2 # $Id: robot.tcl,v 1.25 2001/11/07 11:50:07 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 {<!--[^-]*->} $body { } abody
445 regsub -all -nocase {<script[^<]*</script>} $abody {} bbody
446 regsub -all {<[^\>]+>} $bbody {} nbody
447 puts $out "<documentcontent>"
449 puts $out "</documentcontent>"
451 if {![info exists parm(href)]} {
455 if {[expr $distance <= $maxdistance]} {
456 set href [string trim $parm(href)]
457 if {![RobotHref $url href host path]} continue
460 puts $out "<identifier>$href</identifier>"
461 puts $out "<description>$body</description>"
464 if {![RobotFileExist visited $host $path]} {
466 if {![RobotFileExist bad $host $path]} {
467 if {[RobotFileExist unvisited $host $path]} {
468 set inf [RobotFileOpen unvisited $host $path r]
469 RobotReadRecord $inf oldurl olddistance
475 if {[string length $olddistance] == 0} {
478 if {[expr $distance < $olddistance]} {
479 set outf [RobotFileOpen unvisited $host $path]
480 RobotWriteRecord $outf $url $distance
483 } elseif {[string compare $href $url]} {
484 set inf [RobotFileOpen visited $host $path r]
485 RobotReadRecord $inf xurl olddistance
487 if {[string length $olddistance] == 0} {
490 if {[expr $distance < $olddistance]} {
491 puts "OK remarking url=$url href=$href"
492 puts "olddistance = $olddistance"
493 puts "newdistance = $distance"
494 set outf [RobotFileOpen unvisited $host $path]
495 RobotWriteRecord $outf $url $distance
501 if {![info exists parm(href)]} {
505 if {[expr $distance <= $maxdistance]} {
506 set href [string trim $parm(href)]
507 if {![RobotHref $url href host path]} continue
510 puts $out "<identifier>$href</identifier>"
511 puts $out "<description></description>"
514 if {![RobotFileExist visited $host $path]} {
516 if {![RobotFileExist bad $host $path]} {
517 if {[RobotFileExist unvisited $host $path]} {
518 set inf [RobotFileOpen unvisited $host $path r]
519 RobotReadRecord $inf oldurl olddistance
525 if {[string length $olddistance] == 0} {
528 if {[expr $distance < $olddistance]} {
529 set outf [RobotFileOpen unvisited $host $path]
530 RobotWriteRecord $outf $url $distance
533 } elseif {[string compare $href $url]} {
534 set inf [RobotFileOpen visited $host $path r]
535 RobotReadRecord $inf xurl olddistance
537 if {[string length $olddistance] == 0} {
540 if {[expr $distance < $olddistance]} {
541 puts "OK remarking url=$url href=$href"
542 puts "olddistance = $olddistance"
543 puts "newdistance = $distance"
544 set outf [RobotFileOpen unvisited $host $path]
545 RobotWriteRecord $outf $url $distance
553 proc RobotsTxt {url} {
556 RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
559 proc RobotsTxt0 {v buf} {
562 foreach l [split $buf \n] {
563 if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
564 puts "cmd=$cmd arg=$arg"
565 switch -- [string tolower $cmd] {
568 set pat [string tolower $arg]*
569 set section [string match $pat $agent]
573 puts "rule [list 0 $arg]"
574 lappend $v [list 0 $arg]
579 puts "rule [list 1 $arg]"
580 lappend $v [list 1 $arg]
588 proc RobotTextPlain {url out} {
591 puts $out "<documentcontent>"
592 regsub -all {<} $URL($url,buf) {\<} content
594 puts $out "</documentcontent>"
596 if {![string compare $URL($url,path) /robots.txt]} {
601 proc RobotWriteMetadata {url out} {
607 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
608 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
609 RobotReadRecord $inf fromurl distance
612 set URL($url,dist) $distance
613 puts $out "<distance>"
614 puts $out " $distance"
615 puts $out "</distance>"
617 puts "Parsing $url distance=$distance"
618 switch $URL($url,head,content-type) {
620 if {[string length $distance]} {
621 RobotTextHtml $url $out
625 RobotTextPlain $url $out
628 set pdff [open test.pdf w]
629 puts -nonewline $pdff $URL($url,buf)
636 proc Robot200 {url} {
639 set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
640 puts -nonewline $out $URL($url,buf)
643 if {![checkrule mime $URL($url,head,content-type)]} {
644 RobotError $url mimedeny
647 set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
648 RobotWriteMetadata $url $out
651 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
654 proc RobotReadContent {url sock binary} {
657 set buffer [read $sock 16384]
658 set readCount [string length $buffer]
660 if {$readCount <= 0} {
662 RobotRestart $url $sock
663 } elseif {!$binary && [string first \0 $buffer] >= 0} {
665 RobotRestart $url $sock
667 # puts "Got $readCount bytes"
668 set URL($url,buf) $URL($url,buf)$buffer
672 proc RobotReadHeader {url sock} {
673 global URL debuglevel
675 if {$debuglevel > 1} {
676 puts "HTTP head $url"
678 if {[catch {set buffer [read $sock 2148]}]} {
680 RobotRestart $url $sock
682 set readCount [string length $buffer]
684 if {$readCount <= 0} {
686 RobotRestart $url $sock
688 # puts "Got $readCount bytes"
689 set URL($url,buf) $URL($url,buf)$buffer
691 set n [string first \r\n\r\n $URL($url,buf)]
695 set headbuf [string range $URL($url,buf) 0 $n]
697 set URL($url,buf) [string range $URL($url,buf) $n end]
699 regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
700 set lines [split $headbuf \n]
701 foreach line $lines {
702 if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
703 set URL($url,head,[string tolower $name]) [string trim $value]
706 puts "HTTP CODE $code"
707 set URL($url,state) skip
710 RobotRedirect $url $URL($url,head,location) 301
711 RobotRestart $url $sock
714 RobotRedirect $url $URL($url,head,location) 302
715 RobotRestart $url $sock
718 if {![info exists URL($url,head,content-type)]} {
719 set URL($url,head,content-type) {}
722 switch $URL($url,head,content-type) {
727 fileevent $sock readable [list RobotReadContent $url $sock $binary]
730 RobotError $url $code
731 RobotRestart $url $sock
738 proc RobotSockCancel {url sock} {
740 puts "RobotSockCancel sock=$sock url=$url"
742 RobotRestart $url $sock
745 proc RobotConnect {url sock} {
746 global URL agent acceptLanguage
748 fconfigure $sock -translation {lf crlf} -blocking 0
749 fileevent $sock readable [list RobotReadHeader $url $sock]
750 puts $sock "GET $URL($url,path) HTTP/1.0"
751 puts $sock "Host: $URL($url,host)"
752 puts $sock "User-Agent: $agent"
753 if {[string length $acceptLanguage]} {
754 puts $sock "Accept-Language: $acceptLanguage"
758 set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
765 proc RobotGetUrl {url phost} {
766 global URL robotsRunning
768 puts "Retrieve $robotsRunning url=$url"
769 if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
772 if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
776 set URL($url,method) $method
777 set URL($url,host) $host
778 set URL($url,hostport) $hostport
779 set URL($url,path) $path
780 set URL($url,state) head
783 if {[string compare $path /robots.txt]} {
785 if {![info exists URL($hostport,robots)]} {
786 puts "READING robots.txt for host $hostport"
787 if {[RobotFileExist visited $hostport /robots.txt]} {
788 set inf [RobotFileOpen visited $hostport /robots.txt r]
789 set buf [read $inf 32768]
792 set buf "User-agent: *\nAllow: /\n"
794 RobotsTxt0 URL($hostport,robots) $buf
796 if {[info exists URL($hostport,robots)]} {
797 foreach l $URL($hostport,robots) {
798 if {[string first [lindex $l 1] $path] == 0} {
805 puts "skipped due to robots.txt"
809 if [catch {set sock [socket -async $host $port]}] {
812 RobotConnect $url $sock
817 if {![llength [info commands htmlSwitch]]} {
818 set e [info sharedlibextension]
819 if {[catch {load ./tclrobot$e}]} {
824 set agent "zmbot/0.1"
825 if {![catch {set os [exec uname -s -r]}]} {
826 set agent "$agent ($os)"
841 set acceptLanguage {}
844 set l [llength $argv]
847 puts {tclrobot: usage:}
848 puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-r rules] [url ..]}
849 puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
853 # Rules: allow, deny, url
856 proc checkrule {type this} {
860 if {$debuglevel > 3} {
861 puts "CHECKRULE $type $this"
863 if {[info exist alrules]} {
865 if {$debuglevel > 3} {
869 if {[lindex $l 1] != $type} continue
870 # consider mask (! negates)
871 set masks [lindex $l 2]
873 foreach mask $masks {
874 if {$debuglevel > 4} {
875 puts "consider single mask $mask"
877 if {[string index $mask 0] == "!"} {
878 set mask [string range $mask 1 end]
879 if {[string match $mask $this]} continue
881 if {![string match $mask $this]} continue
885 if {$debuglevel > 4} {
889 # OK, we have a match
890 if {[lindex $l 0] == "allow"} {
891 if {$debuglevel > 3} {
892 puts "CHECKRULE MATCH OK"
896 if {$debuglevel > 3} {
897 puts "CHECKFULE MATCH FAIL"
903 if {$debuglevel > 3} {
904 puts "CHECKRULE MATCH OK"
913 if {[RobotHref http://www.indexdata.dk/ href host path]} {
914 if {![RobotFileExist visited $host $path]} {
915 set outf [RobotFileOpen unvisited $host $path]
916 RobotWriteRecord $outf href 0
922 proc deny {type stuff} {
925 lappend alrules [list deny $type $stuff]
928 proc allow {type stuff} {
931 lappend alrules [list allow $type $stuff]
937 set debuglevel $level
943 set arg [lindex $argv $i]
944 switch -glob -- $arg {
946 set robotsMax [string range $arg 2 end]
947 if {![string length $robotsMax]} {
948 set robotsMax [lindex $argv [incr i]]
952 set maxdistance [string range $arg 2 end]
953 if {![string length $maxdistance]} {
954 set maxdistance [lindex $argv [incr i]]
958 set dom [string range $arg 2 end]
959 if {![string length $dom]} {
960 set dom [lindex $argv [incr i]]
965 set idletime [string range $arg 2 end]
966 if {![string length $idletime]} {
967 set idletime [lindex $argv [incr i]]
971 set acceptLanguage [string range $arg 2 end]
972 if {![string length $acceptLanguage]} {
973 set acceptLanguage [lindex $argv [incr i]]
977 set rfile [string range $arg 2 end]
978 if {![string length $rfile]} {
979 set rfile [lindex $argv [incr i]]
985 if {[RobotHref http://www.indexdata.dk/ href host path]} {
986 if {![RobotFileExist visited $host $path]} {
987 set outf [RobotFileOpen unvisited $host $path]
988 RobotWriteRecord $outf href 0
997 if {![info exist domains]} {
1000 if {![info exist maxdistance]} {
1003 if {![info exist robotsMax]} {
1007 puts "domains=$domains"
1008 puts "max distance=$maxdistance"
1009 puts "max jobs=$robotsMax"
1013 while {$robotsRunning} {