2 # $Id: robot.tcl,v 1.11 2001/01/23 11:26:43 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 first {?} $url] >= 0 && [string first {?} $href] >= 0} {
270 # get method (if any)
271 if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
275 if {[string compare $method http]} {
280 if {[regexp {^//([^/]+)([^\#]*)} $hpath x host surl]} {
281 if {![string length $surl]} {
285 foreach domain $domains {
286 if {[string match $domain $host]} {
295 regexp {^([^\#]*)} $hpath x surl
296 set host $URL($url,hostport)
298 if {![string length $surl]} {
301 if {[string first / $surl]} {
303 regexp {^([^\#?]*)} $URL($url,path) x dpart
304 set l [string last / $dpart]
305 if {[expr $l >= 0]} {
306 set surl [string range $dpart 0 $l]$surl
308 set surl $dpart/$surl
311 set c [split $surl /]
314 set path [lindex $c $i]
317 switch -- [lindex $c $i] {
328 set path [lindex $c $i]/$path
333 regsub -all {~} $path {%7E} path
334 set href "$method://$host$path"
335 puts "Ref href = $href"
339 proc RobotError {url code} {
342 puts "Bad URL $url, $code"
345 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
346 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
347 RobotReadRecord $inf fromurl distance
350 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
351 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
352 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
353 RobotWriteRecord $outf $fromurl $distance
358 proc RobotRedirect {url tourl code} {
361 puts "Redirecting from $url to $tourl"
365 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
366 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
367 RobotReadRecord $inf fromurl distance
370 if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
371 set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
372 RobotWriteRecord $outf $fromurl $distance
375 if {[RobotHref $url tourl host path]} {
376 if {![RobotFileExist visited $host $path]} {
377 if {![RobotFileExist unvisited $host $path]} {
378 set outf [RobotFileOpen unvisited $host $path]
379 RobotWriteRecord $outf $fromurl $distance
384 set inf [RobotFileOpen visited $host $path r]
385 RobotReadRecord $inf oldurl olddistance
387 if {[string length $olddistance] == 0} {
390 if {[string length $distance] == 0} {
393 puts "distance=$distance olddistance=$olddistance"
394 if {[expr $distance < $olddistance]} {
395 set outf [RobotFileOpen unvisited $host $path]
396 RobotWriteRecord $outf $tourl $distance
401 if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
407 proc RobotTextHtml {url out} {
408 global URL maxDistance
411 if {$maxDistance < 1000 && [info exists URL($url,dist)]} {
412 set distance [expr $URL($url,dist) + 1]
414 htmlSwitch $URL($url,buf) \
416 puts $out "<title>$body</title>"
418 puts -nonewline $out "<meta"
419 foreach a [array names parm] {
420 puts -nonewline $out " $a"
421 puts -nonewline $out {="}
422 puts -nonewline $out $parm($a)
423 puts -nonewline $out {"}
427 regsub -all -nocase {<script.*</script>} $body {} abody
428 regsub -all {<[^\>]+>} $abody {} nbody
429 puts $out "<documentcontent>"
431 puts $out "</documentcontent>"
433 if {![info exists parm(href)]} {
437 if {[expr $distance <= $maxDistance]} {
438 set href [string trim $parm(href)]
439 if {![RobotHref $url href host path]} continue
442 puts $out "<identifier>$href</identifier>"
443 puts $out "<description>$body</description>"
446 if {![RobotFileExist visited $host $path]} {
448 if {![RobotFileExist bad $host $path]} {
449 if {[RobotFileExist unvisited $host $path]} {
450 set inf [RobotFileOpen unvisited $host $path r]
451 RobotReadRecord $inf oldurl olddistance
457 if {[string length $olddistance] == 0} {
460 if {[expr $distance < $olddistance]} {
461 set outf [RobotFileOpen unvisited $host $path]
462 RobotWriteRecord $outf $url $distance
465 } elseif {[string compare $href $url]} {
466 set inf [RobotFileOpen visited $host $path r]
467 RobotReadRecord $inf xurl olddistance
469 if {[string length $olddistance] == 0} {
472 if {[expr $distance < $olddistance]} {
473 puts "OK remarking url=$url href=$href"
474 puts "olddistance = $olddistance"
475 puts "newdistance = $distance"
476 set outf [RobotFileOpen unvisited $host $path]
477 RobotWriteRecord $outf $url $distance
485 proc RobotsTxt {url} {
488 RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
491 proc RobotsTxt0 {v buf} {
494 foreach l [split $buf \n] {
495 if {[regexp {([-A-Za-z]+):[ \t]*([^\#]+)} $l match cmd arg]} {
496 puts "cmd=$cmd arg=$arg"
500 set pat [string tolower $arg]*
501 set section [string match $pat $agent]
505 puts "rule [list 0 $arg]"
506 lappend $v [list 0 $arg]
511 puts "rule [list 1 $arg]"
512 lappend $v [list 1 $arg]
520 proc RobotTextPlain {url out} {
523 puts $out "<documentcontent>"
524 puts $out $URL($url,buf)
525 puts $out "</documentcontent>"
527 if {![string compare $URL($url,path) /robots.txt]} {
532 proc Robot200 {url} {
535 set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
539 if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
540 set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
541 RobotReadRecord $inf fromurl distance
544 set URL($url,dist) $distance
545 puts $out "<distance>"
546 puts $out " $distance"
547 puts $out "</distance>"
549 puts "Parsing $url distance=$distance"
550 switch $URL($url,head,content-type) {
552 if {[string length $distance]} {
553 RobotTextHtml $url $out
557 RobotTextPlain $url $out
560 set pdff [open test.pdf w]
561 puts -nonewline $pdff $URL($url,buf)
567 # puts "Parsing done"
568 RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
571 proc RobotReadContent {url sock binary} {
574 puts "RobotReadContent $url"
575 set buffer [read $sock 16384]
576 set readCount [string length $buffer]
578 if {$readCount <= 0} {
580 RobotRestart $url $sock
581 } elseif {!$binary && [string first \0 $buffer] >= 0} {
583 RobotRestart $url $sock
585 # puts "Got $readCount bytes"
586 set URL($url,buf) $URL($url,buf)$buffer
590 proc RobotReadHeader {url sock} {
593 puts "RobotReadHeader $url"
594 if {[catch {set buffer [read $sock 2148]}]} {
596 RobotRestart $url $sock
598 set readCount [string length $buffer]
600 if {$readCount <= 0} {
602 RobotRestart $url $sock
604 # puts "Got $readCount bytes"
605 set URL($url,buf) $URL($url,buf)$buffer
607 set n [string first \r\n\r\n $URL($url,buf)]
611 set headbuf [string range $URL($url,buf) 0 $n]
613 set URL($url,buf) [string range $URL($url,buf) $n end]
615 regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
616 set lines [split $headbuf \n]
617 foreach line $lines {
618 if {[regexp {^([^:]+):[ ]+(.*)} $line x name value]} {
619 set URL($url,head,[string tolower $name]) [string trim $value]
623 set URL($url,state) skip
626 RobotRedirect $url $URL($url,head,location) 301
627 RobotRestart $url $sock
630 RobotRedirect $url $URL($url,head,location) 302
631 RobotRestart $url $sock
634 if {![info exists URL($url,head,content-type)]} {
635 set URL($url,head,content-type) {}
638 switch $URL($url,head,content-type) {
643 fileevent $sock readable [list RobotReadContent $url $sock $binary]
646 RobotError $url $code
647 RobotRestart $url $sock
654 proc RobotSockCancel {url sock} {
656 puts "RobotSockCancel sock=$sock url=$url"
658 RobotRestart $url $sock
661 proc RobotConnect {url sock} {
664 fconfigure $sock -translation {lf crlf} -blocking 0
665 fileevent $sock readable [list RobotReadHeader $url $sock]
666 puts $sock "GET $URL($url,path) HTTP/1.0"
667 puts $sock "Host: $URL($url,host)"
668 puts $sock "User-Agent: $agent"
671 set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
678 proc RobotGetUrl {url phost} {
679 global URL robotsRunning
681 puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
682 if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
685 if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
689 set URL($url,method) $method
690 set URL($url,host) $host
691 set URL($url,hostport) $hostport
692 set URL($url,path) $path
693 set URL($url,state) head
696 if {[string compare $path /robots.txt]} {
698 if {![info exists URL($hostport,robots)]} {
699 puts "READING robots.txt for host $hostport"
700 if {[RobotFileExist visited $hostport /robots.txt]} {
701 set inf [RobotFileOpen visited $hostport /robots.txt r]
702 set buf [read $inf 32768]
705 set buf "User-Agent: *\nAllow: /\n"
707 RobotsTxt0 URL($hostport,robots) $buf
709 if {[info exists URL($hostport,robots)]} {
710 foreach l $URL($hostport,robots) {
711 if {[string first [lindex $l 1] $path] == 0} {
721 if [catch {set sock [socket -async $host $port]}] {
724 RobotConnect $url $sock
729 if {![llength [info commands htmlSwitch]]} {
730 set e [info sharedlibextension]
731 if {[catch {load ./tclrobot$e}]} {
736 set agent "zmbot/0.0"
737 if {![catch {set os [exec uname -s -r]}]} {
738 set agent "$agent ($os)"
755 set l [llength $argv]
758 puts {tclrobot: usage [-j jobs] [-c count] [-d domain] [url ..]}
759 puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
764 set arg [lindex $argv $i]
765 switch -glob -- $arg {
767 set robotsMax [string range $arg 2 end]
768 if {![string length $robotsMax]} {
769 set robotsMax [lindex $argv [incr i]]
773 set maxDistance [string range $arg 2 end]
774 if {![string length $maxDistance]} {
775 set maxDistance [lindex $argv [incr i]]
779 set dom [string range $arg 2 end]
780 if {![string length $dom]} {
781 set dom [lindex $argv [incr i]]
787 if {[RobotHref http://www.indexdata.dk/ href host path]} {
788 if {![RobotFileExist visited $host $path]} {
789 set outf [RobotFileOpen unvisited $host $path]
790 RobotWriteRecord $outf href 0
799 if {![info exist domains]} {
802 if {![info exist maxDistance]} {
805 if {![info exist robotsMax]} {
809 puts "domains=$domains"
810 puts "max distance=$maxDistance"
811 puts "max jobs=$robotsMax"
815 while {$robotsRunning} {