {
+ set inf [RobotFileOpen unvisited $host $path r]
+ RobotReadRecord $inf oldurl olddistance
+ RobotFileClose $inf
+ }
+ } else {
+ set olddistance 0
+ }
+ if {[string length $olddistance] == 0} {
+ set olddistance 1000
+ }
+ if {[expr $distance < $olddistance]} {
+ set outf [RobotFileOpen unvisited $host $path]
+ RobotWriteRecord $outf $url $distance
+ RobotFileClose $outf
+ }
+ } elseif {[string compare $href $url]} {
+ set inf [RobotFileOpen visited $host $path r]
+ RobotReadRecord $inf xurl olddistance
+ close $inf
+ if {[string length $olddistance] == 0} {
+ set olddistance 1000
+ }
+ if {[expr $distance < $olddistance]} {
+ puts "OK remarking url=$url href=$href"
+ puts "olddistance = $olddistance"
+ puts "newdistance = $distance"
+ set outf [RobotFileOpen unvisited $host $path]
+ RobotWriteRecord $outf $url $distance
+ RobotFileClose $outf
+ }
+ }
+ }
}
}
global URL agent
set section 0
foreach l [split $buf \n] {
- if {[regexp {([-A-Za-z]+):[ \t]*([^\#]+)} $l match cmd arg]} {
+ if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
puts "cmd=$cmd arg=$arg"
- switch $cmd {
- User-Agent {
+ switch -- [string tolower $cmd] {
+ user-agent {
if {$section} break
set pat [string tolower $arg]*
set section [string match $pat $agent]
}
- Disallow {
+ disallow {
if {$section} {
puts "rule [list 0 $arg]"
lappend $v [list 0 $arg]
}
}
- Allow {
+ allow {
if {$section} {
puts "rule [list 1 $arg]"
lappend $v [list 1 $arg]
global URL
puts $out "<documentcontent>"
- puts $out $URL($url,buf)
+ regsub -all {<} $URL($url,buf) {\<} content
+ puts $out $content
puts $out "</documentcontent>"
if {![string compare $URL($url,path) /robots.txt]} {
proc Robot200 {url} {
global URL domains
+ set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
+ puts -nonewline $out $URL($url,buf)
+ RobotFileClose $out
+
set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
puts $out "<zmbot>"
regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
set lines [split $headbuf \n]
foreach line $lines {
- if {[regexp {^([^:]+):[ ]+(.*)} $line x name value]} {
+ if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
set URL($url,head,[string tolower $name]) [string trim $value]
}
}
}
proc RobotConnect {url sock} {
- global URL agent
+ global URL agent acceptLanguage
fconfigure $sock -translation {lf crlf} -blocking 0
fileevent $sock readable [list RobotReadHeader $url $sock]
puts $sock "GET $URL($url,path) HTTP/1.0"
puts $sock "Host: $URL($url,host)"
puts $sock "User-Agent: $agent"
+ if {[string length $acceptLanguage]} {
+ puts $sock "Accept-Language: $acceptLanguage"
+ }
puts $sock ""
flush $sock
set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
set buf [read $inf 32768]
close $inf
} else {
- set buf "User-Agent: *\nAllow: /\n"
+ set buf "User-agent: *\nAllow: /\n"
}
RobotsTxt0 URL($hostport,robots) $buf
}
}
}
if {!$ok} {
+ puts "skipped due to robots.txt"
return -1
}
}
set robotSeq 0
set workdir [pwd]
set idleTime 60000
+set acceptLanguage {}
set i 0
set l [llength $argv]
+# For testing only
+if {0} {
+ set url "http://www.sportsfiskeren.dk/sportsfiskeren/corner/index.htm"
+ set href "../../data/../../data2/newsovs.asp?Mode=5"
+
+ set URL($url,path) /sportsfiskeren/corner/index.htm
+ set URL($url,hostport) www.sportsfiskeren.dk
+ RobotHref $url href host path
+ exit 0
+}
+
if {$l < 2} {
- puts {tclrobot: usage [-j jobs] [-c count] [-d domain] [url ..]}
+ puts {tclrobot: usage [-j jobs] [-i idle] [-c count] [-d domain] [url ..]}
puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
exit 1
}
}
lappend domains $dom
}
+ -i* {
+ set idleTime [string range $arg 2 end]
+ if {![string length $idleTime]} {
+ set idleTime [lindex $argv [incr i]]
+ }
+ }
+ -l* {
+ set acceptLanguage [string range $arg 2 end]
+ if {![string length $acceptLanguage]} {
+ set acceptLanguage [lindex $argv [incr i]]
+ }
+ }
default {
set href $arg
if {[RobotHref http://www.indexdata.dk/ href host path]} {
set domains {*}
}
if {![info exist maxDistance]} {
- set maxDistance 3
+ set maxDistance 50
}
if {![info exist robotsMax]} {
set robotsMax 5
projects / tclrobot.git / blobdiff