2 # $Id: robot.tcl,v 1.19 2001/06/29 21:47:31 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]} {
70 puts "------------ N E X T R O U N D --------"
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} {
262 puts "Ref url = $url href=$href"
264 if {[string first { } $href] >= 0} {
267 if {[string length $href] > 256} {
270 if {[string first {?} $href] >= 0} {
273 if {[string first {?} $url] >= 0 && [string first {?} $href] >= 0} {
276 # get method (if any)
277 if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
281 if {[string compare $method http]} {
286 if {[regexp {^//([^/]+)([^\#]*)} $hpath x host surl]} {
287 if {![string length $surl]} {
290 if {[info exist domains]} {
292 foreach domain $domains {
293 if {[string match $domain $host]} {
303 regexp {^([^\#]*)} $hpath x surl
304 set host $URL($url,hostport)
306 if {![string length $surl]} {
309 if {[string first / $surl]} {
311 regexp {^([^\#?]*)} $URL($url,path) x dpart
312 set l [string last / $dpart]
313 if {[expr $l >= 0]} {
314 set surl [string range $dpart 0 $l]$surl
316 set surl $dpart/$surl
319 set surllist [split $surl /]
322 foreach c $surllist {
327 set path [lrange $path 0 $pathl]
340 set path [join $path /]
344 regsub -all {~} $path {%7E} path
345 set href "$method://$host$path"
346 puts "Ref href = $href"
350 proc RobotError {url code} {
353 puts "Bad URL $url, $code"
356 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
357 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
358 RobotReadRecord $inf fromurl distance
361 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
362 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
363 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
364 RobotWriteRecord $outf $fromurl $distance
369 proc RobotRedirect {url tourl code} {
372 puts "Redirecting from $url to $tourl"
376 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
377 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
378 RobotReadRecord $inf fromurl distance
381 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
382 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
383 RobotWriteRecord $outf $fromurl $distance
386 if {[RobotHref $url tourl host path]} {
387 if {![RobotFileExist visited $host $path]} {
388 if {![RobotFileExist unvisited $host $path]} {
389 set outf [RobotFileOpen unvisited $host $path]
390 RobotWriteRecord $outf $fromurl $distance
395 set inf [RobotFileOpen visited $host $path r]
396 RobotReadRecord $inf oldurl olddistance
398 if {[string length $olddistance] == 0} {
401 if {[string length $distance] == 0} {
404 puts "distance=$distance olddistance=$olddistance"
405 if {[expr $distance < $olddistance]} {
406 set outf [RobotFileOpen unvisited $host $path]
407 RobotWriteRecord $outf $tourl $distance
412 if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
418 proc RobotTextHtml {url out} {
419 global URL maxDistance
422 if {$maxDistance < 1000 && [info exists URL($url,dist)]} {
423 set distance [expr $URL($url,dist) + 1]
425 htmlSwitch $URL($url,buf) \
427 puts $out "<title>$body</title>"
429 puts -nonewline $out "<meta"
430 foreach a [array names parm] {
431 puts -nonewline $out " $a"
432 puts -nonewline $out {="}
433 puts -nonewline $out $parm($a)
434 puts -nonewline $out {"}
438 regsub -all -nocase {<script([^<]|(<!.*>))*</script>} $body {} abody
439 regsub -all {<[^\>]+>} $abody {} nbody
440 puts $out "<documentcontent>"
442 puts $out "</documentcontent>"
444 if {![info exists parm(href)]} {
448 if {[expr $distance <= $maxDistance]} {
449 set href [string trim $parm(href)]
450 if {![RobotHref $url href host path]} continue
453 puts $out "<identifier>$href</identifier>"
454 puts $out "<description>$body</description>"
457 if {![RobotFileExist visited $host $path]} {
459 if {![RobotFileExist bad $host $path]} {
460 if {[RobotFileExist unvisited $host $path]} {
461 set inf [RobotFileOpen unvisited $host $path r]
462 RobotReadRecord $inf oldurl olddistance
468 if {[string length $olddistance] == 0} {
471 if {[expr $distance < $olddistance]} {
472 set outf [RobotFileOpen unvisited $host $path]
473 RobotWriteRecord $outf $url $distance
476 } elseif {[string compare $href $url]} {
477 set inf [RobotFileOpen visited $host $path r]
478 RobotReadRecord $inf xurl olddistance
480 if {[string length $olddistance] == 0} {
483 if {[expr $distance < $olddistance]} {
484 puts "OK remarking url=$url href=$href"
485 puts "olddistance = $olddistance"
486 puts "newdistance = $distance"
487 set outf [RobotFileOpen unvisited $host $path]
488 RobotWriteRecord $outf $url $distance
494 if {![info exists parm(href)]} {
498 if {[expr $distance <= $maxDistance]} {
499 set href [string trim $parm(href)]
500 if {![RobotHref $url href host path]} continue
503 puts $out "<identifier>$href</identifier>"
504 puts $out "<description></description>"
507 if {![RobotFileExist visited $host $path]} {
509 if {![RobotFileExist bad $host $path]} {
510 if {[RobotFileExist unvisited $host $path]} {
511 set inf [RobotFileOpen unvisited $host $path r]
512 RobotReadRecord $inf oldurl olddistance
518 if {[string length $olddistance] == 0} {
521 if {[expr $distance < $olddistance]} {
522 set outf [RobotFileOpen unvisited $host $path]
523 RobotWriteRecord $outf $url $distance
526 } elseif {[string compare $href $url]} {
527 set inf [RobotFileOpen visited $host $path r]
528 RobotReadRecord $inf xurl olddistance
530 if {[string length $olddistance] == 0} {
533 if {[expr $distance < $olddistance]} {
534 puts "OK remarking url=$url href=$href"
535 puts "olddistance = $olddistance"
536 puts "newdistance = $distance"
537 set outf [RobotFileOpen unvisited $host $path]
538 RobotWriteRecord $outf $url $distance
546 proc RobotsTxt {url} {
549 RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
552 proc RobotsTxt0 {v buf} {
555 foreach l [split $buf \n] {
556 if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
557 puts "cmd=$cmd arg=$arg"
558 switch -- [string tolower $cmd] {
561 set pat [string tolower $arg]*
562 set section [string match $pat $agent]
566 puts "rule [list 0 $arg]"
567 lappend $v [list 0 $arg]
572 puts "rule [list 1 $arg]"
573 lappend $v [list 1 $arg]
581 proc RobotTextPlain {url out} {
584 puts $out "<documentcontent>"
585 regsub -all {<} $URL($url,buf) {\<} content
587 puts $out "</documentcontent>"
589 if {![string compare $URL($url,path) /robots.txt]} {
594 proc Robot200 {url} {
597 set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
598 puts -nonewline $out $URL($url,buf)
601 set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
605 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
606 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
607 RobotReadRecord $inf fromurl distance
610 set URL($url,dist) $distance
611 puts $out "<distance>"
612 puts $out " $distance"
613 puts $out "</distance>"
615 puts "Parsing $url distance=$distance"
616 switch $URL($url,head,content-type) {
618 if {[string length $distance]} {
619 RobotTextHtml $url $out
623 RobotTextPlain $url $out
626 set pdff [open test.pdf w]
627 puts -nonewline $pdff $URL($url,buf)
633 # puts "Parsing done"
634 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
637 proc RobotReadContent {url sock binary} {
640 puts "RobotReadContent $url"
641 set buffer [read $sock 16384]
642 set readCount [string length $buffer]
644 if {$readCount <= 0} {
646 RobotRestart $url $sock
647 } elseif {!$binary && [string first \0 $buffer] >= 0} {
649 RobotRestart $url $sock
651 # puts "Got $readCount bytes"
652 set URL($url,buf) $URL($url,buf)$buffer
656 proc RobotReadHeader {url sock} {
659 puts "RobotReadHeader $url"
660 if {[catch {set buffer [read $sock 2148]}]} {
662 RobotRestart $url $sock
664 set readCount [string length $buffer]
666 if {$readCount <= 0} {
668 RobotRestart $url $sock
670 # puts "Got $readCount bytes"
671 set URL($url,buf) $URL($url,buf)$buffer
673 set n [string first \r\n\r\n $URL($url,buf)]
677 set headbuf [string range $URL($url,buf) 0 $n]
679 set URL($url,buf) [string range $URL($url,buf) $n end]
681 regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
682 set lines [split $headbuf \n]
683 foreach line $lines {
684 if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
685 set URL($url,head,[string tolower $name]) [string trim $value]
689 set URL($url,state) skip
692 RobotRedirect $url $URL($url,head,location) 301
693 RobotRestart $url $sock
696 RobotRedirect $url $URL($url,head,location) 302
697 RobotRestart $url $sock
700 if {![info exists URL($url,head,content-type)]} {
701 set URL($url,head,content-type) {}
704 switch $URL($url,head,content-type) {
709 fileevent $sock readable [list RobotReadContent $url $sock $binary]
712 RobotError $url $code
713 RobotRestart $url $sock
720 proc RobotSockCancel {url sock} {
722 puts "RobotSockCancel sock=$sock url=$url"
724 RobotRestart $url $sock
727 proc RobotConnect {url sock} {
728 global URL agent acceptLanguage
730 fconfigure $sock -translation {lf crlf} -blocking 0
731 fileevent $sock readable [list RobotReadHeader $url $sock]
732 puts $sock "GET $URL($url,path) HTTP/1.0"
733 puts $sock "Host: $URL($url,host)"
734 puts $sock "User-Agent: $agent"
735 if {[string length $acceptLanguage]} {
736 puts $sock "Accept-Language: $acceptLanguage"
740 set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
747 proc RobotGetUrl {url phost} {
748 global URL robotsRunning
750 puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
751 if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
754 if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
758 set URL($url,method) $method
759 set URL($url,host) $host
760 set URL($url,hostport) $hostport
761 set URL($url,path) $path
762 set URL($url,state) head
765 if {[string compare $path /robots.txt]} {
767 if {![info exists URL($hostport,robots)]} {
768 puts "READING robots.txt for host $hostport"
769 if {[RobotFileExist visited $hostport /robots.txt]} {
770 set inf [RobotFileOpen visited $hostport /robots.txt r]
771 set buf [read $inf 32768]
774 set buf "User-agent: *\nAllow: /\n"
776 RobotsTxt0 URL($hostport,robots) $buf
778 if {[info exists URL($hostport,robots)]} {
779 foreach l $URL($hostport,robots) {
780 if {[string first [lindex $l 1] $path] == 0} {
787 puts "skipped due to robots.txt"
791 if [catch {set sock [socket -async $host $port]}] {
794 RobotConnect $url $sock
799 if {![llength [info commands htmlSwitch]]} {
800 set e [info sharedlibextension]
801 if {[catch {load ./tclrobot$e}]} {
806 set agent "zmbot/0.0"
807 if {![catch {set os [exec uname -s -r]}]} {
808 set agent "$agent ($os)"
823 set acceptLanguage {}
826 set l [llength $argv]
830 set url "http://www.sportsfiskeren.dk/sportsfiskeren/corner/index.htm"
831 set href "../../data/../../data2/newsovs.asp?Mode=5"
833 set URL($url,path) /sportsfiskeren/corner/index.htm
834 set URL($url,hostport) www.sportsfiskeren.dk
835 RobotHref $url href host path
840 puts {tclrobot: usage [-j jobs] [-i idle] [-c count] [-d domain] [url ..]}
841 puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
846 set arg [lindex $argv $i]
847 switch -glob -- $arg {
849 set robotsMax [string range $arg 2 end]
850 if {![string length $robotsMax]} {
851 set robotsMax [lindex $argv [incr i]]
855 set maxDistance [string range $arg 2 end]
856 if {![string length $maxDistance]} {
857 set maxDistance [lindex $argv [incr i]]
861 set dom [string range $arg 2 end]
862 if {![string length $dom]} {
863 set dom [lindex $argv [incr i]]
868 set idleTime [string range $arg 2 end]
869 if {![string length $idleTime]} {
870 set idleTime [lindex $argv [incr i]]
874 set acceptLanguage [string range $arg 2 end]
875 if {![string length $acceptLanguage]} {
876 set acceptLanguage [lindex $argv [incr i]]
881 if {[RobotHref http://www.indexdata.dk/ href host path]} {
882 if {![RobotFileExist visited $host $path]} {
883 set outf [RobotFileOpen unvisited $host $path]
884 RobotWriteRecord $outf href 0
893 if {![info exist domains]} {
896 if {![info exist maxDistance]} {
899 if {![info exist robotsMax]} {
903 puts "domains=$domains"
904 puts "max distance=$maxDistance"
905 puts "max jobs=$robotsMax"
909 while {$robotsRunning} {