2 # $Id: robot.tcl,v 1.16 2001/06/06 07:10:31 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
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 {?} $url] >= 0 && [string first {?} $href] >= 0} {
273 # get method (if any)
274 if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
278 if {[string compare $method http]} {
283 if {[regexp {^//([^/]+)([^\#]*)} $hpath x host surl]} {
284 if {![string length $surl]} {
287 if {[info exist domains]} {
289 foreach domain $domains {
290 if {[string match $domain $host]} {
300 regexp {^([^\#]*)} $hpath x surl
301 set host $URL($url,hostport)
303 if {![string length $surl]} {
306 if {[string first / $surl]} {
308 regexp {^([^\#?]*)} $URL($url,path) x dpart
309 set l [string last / $dpart]
310 if {[expr $l >= 0]} {
311 set surl [string range $dpart 0 $l]$surl
313 set surl $dpart/$surl
316 set c [split $surl /]
319 set path [lindex $c $i]
322 switch -- [lindex $c $i] {
333 set path [lindex $c $i]/$path
338 regsub -all {~} $path {%7E} path
339 set href "$method://$host$path"
340 puts "Ref href = $href"
344 proc RobotError {url code} {
347 puts "Bad URL $url, $code"
350 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
351 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
352 RobotReadRecord $inf fromurl distance
355 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
356 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
357 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
358 RobotWriteRecord $outf $fromurl $distance
363 proc RobotRedirect {url tourl code} {
366 puts "Redirecting from $url to $tourl"
370 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
371 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
372 RobotReadRecord $inf fromurl distance
375 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
376 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
377 RobotWriteRecord $outf $fromurl $distance
380 if {[RobotHref $url tourl host path]} {
381 if {![RobotFileExist visited $host $path]} {
382 if {![RobotFileExist unvisited $host $path]} {
383 set outf [RobotFileOpen unvisited $host $path]
384 RobotWriteRecord $outf $fromurl $distance
389 set inf [RobotFileOpen visited $host $path r]
390 RobotReadRecord $inf oldurl olddistance
392 if {[string length $olddistance] == 0} {
395 if {[string length $distance] == 0} {
398 puts "distance=$distance olddistance=$olddistance"
399 if {[expr $distance < $olddistance]} {
400 set outf [RobotFileOpen unvisited $host $path]
401 RobotWriteRecord $outf $tourl $distance
406 if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
412 proc RobotTextHtml {url out} {
413 global URL maxDistance
416 if {$maxDistance < 1000 && [info exists URL($url,dist)]} {
417 set distance [expr $URL($url,dist) + 1]
419 htmlSwitch $URL($url,buf) \
421 puts $out "<title>$body</title>"
423 puts -nonewline $out "<meta"
424 foreach a [array names parm] {
425 puts -nonewline $out " $a"
426 puts -nonewline $out {="}
427 puts -nonewline $out $parm($a)
428 puts -nonewline $out {"}
432 regsub -all -nocase {<script.*</script>} $body {} abody
433 regsub -all {<[^\>]+>} $abody {} nbody
434 puts $out "<documentcontent>"
436 puts $out "</documentcontent>"
438 if {![info exists parm(href)]} {
442 if {[expr $distance <= $maxDistance]} {
443 set href [string trim $parm(href)]
444 if {![RobotHref $url href host path]} continue
447 puts $out "<identifier>$href</identifier>"
448 puts $out "<description>$body</description>"
451 if {![RobotFileExist visited $host $path]} {
453 if {![RobotFileExist bad $host $path]} {
454 if {[RobotFileExist unvisited $host $path]} {
455 set inf [RobotFileOpen unvisited $host $path r]
456 RobotReadRecord $inf oldurl olddistance
462 if {[string length $olddistance] == 0} {
465 if {[expr $distance < $olddistance]} {
466 set outf [RobotFileOpen unvisited $host $path]
467 RobotWriteRecord $outf $url $distance
470 } elseif {[string compare $href $url]} {
471 set inf [RobotFileOpen visited $host $path r]
472 RobotReadRecord $inf xurl olddistance
474 if {[string length $olddistance] == 0} {
477 if {[expr $distance < $olddistance]} {
478 puts "OK remarking url=$url href=$href"
479 puts "olddistance = $olddistance"
480 puts "newdistance = $distance"
481 set outf [RobotFileOpen unvisited $host $path]
482 RobotWriteRecord $outf $url $distance
488 if {![info exists parm(href)]} {
492 if {[expr $distance <= $maxDistance]} {
493 set href [string trim $parm(href)]
494 if {![RobotHref $url href host path]} continue
497 puts $out "<identifier>$href</identifier>"
498 puts $out "<description></description>"
501 if {![RobotFileExist visited $host $path]} {
503 if {![RobotFileExist bad $host $path]} {
504 if {[RobotFileExist unvisited $host $path]} {
505 set inf [RobotFileOpen unvisited $host $path r]
506 RobotReadRecord $inf oldurl olddistance
512 if {[string length $olddistance] == 0} {
515 if {[expr $distance < $olddistance]} {
516 set outf [RobotFileOpen unvisited $host $path]
517 RobotWriteRecord $outf $url $distance
520 } elseif {[string compare $href $url]} {
521 set inf [RobotFileOpen visited $host $path r]
522 RobotReadRecord $inf xurl olddistance
524 if {[string length $olddistance] == 0} {
527 if {[expr $distance < $olddistance]} {
528 puts "OK remarking url=$url href=$href"
529 puts "olddistance = $olddistance"
530 puts "newdistance = $distance"
531 set outf [RobotFileOpen unvisited $host $path]
532 RobotWriteRecord $outf $url $distance
540 proc RobotsTxt {url} {
543 RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
546 proc RobotsTxt0 {v buf} {
549 foreach l [split $buf \n] {
550 if {[regexp {([-A-Za-z]+):[ \t]*([^\#\t ]+)} $l match cmd arg]} {
551 puts "cmd=$cmd arg=$arg"
552 switch -- [string tolower $cmd] {
555 set pat [string tolower $arg]*
556 set section [string match $pat $agent]
560 puts "rule [list 0 $arg]"
561 lappend $v [list 0 $arg]
566 puts "rule [list 1 $arg]"
567 lappend $v [list 1 $arg]
575 proc RobotTextPlain {url out} {
578 puts $out "<documentcontent>"
579 puts $out $URL($url,buf)
580 puts $out "</documentcontent>"
582 if {![string compare $URL($url,path) /robots.txt]} {
587 proc Robot200 {url} {
590 set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
594 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
595 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
596 RobotReadRecord $inf fromurl distance
599 set URL($url,dist) $distance
600 puts $out "<distance>"
601 puts $out " $distance"
602 puts $out "</distance>"
604 puts "Parsing $url distance=$distance"
605 switch $URL($url,head,content-type) {
607 if {[string length $distance]} {
608 RobotTextHtml $url $out
612 RobotTextPlain $url $out
615 set pdff [open test.pdf w]
616 puts -nonewline $pdff $URL($url,buf)
622 # puts "Parsing done"
623 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
626 proc RobotReadContent {url sock binary} {
629 puts "RobotReadContent $url"
630 set buffer [read $sock 16384]
631 set readCount [string length $buffer]
633 if {$readCount <= 0} {
635 RobotRestart $url $sock
636 } elseif {!$binary && [string first \0 $buffer] >= 0} {
638 RobotRestart $url $sock
640 # puts "Got $readCount bytes"
641 set URL($url,buf) $URL($url,buf)$buffer
645 proc RobotReadHeader {url sock} {
648 puts "RobotReadHeader $url"
649 if {[catch {set buffer [read $sock 2148]}]} {
651 RobotRestart $url $sock
653 set readCount [string length $buffer]
655 if {$readCount <= 0} {
657 RobotRestart $url $sock
659 # puts "Got $readCount bytes"
660 set URL($url,buf) $URL($url,buf)$buffer
662 set n [string first \r\n\r\n $URL($url,buf)]
666 set headbuf [string range $URL($url,buf) 0 $n]
668 set URL($url,buf) [string range $URL($url,buf) $n end]
670 regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
671 set lines [split $headbuf \n]
672 foreach line $lines {
673 if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
674 set URL($url,head,[string tolower $name]) [string trim $value]
678 set URL($url,state) skip
681 RobotRedirect $url $URL($url,head,location) 301
682 RobotRestart $url $sock
685 RobotRedirect $url $URL($url,head,location) 302
686 RobotRestart $url $sock
689 if {![info exists URL($url,head,content-type)]} {
690 set URL($url,head,content-type) {}
693 switch $URL($url,head,content-type) {
698 fileevent $sock readable [list RobotReadContent $url $sock $binary]
701 RobotError $url $code
702 RobotRestart $url $sock
709 proc RobotSockCancel {url sock} {
711 puts "RobotSockCancel sock=$sock url=$url"
713 RobotRestart $url $sock
716 proc RobotConnect {url sock} {
719 fconfigure $sock -translation {lf crlf} -blocking 0
720 fileevent $sock readable [list RobotReadHeader $url $sock]
721 puts $sock "GET $URL($url,path) HTTP/1.0"
722 puts $sock "Host: $URL($url,host)"
723 puts $sock "User-Agent: $agent"
726 set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
733 proc RobotGetUrl {url phost} {
734 global URL robotsRunning
736 puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
737 if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
740 if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
744 set URL($url,method) $method
745 set URL($url,host) $host
746 set URL($url,hostport) $hostport
747 set URL($url,path) $path
748 set URL($url,state) head
751 if {[string compare $path /robots.txt]} {
753 if {![info exists URL($hostport,robots)]} {
754 puts "READING robots.txt for host $hostport"
755 if {[RobotFileExist visited $hostport /robots.txt]} {
756 set inf [RobotFileOpen visited $hostport /robots.txt r]
757 set buf [read $inf 32768]
760 set buf "User-agent: *\nAllow: /\n"
762 RobotsTxt0 URL($hostport,robots) $buf
764 if {[info exists URL($hostport,robots)]} {
765 foreach l $URL($hostport,robots) {
766 if {[string first [lindex $l 1] $path] == 0} {
773 puts "skipped due to robots.txt"
777 if [catch {set sock [socket -async $host $port]}] {
780 RobotConnect $url $sock
785 if {![llength [info commands htmlSwitch]]} {
786 set e [info sharedlibextension]
787 if {[catch {load ./tclrobot$e}]} {
792 set agent "zmbot/0.0"
793 if {![catch {set os [exec uname -s -r]}]} {
794 set agent "$agent ($os)"
811 set l [llength $argv]
814 puts {tclrobot: usage [-j jobs] [-i idle] [-c count] [-d domain] [url ..]}
815 puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
820 set arg [lindex $argv $i]
821 switch -glob -- $arg {
823 set robotsMax [string range $arg 2 end]
824 if {![string length $robotsMax]} {
825 set robotsMax [lindex $argv [incr i]]
829 set maxDistance [string range $arg 2 end]
830 if {![string length $maxDistance]} {
831 set maxDistance [lindex $argv [incr i]]
835 set dom [string range $arg 2 end]
836 if {![string length $dom]} {
837 set dom [lindex $argv [incr i]]
842 set idleTime [string range $arg 2 end]
843 if {![string length $idleTime]} {
844 set idleTime [lindex $argv [incr i]]
849 if {[RobotHref http://www.indexdata.dk/ href host path]} {
850 if {![RobotFileExist visited $host $path]} {
851 set outf [RobotFileOpen unvisited $host $path]
852 RobotWriteRecord $outf href 0
861 if {![info exist domains]} {
864 if {![info exist maxDistance]} {
867 if {![info exist robotsMax]} {
871 puts "domains=$domains"
872 puts "max distance=$maxDistance"
873 puts "max jobs=$robotsMax"
877 while {$robotsRunning} {