#!/usr/bin/tclsh
-# $Id: robot.tcl,v 1.20 2001/06/29 22:25:55 adam Exp $
+# $Id: robot.tcl,v 1.28 2001/11/13 11:17:26 adam Exp $
#
proc RobotFileNext1 {area lead} {
# puts "RobotFileNext1 area=$area lead=$lead"
}
proc RobotFileNext {area} {
- global robotSeq global idleTime ns
+ global robotSeq
+ global idletime ns
+ global status
# puts "RobotFileNext robotSeq=$robotSeq"
if {$robotSeq < 0} {
if {![string length $n]} {
set robotSeq -1
flush stdout
- puts "------------ N E X T R O U N D --------"
+ puts "Round robin un,ba,vi=$status(unvisited),$status(bad),$status(visited)"
return wait
}
incr robotSeq
}
proc RobotFileUnlink {area host path} {
+ global status
# puts "RobotFileUnlink begin"
# puts "area=$area host=$host path=$path"
set lpath [split $path /]
set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
# puts "npath=$npath"
set comp [split $npath /]
+ if {[catch {exec rm [join $comp /]}]} return
+
set l [llength $comp]
incr l -1
- if {[catch {exec rm [join $comp /]}]} return
incr l -1
+ incr status($area) -1
for {set i $l} {$i > 0} {incr i -1} {
set path [join [lrange $comp 0 $i] /]
if {![catch {glob $path/*}]} return
proc RobotFileOpen {area host path {mode w}} {
set orgPwd [pwd]
global workdir
+ global status
if {![info exists workdir]} {
return stdout
set out [open frobots.txt w]
puts "creating robots.txt in $d"
close $out
+ incr status(unvisited)
}
}
}
} else {
set out [open f $mode]
}
+ if {$mode == "w"} {
+ incr status($area)
+ }
cd $orgPwd
return $out
}
proc RobotStart {} {
global URL
- global robotsRunning robotsMax idleTime
+ global robotsRunning robotsMax idletime
# puts "RobotStart"
while {1} {
}
incr robotsRunning
if {[string compare $url wait] == 0} {
- after $idleTime RobotRR
+ after $idletime RobotRR
return
}
set r [RobotGetUrl $url {}]
}
proc RobotHref {url hrefx hostx pathx} {
- global URL domains
+ global URL domains debuglevel
upvar $hrefx href
upvar $hostx host
upvar $pathx path
- puts "Ref url = $url href=$href"
+ if {$debuglevel > 1} {
+ puts "Ref input url = $url href=$href"
+ }
if {[string first { } $href] >= 0} {
return 0
foreach c $surllist {
switch -- $c {
.. {
- if {$pathl > 0} {
+ if {$pathl > 1} {
incr pathl -2
set path [lrange $path 0 $pathl]
incr pathl
}
}
}
- if {$pathl} {
- set path [join $path /]
- } else {
- set path ""
+ if {$debuglevel > 4} {
+ puts "pathl=$pathl output path=$path"
+ }
+ set path [join $path /]
+ if {![string length $path]} {
+ set path /
}
regsub -all {~} $path {%7E} path
set href "$method://$host$path"
- puts "Ref href = $href"
- return 1
+
+ if {$debuglevel > 1} {
+ puts "Ref result = $href"
+ }
+ return [checkrule url $href]
}
proc RobotError {url code} {
global URL
- puts "Bad URL $url, $code"
+ puts "Bad URL $url (code $code)"
set fromurl {}
set distance -1
if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
}
proc RobotTextHtml {url out} {
- global URL maxDistance
+ global URL maxdistance
set distance 0
- if {$maxDistance < 1000 && [info exists URL($url,dist)]} {
+ if {$maxdistance < 1000 && [info exists URL($url,dist)]} {
set distance [expr $URL($url,dist) + 1]
}
htmlSwitch $URL($url,buf) \
}
puts $out {></meta>}
} body {
- regsub -all -nocase {<script([^<]|(<!.*>))*</script>} $body {} abody
- regsub -all {<[^\>]+>} $abody {} nbody
+ regsub -all {<!--[^-]*->} $body { } abody
+ regsub -all -nocase {<script[^<]*</script>} $abody {} bbody
+ regsub -all {<[^\>]+>} $bbody {} nbody
puts $out "<documentcontent>"
puts $out $nbody
puts $out "</documentcontent>"
puts "no href"
continue
}
- if {[expr $distance <= $maxDistance]} {
+ if {[expr $distance <= $maxdistance]} {
set href [string trim $parm(href)]
if {![RobotHref $url href host path]} continue
puts "no href"
continue
}
- if {[expr $distance <= $maxDistance]} {
+ if {[expr $distance <= $maxdistance]} {
set href [string trim $parm(href)]
if {![RobotHref $url href host path]} continue
}
}
}
+ } -nonest frame {
+ if {![info exists parm(src)]} {
+ puts "no src"
+ continue
+ }
+ if {[expr $distance <= $maxdistance]} {
+ set href [string trim $parm(src)]
+ 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
+ }
+ }
+ }
}
}
}
}
-proc Robot200 {url} {
+proc RobotWriteMetadata {url out} {
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>"
set distance 1000
headSave $url $out
puts "Parsing $url distance=$distance"
switch $URL($url,head,content-type) {
- text/html {
- if {[string length $distance]} {
- RobotTextHtml $url $out
- }
- }
- text/plain {
- RobotTextPlain $url $out
- }
- application/pdf {
- set pdff [open test.pdf w]
- puts -nonewline $pdff $URL($url,buf)
- close $pdff
- }
+ text/html {
+ if {[string length $distance]} {
+ RobotTextHtml $url $out
+ }
+ }
+ text/plain {
+ RobotTextPlain $url $out
+ }
}
puts $out "</zmbot>"
+}
+
+proc Robot200 {url} {
+ global URL domains
+
+ set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
+ puts -nonewline $out $URL($url,buf)
RobotFileClose $out
- # puts "Parsing done"
+
+ set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
+ RobotWriteMetadata $url $out
+ RobotFileClose $out
+
RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
}
}
proc RobotReadHeader {url sock} {
- global URL
+ global URL debuglevel
- puts "RobotReadHeader $url"
+ if {$debuglevel > 1} {
+ puts "HTTP head $url"
+ }
if {[catch {set buffer [read $sock 2148]}]} {
RobotError $url 404
RobotRestart $url $sock
+ return
}
set readCount [string length $buffer]
set URL($url,head,[string tolower $name]) [string trim $value]
}
}
- puts "code = $code"
+ puts "HTTP CODE $code"
set URL($url,state) skip
switch $code {
301 {
if {![info exists URL($url,head,content-type)]} {
set URL($url,head,content-type) {}
}
- set binary 0
- switch $URL($url,head,content-type) {
- application/pdf {
- set binary 1
+ set binary 1
+ switch -glob -- $URL($url,head,content-type) {
+ text/* {
+ set binary 0
}
}
+ if {![regexp {/robots.txt$} $url]} {
+ if {![checkrule mime $URL($url,head,content-type)]} {
+ RobotError $url mimedeny
+ RobotRestart $url $sock
+ return
+ }
+ }
fileevent $sock readable [list RobotReadContent $url $sock $binary]
}
default {
proc RobotGetUrl {url phost} {
global URL robotsRunning
flush stdout
- puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
+ puts "Retrieve $robotsRunning url=$url"
if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
return -1
}
}
}
-set agent "zmbot/0.0"
+set agent "zmbot/0.1"
if {![catch {set os [exec uname -s -r]}]} {
set agent "$agent ($os)"
}
set robotsRunning 0
set robotSeq 0
set workdir [pwd]
-set idleTime 60000
+set idletime 60000
set acceptLanguage {}
+set debuglevel 0
+set status(unvisited) 0
+set status(visited) 0
+set status(bad) 0
+set status(raw) 0
-set i 0
-set l [llength $argv]
-# For testing only
-if {0} {
- set url "http://www.sportsfiskeren.dk/sportsfiskeren/corner/index.htm"
- set href "../../data/../../data2/newsovs.asp?Mode=5"
+# Rules: allow, deny, url
+
+proc checkrule {type this} {
+ global alrules
+ global debuglevel
+
+ if {$debuglevel > 3} {
+ puts "CHECKRULE $type $this"
+ }
+ if {[info exist alrules]} {
+ foreach l $alrules {
+ if {$debuglevel > 3} {
+ puts "consider $l"
+ }
+ # consider type
+ if {[lindex $l 1] != $type} continue
+ # consider mask (! negates)
+ set masks [lindex $l 2]
+ set ok 0
+ foreach mask $masks {
+ if {$debuglevel > 4} {
+ puts "consider single mask $mask"
+ }
+ if {[string index $mask 0] == "!"} {
+ set mask [string range $mask 1 end]
+ if {[string match $mask $this]} continue
+ } else {
+ if {![string match $mask $this]} continue
+ }
+ set ok 1
+ }
+ if {$debuglevel > 4} {
+ puts "ok = $ok"
+ }
+ if {!$ok} continue
+ # OK, we have a match
+ if {[lindex $l 0] == "allow"} {
+ if {$debuglevel > 3} {
+ puts "CHECKRULE MATCH OK"
+ }
+ return 1
+ } else {
+ if {$debuglevel > 3} {
+ puts "CHECKFULE MATCH FAIL"
+ }
+ return 0
+ }
+ }
+ }
+ if {$debuglevel > 3} {
+ puts "CHECKRULE MATCH OK"
+ }
+ return 1
+}
+
+
+proc url {href} {
+ global debuglevel
+
+ if {[RobotHref http://www.indexdata.dk/ href host path]} {
+ if {![RobotFileExist visited $host $path]} {
+ set outf [RobotFileOpen unvisited $host $path]
+ RobotWriteRecord $outf href 0
+ RobotFileClose $outf
+ }
+ }
+}
+
+proc deny {type stuff} {
+ global alrules
+
+ lappend alrules [list deny $type $stuff]
+}
+
+proc allow {type stuff} {
+ global alrules
+
+ lappend alrules [list allow $type $stuff]
+}
+
+proc debug {level} {
+ global debuglevel
- set URL($url,path) /sportsfiskeren/corner/index.htm
- set URL($url,hostport) www.sportsfiskeren.dk
- RobotHref $url href host path
- exit 0
+ set debuglevel $level
}
+# Parse options
+
+set i 0
+set l [llength $argv]
+
if {$l < 2} {
- puts {tclrobot: usage [-j jobs] [-i idle] [-c count] [-d domain] [url ..]}
+ puts {tclrobot: usage:}
+ puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-r rules] [url ..]}
puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
+
exit 1
}
-
while {$i < $l} {
set arg [lindex $argv $i]
switch -glob -- $arg {
}
}
-c* {
- set maxDistance [string range $arg 2 end]
- if {![string length $maxDistance]} {
- set maxDistance [lindex $argv [incr i]]
+ set maxdistance [string range $arg 2 end]
+ if {![string length $maxdistance]} {
+ set maxdistance [lindex $argv [incr i]]
}
}
-d* {
lappend domains $dom
}
-i* {
- set idleTime [string range $arg 2 end]
- if {![string length $idleTime]} {
- set idleTime [lindex $argv [incr i]]
+ set idletime [string range $arg 2 end]
+ if {![string length $idletime]} {
+ set idletime [lindex $argv [incr i]]
}
}
-l* {
set acceptLanguage [lindex $argv [incr i]]
}
}
+ -r* {
+ set rfile [string range $arg 2 end]
+ if {![string length $rfile]} {
+ set rfile [lindex $argv [incr i]]
+ }
+ source $rfile
+ }
default {
set href $arg
if {[RobotHref http://www.indexdata.dk/ href host path]} {
if {![info exist domains]} {
set domains {*}
}
-if {![info exist maxDistance]} {
- set maxDistance 50
+if {![info exist maxdistance]} {
+ set maxdistance 50
}
if {![info exist robotsMax]} {
set robotsMax 5
}
puts "domains=$domains"
-puts "max distance=$maxDistance"
+puts "max distance=$maxdistance"
puts "max jobs=$robotsMax"
+
RobotStart
+
while {$robotsRunning} {
vwait robotsRunning
}
+
+puts "End un,ba,vi=$status(unvisited),$status(bad),$status(visited)"
projects / tclrobot.git / blobdiff