2 # $Id: robot.tcl,v 1.22 2001/10/30 08:29:54 adam Exp $
4 proc RobotFileNext1 {area lead} {
5 # puts "RobotFileNext1 area=$area lead=$lead"
6 if {[catch {set ns [glob ${area}/*]}]} {
10 if {[file isfile $n]} {
11 set off [string last / $n]
13 return $lead/[string range $n $off end]
17 if {[file isdirectory $n]} {
18 set off [string last / $n]
20 set sb [RobotFileNext1 $n $lead/[string range $n $off end]]
21 if {[string length $sb]} {
29 proc RobotWriteRecord {outf fromurl distance} {
31 puts $outf "<distance>"
33 puts $outf "</distance>"
34 puts $outf "<fromurl>"
36 puts $outf "</fromurl>"
40 proc RobotReadRecord {inf fromurlx distancex} {
41 upvar $fromurlx fromurl
42 upvar $distancex distance
45 set distance [string trim [gets $inf]]
46 # puts "got distance = $distance"
49 set fromurl [string trim [gets $inf]]
52 proc RobotFileNext {area} {
53 global robotSeq global idletime ns
55 # puts "RobotFileNext robotSeq=$robotSeq"
60 if {[catch {set ns [glob ${area}/*]}]} {
64 set off [string length $area]
66 set n [lindex $ns $robotSeq]
67 if {![string length $n]} {
74 if {[file isfile $n/frobots.txt]} {
75 puts "ok returning http://[string range $n $off end]/robots.txt"
76 return http://[string range $n $off end]/robots.txt
77 } elseif {[file isdirectory $n]} {
78 set sb [RobotFileNext1 $n http://[string range $n $off end]]
79 if {[string length $sb]} {
83 puts "no more work at end of RobotFileNext n=$n"
89 proc RobotFileExist {area host path} {
90 # puts "RobotFileExist begin area=$area host=$host path=$path"
91 set lpath [split $path /]
92 set l [llength $lpath]
94 set t [lindex $lpath $l]
96 set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
97 # puts "RobotFileExist end npath=$npath"
98 return [file exists $npath]
101 proc RobotFileUnlink {area host path} {
102 # puts "RobotFileUnlink begin"
103 # puts "area=$area host=$host path=$path"
104 set lpath [split $path /]
105 set l [llength $lpath]
107 set t [lindex $lpath $l]
109 set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
110 # puts "npath=$npath"
111 set comp [split $npath /]
112 set l [llength $comp]
114 if {[catch {exec rm [join $comp /]}]} return
116 for {set i $l} {$i > 0} {incr i -1} {
117 set path [join [lrange $comp 0 $i] /]
118 if {![catch {glob $path/*}]} return
121 # puts "RobotFileUnlink end"
124 proc RobotFileClose {out} {
125 if [string compare $out stdout] {
130 proc RobotFileOpen {area host path {mode w}} {
134 if {![info exists workdir]} {
137 #puts "RobotFileOpen orgPwd=$orgPwd area=$area host=$host path=$path mode=$mode"
138 if {[string compare $orgPwd $workdir]} {
139 puts "ooops. RobotFileOpen failed"
140 puts "workdir = $workdir"
144 set comp [split $area/$host$path /]
145 set len [llength $comp]
147 for {set i 0} {$i < $len} {incr i} {
149 set d "d[lindex $comp $i]"
151 set d [lindex $comp $i]
153 if {[catch {cd ./$d}]} {
156 if {![string compare $area unvisited] && $i == 1 && $mode == "w"} {
157 set out [open frobots.txt w]
158 puts "creating robots.txt in $d"
163 set d [lindex $comp $len]
164 if {[string length $d]} {
165 if {[file isdirectory $d]} {
166 set out [open $d/f $mode]
168 set out [open f$d $mode]
171 set out [open f $mode]
178 global robotSeq robotsRunning
180 incr robotsRunning -1
181 while {$robotsRunning} {
188 proc RobotRestart {url sock} {
189 global URL robotsRunning
192 after cancel $URL($sock,cancel)
194 foreach v [array names URL $url,*] {
198 incr robotsRunning -1
204 global robotsRunning robotsMax idletime
208 set url [RobotFileNext unvisited]
209 if {![string length $url]} {
213 if {[string compare $url wait] == 0} {
214 after $idletime RobotRR
217 set r [RobotGetUrl $url {}]
219 if {$robotsRunning >= $robotsMax} return
221 incr robotsRunning -1
222 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
223 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
226 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
231 proc headSave {url out} {
234 if {[info exists URL($url,head,last-modified)]} {
235 puts $out "<lastmodified>$URL($url,head,last-modified)</lastmodified>"
238 if {[info exists URL($url,head,date)]} {
239 puts $out " <date>$URL($url,head,date)</date>"
241 if {[info exists URL($url,head,content-length)]} {
242 puts $out " <by>$URL($url,head,content-length)</by>"
244 if {[info exists URL($url,head,server)]} {
245 puts $out " <format>$URL($url,head,server)</format>"
248 puts $out {<publisher>}
249 puts $out " <identifier>$url</identifier>"
250 if {[info exists URL($url,head,content-type)]} {
251 puts $out " <type>$URL($url,head,content-type)</type>"
253 puts $out {</publisher>}
256 proc RobotHref {url hrefx hostx pathx} {
257 global URL domains debuglevel
262 if {$debuglevel > 1} {
263 puts "Ref input url = $url href=$href"
266 if {[string first { } $href] >= 0} {
269 if {[string length $href] > 256} {
272 if {[string first {?} $href] >= 0} {
275 if {[string first {?} $url] >= 0 && [string first {?} $href] >= 0} {
278 # get method (if any)
279 if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
283 if {[string compare $method http]} {
288 if {[regexp {^//([^/]+)([^\#]*)} $hpath x host surl]} {
289 if {![string length $surl]} {
292 if {[info exist domains]} {
294 foreach domain $domains {
295 if {[string match $domain $host]} {
305 regexp {^([^\#]*)} $hpath x surl
306 set host $URL($url,hostport)
308 if {![string length $surl]} {
311 if {[string first / $surl]} {
313 regexp {^([^\#?]*)} $URL($url,path) x dpart
314 set l [string last / $dpart]
315 if {[expr $l >= 0]} {
316 set surl [string range $dpart 0 $l]$surl
318 set surl $dpart/$surl
321 set surllist [split $surl /]
324 foreach c $surllist {
329 set path [lrange $path 0 $pathl]
343 set path [join $path /]
347 regsub -all {~} $path {%7E} path
348 set href "$method://$host$path"
350 if {$debuglevel > 1} {
351 puts "Ref result = $href"
353 return [checkrule url $href]
356 proc RobotError {url code} {
359 puts "Bad URL $url (code $code)"
362 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
363 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
364 RobotReadRecord $inf fromurl distance
367 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
368 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
369 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
370 RobotWriteRecord $outf $fromurl $distance
375 proc RobotRedirect {url tourl code} {
378 puts "Redirecting from $url to $tourl"
382 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
383 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
384 RobotReadRecord $inf fromurl distance
387 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
388 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
389 RobotWriteRecord $outf $fromurl $distance
392 if {[RobotHref $url tourl host path]} {
393 if {![RobotFileExist visited $host $path]} {
394 if {![RobotFileExist unvisited $host $path]} {
395 set outf [RobotFileOpen unvisited $host $path]
396 RobotWriteRecord $outf $fromurl $distance
401 set inf [RobotFileOpen visited $host $path r]
402 RobotReadRecord $inf oldurl olddistance
404 if {[string length $olddistance] == 0} {
407 if {[string length $distance] == 0} {
410 puts "distance=$distance olddistance=$olddistance"
411 if {[expr $distance < $olddistance]} {
412 set outf [RobotFileOpen unvisited $host $path]
413 RobotWriteRecord $outf $tourl $distance
418 if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
424 proc RobotTextHtml {url out} {
425 global URL maxdistance
428 if {$maxdistance < 1000 && [info exists URL($url,dist)]} {
429 set distance [expr $URL($url,dist) + 1]
431 htmlSwitch $URL($url,buf) \
433 puts $out "<title>$body</title>"
435 puts -nonewline $out "<meta"
436 foreach a [array names parm] {
437 puts -nonewline $out " $a"
438 puts -nonewline $out {="}
439 puts -nonewline $out $parm($a)
440 puts -nonewline $out {"}
444 regsub -all -nocase {<script([^<]|(<!.*>))*</script>} $body {} abody
445 regsub -all {<[^\>]+>} $abody {} nbody
446 puts $out "<documentcontent>"
448 puts $out "</documentcontent>"
450 if {![info exists parm(href)]} {
454 if {[expr $distance <= $maxdistance]} {
455 set href [string trim $parm(href)]
456 if {![RobotHref $url href host path]} continue
459 puts $out "<identifier>$href</identifier>"
460 puts $out "<description>$body</description>"
463 if {![RobotFileExist visited $host $path]} {
465 if {![RobotFileExist bad $host $path]} {
466 if {[RobotFileExist unvisited $host $path]} {
467 set inf [RobotFileOpen unvisited $host $path r]
468 RobotReadRecord $inf oldurl olddistance
474 if {[string length $olddistance] == 0} {
477 if {[expr $distance < $olddistance]} {
478 set outf [RobotFileOpen unvisited $host $path]
479 RobotWriteRecord $outf $url $distance
482 } elseif {[string compare $href $url]} {
483 set inf [RobotFileOpen visited $host $path r]
484 RobotReadRecord $inf xurl olddistance
486 if {[string length $olddistance] == 0} {
489 if {[expr $distance < $olddistance]} {
490 puts "OK remarking url=$url href=$href"
491 puts "olddistance = $olddistance"
492 puts "newdistance = $distance"
493 set outf [RobotFileOpen unvisited $host $path]
494 RobotWriteRecord $outf $url $distance
500 if {![info exists parm(href)]} {
504 if {[expr $distance <= $maxdistance]} {
505 set href [string trim $parm(href)]
506 if {![RobotHref $url href host path]} continue
509 puts $out "<identifier>$href</identifier>"
510 puts $out "<description></description>"
513 if {![RobotFileExist visited $host $path]} {
515 if {![RobotFileExist bad $host $path]} {
516 if {[RobotFileExist unvisited $host $path]} {
517 set inf [RobotFileOpen unvisited $host $path r]
518 RobotReadRecord $inf oldurl olddistance
524 if {[string length $olddistance] == 0} {
527 if {[expr $distance < $olddistance]} {
528 set outf [RobotFileOpen unvisited $host $path]
529 RobotWriteRecord $outf $url $distance
532 } elseif {[string compare $href $url]} {
533 set inf [RobotFileOpen visited $host $path r]
534 RobotReadRecord $inf xurl olddistance
536 if {[string length $olddistance] == 0} {
539 if {[expr $distance < $olddistance]} {
540 puts "OK remarking url=$url href=$href"
541 puts "olddistance = $olddistance"
542 puts "newdistance = $distance"
543 set outf [RobotFileOpen unvisited $host $path]
544 RobotWriteRecord $outf $url $distance
552 proc RobotsTxt {url} {
555 RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
558 proc RobotsTxt0 {v buf} {
561 foreach l [split $buf \n] {
562 if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
563 puts "cmd=$cmd arg=$arg"
564 switch -- [string tolower $cmd] {
567 set pat [string tolower $arg]*
568 set section [string match $pat $agent]
572 puts "rule [list 0 $arg]"
573 lappend $v [list 0 $arg]
578 puts "rule [list 1 $arg]"
579 lappend $v [list 1 $arg]
587 proc RobotTextPlain {url out} {
590 puts $out "<documentcontent>"
591 regsub -all {<} $URL($url,buf) {\<} content
593 puts $out "</documentcontent>"
595 if {![string compare $URL($url,path) /robots.txt]} {
600 proc Robot200 {url} {
603 set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
604 puts -nonewline $out $URL($url,buf)
607 if {![checkrule mime $URL($url,head,content-type)]} {
608 RobotError $url mimedeny
612 set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
616 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
617 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
618 RobotReadRecord $inf fromurl distance
621 set URL($url,dist) $distance
622 puts $out "<distance>"
623 puts $out " $distance"
624 puts $out "</distance>"
626 puts "Parsing $url distance=$distance"
627 switch $URL($url,head,content-type) {
629 if {[string length $distance]} {
630 RobotTextHtml $url $out
634 RobotTextPlain $url $out
637 set pdff [open test.pdf w]
638 puts -nonewline $pdff $URL($url,buf)
644 # puts "Parsing done"
645 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
648 proc RobotReadContent {url sock binary} {
651 set buffer [read $sock 16384]
652 set readCount [string length $buffer]
654 if {$readCount <= 0} {
656 RobotRestart $url $sock
657 } elseif {!$binary && [string first \0 $buffer] >= 0} {
659 RobotRestart $url $sock
661 # puts "Got $readCount bytes"
662 set URL($url,buf) $URL($url,buf)$buffer
666 proc RobotReadHeader {url sock} {
667 global URL debuglevel
669 if {$debuglevel > 1} {
670 puts "HTTP head $url"
672 if {[catch {set buffer [read $sock 2148]}]} {
674 RobotRestart $url $sock
676 set readCount [string length $buffer]
678 if {$readCount <= 0} {
680 RobotRestart $url $sock
682 # puts "Got $readCount bytes"
683 set URL($url,buf) $URL($url,buf)$buffer
685 set n [string first \r\n\r\n $URL($url,buf)]
689 set headbuf [string range $URL($url,buf) 0 $n]
691 set URL($url,buf) [string range $URL($url,buf) $n end]
693 regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
694 set lines [split $headbuf \n]
695 foreach line $lines {
696 if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
697 set URL($url,head,[string tolower $name]) [string trim $value]
700 puts "HTTP CODE $code"
701 set URL($url,state) skip
704 RobotRedirect $url $URL($url,head,location) 301
705 RobotRestart $url $sock
708 RobotRedirect $url $URL($url,head,location) 302
709 RobotRestart $url $sock
712 if {![info exists URL($url,head,content-type)]} {
713 set URL($url,head,content-type) {}
716 switch $URL($url,head,content-type) {
721 fileevent $sock readable [list RobotReadContent $url $sock $binary]
724 RobotError $url $code
725 RobotRestart $url $sock
732 proc RobotSockCancel {url sock} {
734 puts "RobotSockCancel sock=$sock url=$url"
736 RobotRestart $url $sock
739 proc RobotConnect {url sock} {
740 global URL agent acceptLanguage
742 fconfigure $sock -translation {lf crlf} -blocking 0
743 fileevent $sock readable [list RobotReadHeader $url $sock]
744 puts $sock "GET $URL($url,path) HTTP/1.0"
745 puts $sock "Host: $URL($url,host)"
746 puts $sock "User-Agent: $agent"
747 if {[string length $acceptLanguage]} {
748 puts $sock "Accept-Language: $acceptLanguage"
752 set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
759 proc RobotGetUrl {url phost} {
760 global URL robotsRunning
762 puts "Retrieve $robotsRunning url=$url"
763 if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
766 if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
770 set URL($url,method) $method
771 set URL($url,host) $host
772 set URL($url,hostport) $hostport
773 set URL($url,path) $path
774 set URL($url,state) head
777 if {[string compare $path /robots.txt]} {
779 if {![info exists URL($hostport,robots)]} {
780 puts "READING robots.txt for host $hostport"
781 if {[RobotFileExist visited $hostport /robots.txt]} {
782 set inf [RobotFileOpen visited $hostport /robots.txt r]
783 set buf [read $inf 32768]
786 set buf "User-agent: *\nAllow: /\n"
788 RobotsTxt0 URL($hostport,robots) $buf
790 if {[info exists URL($hostport,robots)]} {
791 foreach l $URL($hostport,robots) {
792 if {[string first [lindex $l 1] $path] == 0} {
799 puts "skipped due to robots.txt"
803 if [catch {set sock [socket -async $host $port]}] {
806 RobotConnect $url $sock
811 if {![llength [info commands htmlSwitch]]} {
812 set e [info sharedlibextension]
813 if {[catch {load ./tclrobot$e}]} {
818 set agent "zmbot/0.1"
819 if {![catch {set os [exec uname -s -r]}]} {
820 set agent "$agent ($os)"
835 set acceptLanguage {}
838 set l [llength $argv]
841 puts {tclrobot: usage:}
842 puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-r rules] [url ..]}
843 puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
847 # Rules: allow, deny, url
850 proc checkrule {type this} {
854 if {$debuglevel > 3} {
855 puts "CHECKRULE $type $this"
857 if {[info exist alrules]} {
859 if {$debuglevel > 3} {
863 if {[lindex $l 1] != $type} continue
864 # consider mask (! negates)
865 set mask [lindex $l 2]
866 if {[string index $mask 0] == "!"} {
867 set mask [string range $mask 1 end]
868 if {[string match $mask $this]} continue
870 if {![string match $mask $this]} continue
872 # OK, we have a match
873 if {[lindex $l 0] == "allow"} {
874 if {$debuglevel > 3} {
875 puts "CHECKRULE MATH OK"
879 if {$debuglevel > 3} {
880 puts "CHECKFULE MATCH FAIL"
886 if {$debuglevel > 3} {
887 puts "CHECKRULE MATH OK"
896 if {[RobotHref http://www.indexdata.dk/ href host path]} {
897 if {![RobotFileExist visited $host $path]} {
898 set outf [RobotFileOpen unvisited $host $path]
899 RobotWriteRecord $outf href 0
905 proc deny {type stuff} {
908 lappend alrules [list deny $type $stuff]
911 proc allow {type stuff} {
914 lappend alrules [list allow $type $stuff]
920 set debuglevel $level
926 set arg [lindex $argv $i]
927 switch -glob -- $arg {
929 set robotsMax [string range $arg 2 end]
930 if {![string length $robotsMax]} {
931 set robotsMax [lindex $argv [incr i]]
935 set maxdistance [string range $arg 2 end]
936 if {![string length $maxdistance]} {
937 set maxdistance [lindex $argv [incr i]]
941 set dom [string range $arg 2 end]
942 if {![string length $dom]} {
943 set dom [lindex $argv [incr i]]
948 set idletime [string range $arg 2 end]
949 if {![string length $idletime]} {
950 set idletime [lindex $argv [incr i]]
954 set acceptLanguage [string range $arg 2 end]
955 if {![string length $acceptLanguage]} {
956 set acceptLanguage [lindex $argv [incr i]]
960 set rfile [string range $arg 2 end]
961 if {![string length $rfile]} {
962 set rfile [lindex $argv [incr i]]
968 if {[RobotHref http://www.indexdata.dk/ href host path]} {
969 if {![RobotFileExist visited $host $path]} {
970 set outf [RobotFileOpen unvisited $host $path]
971 RobotWriteRecord $outf href 0
980 if {![info exist domains]} {
983 if {![info exist maxdistance]} {
986 if {![info exist robotsMax]} {
990 puts "domains=$domains"
991 puts "max distance=$maxdistance"
992 puts "max jobs=$robotsMax"
996 while {$robotsRunning} {