X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=73d558aad3893ad0b0b823f929247a5cb5d02dac;hb=4355628830cd0f9e27c059d20254d8e1c30896eb;hp=b6466c347186d9954fbdda74c78ed7054fab76c5;hpb=9d3f82cd1140362487d8fa6372cac1b24a49d21e;p=tclrobot.git

diff --git a/robot.tcl b/robot.tcl
index b6466c3..73d558a 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,5 +1,5 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.29 2001/11/14 09:15:23 adam Exp $
+# $Id: robot.tcl,v 1.45 2003/06/11 10:29:41 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
     # puts "RobotFileNext1 area=$area lead=$lead"
@@ -9,14 +9,19 @@ proc RobotFileNext1 {area lead} {
     foreach n $ns {
 	if {[file isfile $n]} {
             set off [string last / $n]
-	    incr off 2
-            return $lead/[string range $n $off end]
+	    # skip /
+	    incr off
+	    set end [string length $n]
+	    # skip _.tkl
+	    incr end -6
+            return $lead/[string range $n $off $end]
         }
     }
     foreach n $ns {
 	if {[file isdirectory $n]} {
             set off [string last / $n]
-	    incr off 2
+	    # skip /
+	    incr off
             set sb [RobotFileNext1 $n $lead/[string range $n $off end]]
             if {[string length $sb]} {
                 return $sb
@@ -27,6 +32,7 @@ proc RobotFileNext1 {area lead} {
 }
 
 proc RobotWriteRecord {outf fromurl distance} {
+    puts $outf {<?xml version="1.0" encoding="ISO-8859-1" standalone="yes"?>}
     puts $outf "<zmbot>"
     puts $outf "<distance>"
     puts $outf $distance
@@ -42,6 +48,7 @@ proc RobotReadRecord {inf fromurlx distancex} {
     upvar $distancex distance
     gets $inf
     gets $inf
+    gets $inf
     set distance [string trim [gets $inf]]
     # puts "got distance = $distance"
     gets $inf
@@ -49,34 +56,36 @@ proc RobotReadRecord {inf fromurlx distancex} {
     set fromurl [string trim [gets $inf]]
 }
 
-proc RobotFileNext {area} {
-    global robotSeq
+proc RobotFileNext {task area} {
+    global control
     global idletime ns
     global status
 
-    # puts "RobotFileNext robotSeq=$robotSeq"
-    if {$robotSeq < 0} {
+    # puts "RobotFileNext seq=$control($task,seq)"
+    if {$control($task,seq) < 0} {
 	return {}
     }
-    if {$robotSeq == 0} {
-	if {[catch {set ns [glob ${area}/*]}]} {
-	    return {}
+    if {$control($task,seq) == 0} {
+	if {[catch {set ns($task) [glob $task/$area/*]}]} {
+	    return done
 	}
     }
-    set off [string length $area]
+    # puts "ns=$ns($task)"
+    set off [string length $task/$area]
     incr off
-    set n [lindex $ns $robotSeq]
+    set n [lindex $ns($task) $control($task,seq)]
+    # puts "n=$n"
     if {![string length $n]} {
-	set robotSeq -1
+	set control($task,seq) -1
 	flush stdout
-        set statusfile [open status w]
-        puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+        set statusfile [open $task/status w]
+        puts $statusfile "$status($task,unvisited) $status($task,bad) $status($task,visited)"
         close $statusfile
 	return wait
     }
-    incr robotSeq
-    if {[file isfile $n/frobots.txt]} {
-	puts "ok returning http://[string range $n $off end]/robots.txt"
+    incr control($task,seq)
+    if {[file isfile $n/robots.txt_.tkl]} {
+	# puts "ok returning http://[string range $n $off end]/robots.txt"
 	return http://[string range $n $off end]/robots.txt
     } elseif {[file isdirectory $n]} {
 	set sb [RobotFileNext1 $n http://[string range $n $off end]]
@@ -85,51 +94,36 @@ proc RobotFileNext {area} {
 	}
     }
     puts "no more work at end of RobotFileNext n=$n"
-    puts "ns=$ns"
+    puts "ns=$ns($task)"
     return {}
 }
 
 
-proc RobotFileExist {area host path} {
+proc RobotFileExist {task area host path} {
     global debuglevel
 
     if {$debuglevel > 3} {
         puts "RobotFileExist begin area=$area host=$host path=$path"
     }
-    set lpath [split $path /]
-    set l [llength $lpath]
-    incr l -1
-    set t [lindex $lpath $l]
-    incr l -1
-    set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
-    if {$debuglevel > 3} {
-        puts "RobotFileExist end npath=$npath"
-    }
-    return [file exists $npath]
+    return [file exists $task/$area/$host${path}_.tkl]
 }
 
-proc RobotFileUnlink {area host path} {
+proc RobotFileUnlink {task area host path} {
     global status
     # puts "RobotFileUnlink begin"
     # puts "area=$area host=$host path=$path"
-    set lpath [split $path /]
-    set l [llength $lpath]
-    incr l -1
-    set t [lindex $lpath $l]
-    incr l -1
-    set npath $area/$host[join [lrange $lpath 0 $l] /d]/f$t
+    set npath $task/$area/$host${path}_.tkl
     # puts "npath=$npath"
     set comp [split $npath /]
-    if {[catch {exec rm [join $comp /]}]} return
+    if {[catch {exec rm $npath}]} return
 
     set l [llength $comp]
-    incr l -1
-    incr l -1
-    incr status($area) -1
+    incr l -2
+    incr status($task,$area) -1
     for {set i $l} {$i > 0} {incr i -1} {
         set path [join [lrange $comp 0 $i] /]
 	if {![catch {glob $path/*}]} return
-        exec rmdir ./$path
+        exec rmdir $path
     }
     # puts "RobotFileUnlink end"
 }
@@ -140,12 +134,14 @@ proc RobotFileClose {out} {
     }
 }
 
-proc RobotFileOpen {area host path {mode w}} {
+proc RobotFileOpen {task area host path {mode w}} {
     set orgPwd [pwd]
     global workdir
     global status
     global debuglevel
 
+    # puts "RobotFileOpen task=$task path=$path"
+
     if {![info exists workdir]} {
 	return stdout
     }
@@ -158,127 +154,291 @@ proc RobotFileOpen {area host path {mode w}} {
 	puts "pwd = $orgPwd"
 	exit 1
     }
-    set comp [split $area/$host$path /]
+
+    set comp [split $task/$area/$host /]
     set len [llength $comp]
     incr len -1
-    for {set i 0} {$i < $len} {incr i} {
-        if {$i > 1} {
-            set d "d[lindex $comp $i]" 
-        } else {
-            set d [lindex $comp $i]
-        }
-        if {[catch {cd ./$d}]} {
+
+    # puts "1 comp=$comp"
+
+    for {set i 0} {$i <= $len} {incr i} {
+        set d [lindex $comp $i]
+	if {[string length $d] == 0} {
+	    cd /
+	} elseif {[catch {cd $d}]} {
             exec mkdir $d
             cd ./$d
-	    if {![string compare $area unvisited] && $i == 1 && $mode == "w"} {
-		set out [open frobots.txt w]
-		puts "creating robots.txt in $d"
-		close $out
-                incr status(unvisited)
+	    if {![string compare $area unvisited] && $i == $len && $mode == "w"} {
+		if {[string compare $path /robots.txt]} {
+		    set out [open robots.txt_.tkl w]
+		    puts "creating robots.txt in $d"
+		    close $out
+                    incr status($task,unvisited)
+		}
 	    }
         }
     }
-    set d [lindex $comp $len]
-    if {[string length $d]} {
-        set out [open f$d $mode]
-        if {0} {
-            if {[file isfile $d/f]} {
-                set out [open $d/f $mode]
-            } else {
-                set out [open f$d $mode]
+
+    set comp [split $path /]
+    set len [llength $comp]
+    incr len -1
+
+    # puts "2 path=$path comp=$comp"
+
+    for {set i 0} {$i < $len} {incr i} {
+        set d [lindex $comp $i]
+        if {[string length $d] > 0} {
+            if {[catch {cd $d}]} {
+                exec mkdir $d
+                cd ./$d
             }
         }
-    } else {
-        set out [open f $mode]
     }
+    set d [lindex $comp $len]
+    set out [open ${d}_.tkl $mode]
     if {$mode == "w"} {
-        incr status($area)
+        incr status($task,$area)
     }
     cd $orgPwd
     return $out
 }
 
-proc RobotRR {} {
-    global robotSeq robotsRunning
+proc RobotStartJob {fname t} {
+    global control
+
+    set f [open $fname r]
+    set xml [read $f]
+    puts "Reading $fname"
+    close $f
+    if {![regexp {<status>([^<]*)</status>} $xml x status]} {
+	return
+    }
+    if {$status == "done"} {
+        puts "already done"
+        return
+    }
+    puts "status = $status"
+    if {![task $t]} {
+        return
+    }
+    htmlSwitch $xml \
+        url {
+            url $body
+        } filter {
+            set type $parm(type)
+            set action $parm(action)
+            if {$type == "domain"} {
+                $action url http://$body/*
+            }
+            if {$type == "url"} {
+                $action url $body
+            }
+            if {$type == "mime"} {
+                $action mime $body
+            }
+        } distance {
+            set control($t,distance) $body
+        } status {
+            set control($t,filestatus) $body
+        }
+    if {$status == "pending"} {
+        regsub {<status>[^<]*</status>} $xml {<status>running</status>} xml2
+        set f [open $fname w]
+        puts -nonewline $f $xml2 
+        close $f
+    }
+}
+
+proc RobotDoneJob {t} {
+    global daemon_dir
 
+    if {![info exists daemon_dir]} {
+        return
+    }
+
+    set fname $t.tkl
+
+    set f [open $fname r]
+    set xml [read $f]
+    puts "Reading $fname"
+    regexp {<status>([^<]*)</status>} $xml x status
+    puts "------"
+    puts "status = $status"
+    close $f
+
+    regsub {<status>[^<]*</status>} $xml {<status>done</status>} xml2
+    set f [open $fname w]
+    puts -nonewline $f $xml2 
+    close $f
+}
+
+proc RobotScanDir {} {
+    global daemon_dir
+
+    if {![info exists daemon_dir]} {
+        return
+    }
+    foreach d $daemon_dir {
+        if {[catch {set files [glob $d/*.tkl]}]} {
+            return
+        }
+        foreach fname $files {
+            if {[file isfile $fname] && [file readable $fname]} {
+                set t [file rootname $fname]
+                RobotStartJob $fname $t
+            }
+        }
+    }
+}
+
+proc RobotRR {task} {
+    global control robotsRunning tasks robotsMax status
+
+    puts "RobotRR -- running=$robotsRunning max=$robotsMax---------------"
     incr robotsRunning -1
+
+    # only one task gets through...
+    if {[string compare [lindex $tasks 0] $task]} {
+        return
+    }
+    puts "RobotRR. task = $task"
     while {$robotsRunning} {
 	vwait robotsRunning
     }
-    set robotSeq 0
-    RobotStart
+    puts "Scan"
+    if {[catch {RobotScanDir} msg]} {
+        puts "RobotScanDir failed"
+        puts $msg
+    }
+    foreach t $tasks {
+        set statusfile [open $t/status w]
+        puts $statusfile "$status($t,unvisited) $status($t,bad) $status($t,visited)"
+        close $statusfile
+        set control($t,seq) 0
+        RobotStart $t
+    }
+}
+
+proc RobotDaemonSig {} {
+    global daemon_cnt
+
+    incr daemon_cnt
+}
+
+proc RobotDaemonLoop {} {
+    global daemon_cnt tasks robotsRunning status
+
+    set daemon_cnt 0
+    while 1 {
+        puts $daemon_cnt
+        
+        RobotScanDir
+        
+        if {[info exists tasks]} {
+            puts "daemon loop tasks $tasks"
+            foreach t $tasks {
+                set control($t,seq) 0
+                RobotStart $t
+            }
+            while {$robotsRunning} {
+                vwait robotsRunning
+            }
+        }
+        after 30000 RobotDaemonSig
+        vwait daemon_cnt
+    }
 }
 
-proc RobotRestart {url sock} {
+proc RobotRestart {task url sock} {
     global URL robotsRunning
 
     close $sock
     after cancel $URL($sock,cancel) 
 
-    foreach v [array names URL $url,*] {
+    foreach v [array names URL $task,$url,*] {
 	unset URL($v)
     }
 
     incr robotsRunning -1
-    RobotStart
+    RobotStart $task
 }
 
-proc RobotStart {} {
+proc RobotStart {task} {
     global URL
-    global robotsRunning robotsMax idletime
+    global robotsRunning robotsMax idletime status tasks
   
-    # puts "RobotStart"
+    # puts "RobotStart $task running=$robotsRunning"
     while {1} {
-        set url [RobotFileNext unvisited]
+        set url [RobotFileNext $task unvisited]
+	if {[string compare $url done] == 0} {
+            puts "In RobotStart task $task done"
+
+            catch {unset ntasks}
+            foreach t $tasks {
+                if {[string compare $t $task]} {
+                    lappend ntasks $t
+                } else {
+                    puts "task $t done"
+                }
+            }
+            if {![info exists ntasks]} {
+                unset tasks
+                puts "all done"
+            } else {
+                set tasks $ntasks
+            }
+            RobotDoneJob $task
+	    return
+	}
         if {![string length $url]} {
 	    return
 	}
-	incr robotsRunning
+        incr robotsRunning
 	if {[string compare $url wait] == 0} {
-	    after $idletime RobotRR
-	    return
+            after $idletime [list RobotRR $task]
+            return
 	}
-        set r [RobotGetUrl $url {}]
+        set r [RobotGetUrl $task $url {}]
         if {!$r} {
 	    if {$robotsRunning >= $robotsMax} return
         } else {
 	    incr robotsRunning -1
-	    if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
-		set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
+	    if {![RobotFileExist $task bad $URL($task,$url,hostport) $URL($task,$url,path)]} {
+		set outf [RobotFileOpen $task bad $URL($task,$url,hostport) $URL($task,$url,path)]
 		RobotFileClose $outf
 	    }
-            RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
+            RobotFileUnlink $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)
 	}
     }
 }
 
-proc headSave {url out} {
+proc headSave {task url out} {
     global URL
     
-    if {[info exists URL($url,head,last-modified)]} {
-        puts $out "<lastmodified>$URL($url,head,last-modified)</lastmodified>"
+    if {[info exists URL($task,$url,head,last-modified)]} {
+        puts $out "<lastmodified>$URL($task,$url,head,last-modified)</lastmodified>"
     }
     puts $out {<si>}
-    if {[info exists URL($url,head,date)]} {
-        puts $out " <date>$URL($url,head,date)</date>"
+    if {[info exists URL($task,$url,head,date)]} {
+        puts $out " <date>$URL($task,$url,head,date)</date>"
     }
-    if {[info exists URL($url,head,content-length)]} {
-        puts $out " <by>$URL($url,head,content-length)</by>"
+    if {[info exists URL($task,$url,head,content-length)]} {
+        puts $out " <by>$URL($task,$url,head,content-length)</by>"
     }
-    if {[info exists URL($url,head,server)]} {
-        puts $out " <format>$URL($url,head,server)</format>"
+    if {[info exists URL($task,$url,head,server)]} {
+        puts $out " <format>$URL($task,$url,head,server)</format>"
     }
     puts $out {</si>}
     puts $out {<publisher>}
     puts $out " <identifier>$url</identifier>"
-    if {[info exists URL($url,head,content-type)]} {
-        puts $out " <type>$URL($url,head,content-type)</type>"
+    if {[info exists URL($task,$url,head,content-type)]} {
+        puts $out " <type>$URL($task,$url,head,content-type)</type>"
     }
     puts $out {</publisher>}
 }
 
-proc RobotHref {url hrefx hostx pathx} {
-    global URL domains debuglevel
+proc RobotHref {task url hrefx hostx pathx} {
+    global URL control debuglevel
     upvar $hrefx href
     upvar $hostx host
     upvar $pathx path
@@ -293,12 +453,11 @@ proc RobotHref {url hrefx hostx pathx} {
     if {[string length $href] > 256} {
 	return 0
     }
-    if {[string first {?} $href] >= 0} {
-	return 0
-    }
-    if {[string first {?} $url] >= 0 && [string first {?} $href] >= 0} {
-	return 0
-    }
+
+#   Skip pages that have ? in them
+#    if {[string first {?} $url] >= 0 && [string first {?} $href] >= 0} {
+#	return 0
+#    }
     # get method (if any)
     if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
 	set hpath $href
@@ -313,13 +472,13 @@ proc RobotHref {url hrefx hostx pathx} {
 	if {![string length $surl]} {
 	    set surl /
 	}
-        if {[info exist domains]} {
+        if {[info exist control($task,domains)]} {
 	    set ok 0
-	    foreach domain $domains {
+	    foreach domain $control($task,domains) {
 	        if {[string match $domain $host]} {
 		    set ok 1
 		    break
-	         }
+                }
 	    }
 	    if {!$ok} {
 	        return 0
@@ -327,16 +486,16 @@ proc RobotHref {url hrefx hostx pathx} {
         }
     } else {
 	regexp {^([^\#]*)} $hpath x surl
-	set host $URL($url,hostport)
+	set host $URL($task,$url,hostport)
     }
     if {![string length $surl]} {
 	return 0
     }
     if {[string first / $surl]} {
 	# relative path
-        set curpath $URL($url,path)
-        if {[info exists URL($url,bpath)]} {
-            set curpath $URL($url,bpath)
+        set curpath $URL($task,$url,path)
+        if {[info exists URL($task,$url,bpath)]} {
+            set curpath $URL($task,$url,bpath)
         }
 	regexp {^([^\#?]*)} $curpath x dpart
 	set l [string last / $dpart]
@@ -380,55 +539,55 @@ proc RobotHref {url hrefx hostx pathx} {
     if {$debuglevel > 1} {
         puts "Ref result = $href"
     }
-    return [checkrule url $href]
+    return [checkrule $task url $href]
 }
 
-proc RobotError {url code} {
+proc RobotError {task url code} {
     global URL
 
     puts "Bad URL $url (code $code)"
     set fromurl {}
     set distance -1
-    if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
-	set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
+    if {[RobotFileExist $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)]} {
+	set inf [RobotFileOpen $task unvisited $URL($task,$url,hostport) $URL($task,$url,path) r]
 	RobotReadRecord $inf fromurl distance
 	RobotFileClose $inf
     }
-    RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
-    if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
-	set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
+    RobotFileUnlink $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)
+    if {![RobotFileExist $task bad $URL($task,$url,hostport) $URL($task,$url,path)]} {
+	set outf [RobotFileOpen $task bad $URL($task,$url,hostport) $URL($task,$url,path)]
 	RobotWriteRecord $outf $fromurl $distance
 	RobotFileClose $outf
     }
 }
 
-proc RobotRedirect {url tourl code} {
+proc RobotRedirect {task url tourl code} {
     global URL
 
     puts "Redirecting from $url to $tourl"
 
     set distance {}
     set fromurl {}
-    if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
-	set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
+    if {[RobotFileExist $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)]} {
+	set inf [RobotFileOpen $task unvisited $URL($task,$url,hostport) $URL($task,$url,path) r]
 	RobotReadRecord $inf fromurl distance
 	RobotFileClose $inf
     }
-    if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
-	set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
+    if {![RobotFileExist $task bad $URL($task,$url,hostport) $URL($task,$url,path)]} {
+	set outf [RobotFileOpen $task bad $URL($task,$url,hostport) $URL($task,$url,path)]
 	RobotWriteRecord $outf $fromurl $distance
 	RobotFileClose $outf
     }
-    if {[RobotHref $url tourl host path]} {
-	if {![RobotFileExist visited $host $path]} {
-	    if {![RobotFileExist unvisited $host $path]} {
-		set outf [RobotFileOpen unvisited $host $path]
+    if {[RobotHref $task $url tourl host path]} {
+	if {![RobotFileExist $task visited $host $path]} {
+	    if {![RobotFileExist $task unvisited $host $path]} {
+		set outf [RobotFileOpen $task unvisited $host $path]
 		RobotWriteRecord $outf $fromurl $distance
 		RobotFileClose $outf
 	    }
 	} else {
 	    set olddistance {}
-	    set inf [RobotFileOpen visited $host $path r]
+	    set inf [RobotFileOpen $task visited $host $path r]
 	    RobotReadRecord $inf oldurl olddistance
 	    RobotFileClose $inf
 	    if {[string length $olddistance] == 0} {
@@ -439,34 +598,44 @@ proc RobotRedirect {url tourl code} {
 	    }
 	    puts "distance=$distance olddistance=$olddistance"
 	    if {[expr $distance < $olddistance]} {
-		set outf [RobotFileOpen unvisited $host $path]
+		set outf [RobotFileOpen $task unvisited $host $path]
 		RobotWriteRecord $outf $tourl $distance
 		RobotFileClose $outf
 	    }
 	}
     }
-    if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
+    if {[catch {RobotFileUnlink $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)}]} {
         puts "unlink failed"
         exit 1
     }
 }
 
-proc link {url out href body distance} {
-    global URL maxdistance
-    if {[expr $distance > $maxdistance]} return
+proc wellform {body} {
+    regsub -all {<!--[^-]*-->} $body { } abody
+    regsub -all -nocase {<script[^<]*</script>} $abody {} body
+    regsub -all {<[^\>]+>} $body {} abody
+    regsub -all {&nbsp;} $abody { } body
+    regsub -all {&} $body {&amp;} abody
+    return $abody
+}
+
+proc link {task url out href body distance} {
+    global URL control
+    if {[expr $distance > $control($task,distance)]} return
     
-    if {![RobotHref $url href host path]} return
+    if {![RobotHref $task $url href host path]} return
     
     puts $out "<cr>"
     puts $out "<identifier>$href</identifier>"
-    puts $out "<description>$body</description>"
+    set abody [wellform $body]
+    puts $out "<description>$abody</description>"
     puts $out "</cr>"
     
-    if {![RobotFileExist visited $host $path]} {
+    if {![RobotFileExist $task visited $host $path]} {
         set olddistance 1000
-        if {![RobotFileExist bad $host $path]} {
-            if {[RobotFileExist unvisited $host $path]} {
-                set inf [RobotFileOpen unvisited $host $path r]
+        if {![RobotFileExist $task bad $host $path]} {
+            if {[RobotFileExist $task unvisited $host $path]} {
+                set inf [RobotFileOpen $task unvisited $host $path r]
                 RobotReadRecord $inf oldurl olddistance
                 RobotFileClose $inf
             }
@@ -477,12 +646,12 @@ proc link {url out href body distance} {
             set olddistance 1000
         }
         if {[expr $distance < $olddistance]} {
-            set outf [RobotFileOpen unvisited $host $path]
+            set outf [RobotFileOpen $task unvisited $host $path]
             RobotWriteRecord $outf $url $distance
             RobotFileClose $outf
         }
     } elseif {[string compare $href $url]} {
-        set inf [RobotFileOpen visited $host $path r]
+        set inf [RobotFileOpen $task visited $host $path r]
         RobotReadRecord $inf xurl olddistance
         close $inf
         if {[string length $olddistance] == 0} {
@@ -492,77 +661,116 @@ proc link {url out href body distance} {
             puts "OK remarking url=$url href=$href"
             puts "olddistance = $olddistance"
             puts "newdistance = $distance"
-            set outf [RobotFileOpen unvisited $host $path]
+            set outf [RobotFileOpen $task unvisited $host $path]
             RobotWriteRecord $outf $url $distance
             RobotFileClose $outf
         }
     }
 }
 
-proc RobotTextHtml {url out} {
-    global URL maxdistance
+proc RobotTextHtml {task url out} {
+    global URL control
+
+    # set title so we can emit it for the body
+    set title {}
+    # if true, nothing will be indexed
+    set noindex 0
+    # if true, nothing will be followed
+    set nofollow 0
 
     set distance 0
     set fdistance 0
-    if {$maxdistance < 1000 && [info exists URL($url,dist)]} {
-        set fdistance $URL($url,dist)
+    if {$control($task,distance) < 1000 && [info exists URL($task,$url,dist)]} {
+        set fdistance $URL($task,$url,dist)
 	set distance [expr $fdistance + 1]
     }
-    htmlSwitch $URL($url,buf) \
+    htmlSwitch $URL($task,$url,buf) \
         title {
-	    puts $out "<title>$body</title>"
+            set title $body
         } -nonest meta {
+            # collect metadata and save NAME= CONTENT=..
+            set metaname {}
+            set metacontent {}
             puts -nonewline $out "<meta"
-            foreach a [array names parm] {
-		puts -nonewline $out " $a"
+	    set al [array names parm]
+            foreach a $al {
+                set al [string tolower $a]
+		puts -nonewline $out " $al"
                 puts -nonewline $out {="}
                 puts -nonewline $out $parm($a)
                 puts -nonewline $out {"}
+                switch -- $al {
+                    "name" {
+                        set metaname [string tolower $parm($a)]
+                    }
+                    "content" {
+                        set metacontent $parm($a)
+                    }
+                }
+	 	unset parm($al)
+            }
+	    puts $out "></meta>"
+            # go through robots directives (af any)
+            if {![string compare $metaname robots]} {
+                set direcs [split [string tolower $metacontent] ,]
+                if {[lsearch $direcs noindex] >= 0} {
+                    set noindex 1
+                }
+                if {[lsearch $direcs nofollow] >= 0} {
+                    set nofollow 1
+                }
             }
-	    puts $out {></meta>}
 	} body {
-	    regsub -all {<!--[^-]*->} $body { } abody
-	    regsub -all -nocase {<script[^<]*</script>} $abody {} bbody
-	    regsub -all {<[^\>]+>} $bbody {} nbody
-	    puts $out "<documentcontent>"
-            puts $out $nbody
-            puts $out "</documentcontent>"
+            # don't print title of document content if noindex is used
+            if {!$noindex} {
+                puts $out "<title>$title</title>"
+		set bbody [wellform $body]
+                puts $out "<documentcontent>"
+                puts $out $bbody
+                puts $out "</documentcontent>"
+            }
         } -nonest base {
+            # <base href=.. >
             if {![info exists parm(href)]} {
 	        continue
             }
             set href [string trim $parm(href)]
-            if {![RobotHref $url href host path]} continue
-            set URL($url,bpath) $path
-        } -nonest a {
+            if {![RobotHref $task $url href host path]} continue
+            set URL($task,$url,bpath) $path
+        } a {
+            # <a href="...."> .. </a> 
+            # we're not using nonest - otherwise body isn't set
+            if {$nofollow} continue
             if {![info exists parm(href)]} {
 	        continue
             }
-            link $url $out [string trim $parm(href)] $body $distance
+            link $task $url $out [string trim $parm(href)] $body $distance
         } -nonest area {
+            if {$nofollow} continue
             if {![info exists parm(href)]} {
 	        continue
             }
-            link $url $out [string trim $parm(href)] $body $distance
+            link $task $url $out [string trim $parm(href)] $body $distance
         } -nonest frame {
             if {![info exists parm(src)]} {
 	        continue
             }
-            link $url $out [string trim $parm(src)] $body $fdistance
+            link $task $url $out [string trim $parm(src)] $body $fdistance
 	}
 }
 
-proc RobotsTxt {url} {
+proc RobotsTxt {task url} {
     global agent URL
 
-    RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
+    RobotsTxt0 $task URL(URL($task,$url,hostport),robots) $URL($task,$url,buf)
 }
 
-proc RobotsTxt0 {v buf} {
+proc RobotsTxt0 {task v buf} {
     global URL agent
     set section 0
     foreach l [split $buf \n] {
 	if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
+            set arg [string trim $arg]
 	    puts "cmd=$cmd arg=$arg"
 	    switch -- [string tolower $cmd] {
 		user-agent {
@@ -587,186 +795,193 @@ proc RobotsTxt0 {v buf} {
     }
 }
 
-proc RobotTextPlain {url out} {
+proc RobotTextPlain {task url out} {
     global URL
 
     puts $out "<documentcontent>"
-    regsub -all {<} $URL($url,buf) {\&lt;} content
+    regsub -all {<} $URL($task,$url,buf) {\&lt;} content
     puts $out $content
     puts $out "</documentcontent>"
 
-    if {![string compare $URL($url,path) /robots.txt]} {
-	RobotsTxt $url
+    if {![string compare $URL($task,$url,path) /robots.txt]} {
+	RobotsTxt $task $url
     }
 }
 
-proc RobotWriteMetadata {url out} {
-    global URL domains
+proc RobotWriteMetadata {task url out} {
+    global URL
 
+    set charset $URL($task,$url,charset)
+    puts $out "<?xml version=\"1.0\" encoding=\"$charset\" standalone=\"yes\"?>"
     puts $out "<zmbot>"
 
     set distance 1000
-    if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
-	set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
+    if {[RobotFileExist $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)]} {
+	set inf [RobotFileOpen $task unvisited $URL($task,$url,hostport) $URL($task,$url,path) r]
 	RobotReadRecord $inf fromurl distance
 	RobotFileClose $inf
     }
-    set URL($url,dist) $distance
+    set URL($task,$url,dist) $distance
     puts $out "<distance>"
     puts $out "  $distance"
     puts $out "</distance>"
-    headSave $url $out
+    headSave $task $url $out
     puts "Parsing $url distance=$distance"
-    switch $URL($url,head,content-type) {
+    switch $URL($task,$url,head,content-type) {
         text/html {
             if {[string length $distance]} {
-                RobotTextHtml $url $out
+                RobotTextHtml $task $url $out
             }
         }
         text/plain {
-            RobotTextPlain $url $out
+            RobotTextPlain $task $url $out
         }
     }
     puts $out "</zmbot>"
 }
 
-proc Robot200 {url} {
-    global URL domains
+proc Robot200 {task url} {
+    global URL
     
-    set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
-    puts -nonewline $out $URL($url,buf)
+    set out [RobotFileOpen $task raw $URL($task,$url,hostport) $URL($task,$url,path)]
+    puts -nonewline $out $URL($task,$url,buf)
     RobotFileClose $out
 
-    set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
-    RobotWriteMetadata $url $out
+    set out [RobotFileOpen $task visited $URL($task,$url,hostport) $URL($task,$url,path)]
+    RobotWriteMetadata $task $url $out
     RobotFileClose $out
 
-    RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
+    RobotFileUnlink $task unvisited $URL($task,$url,hostport) $URL($task,$url,path)
 }
 
-proc RobotReadContent {url sock binary} {
+proc RobotReadContent {task url sock binary} {
     global URL
 
     set buffer [read $sock 16384]
     set readCount [string length $buffer]
 
     if {$readCount <= 0} {
-	Robot200 $url
-	RobotRestart $url $sock
+	Robot200 $task $url
+	RobotRestart $task $url $sock
     } elseif {!$binary && [string first \0 $buffer] >= 0} {
-	Robot200 $url
-	RobotRestart $url $sock
+	Robot200 $task $url
+	RobotRestart $task $url $sock
     } else {
 	# puts "Got $readCount bytes"
-	set URL($url,buf) $URL($url,buf)$buffer
+	set URL($task,$url,buf) $URL($task,$url,buf)$buffer
     }
 }
 
-proc RobotReadHeader {url sock} {
+proc RobotReadHeader {task url sock} {
     global URL debuglevel
 
     if {$debuglevel > 1} {
         puts "HTTP head $url"
     }
     if {[catch {set buffer [read $sock 2148]}]} {
-	RobotError $url 404
-	RobotRestart $url $sock
+	RobotError $task $url 404
+	RobotRestart $task $url $sock
         return
     }
     set readCount [string length $buffer]
     
     if {$readCount <= 0} {
-	RobotError $url 404
-	RobotRestart $url $sock
+	RobotError $task $url 404
+	RobotRestart $task $url $sock
     } else {
 	# puts "Got $readCount bytes"
-	set URL($url,buf) $URL($url,buf)$buffer
+	set URL($task,$url,buf) $URL($task,$url,buf)$buffer
 	
-	set n [string first \r\n\r\n $URL($url,buf)]
+	set n [string first \r\n\r\n $URL($task,$url,buf)]
 	if {$n > 1} {
 	    set code 0
 	    set version {}
-	    set headbuf [string range $URL($url,buf) 0 $n]
+	    set headbuf [string range $URL($task,$url,buf) 0 $n]
 	    incr n 4
-	    set URL($url,buf) [string range $URL($url,buf) $n end]
+	    set URL($task,$url,charset) ISO-8859-1
+	    set URL($task,$url,buf) [string range $URL($task,$url,buf) $n end]
 	    
 	    regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
 	    set lines [split $headbuf \n]
 	    foreach line $lines {
 		if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
-		    set URL($url,head,[string tolower $name]) [string trim $value]
+		    set URL($task,$url,head,[string tolower $name]) [string trim $value]
 		}
+		regexp {^Content-Type:.*charset=([A-Za-z0-9_-]*)} $line x URL($task,$url,charset)
 	    }
 	    puts "HTTP CODE $code"
-	    set URL($url,state) skip
+	    set URL($task,$url,state) skip
 	    switch $code {
 		301 {
-		    RobotRedirect $url $URL($url,head,location) 301
-		    RobotRestart $url $sock
+		    RobotRedirect $task $url $URL($task,$url,head,location) 301
+		    RobotRestart $task $url $sock
 		}
 		302 {
-		    RobotRedirect $url $URL($url,head,location) 302
-		    RobotRestart $url $sock
+		    RobotRedirect $task $url $URL($task,$url,head,location) 302
+		    RobotRestart $task $url $sock
 		}
 		200 {
-		    if {![info exists URL($url,head,content-type)]} {
-			set URL($url,head,content-type) {}
+		    if {![info exists URL($task,$url,head,content-type)]} {
+			set URL($task,$url,head,content-type) {}
 		    }
 		    set binary 1
-		    switch -glob -- $URL($url,head,content-type) {
+		    switch -glob -- $URL($task,$url,head,content-type) {
 			text/* {
 			    set binary 0
 			}
 		    }
                     if {![regexp {/robots.txt$} $url]} {
-                        if {![checkrule mime $URL($url,head,content-type)]} {
-                            RobotError $url mimedeny
-                            RobotRestart $url $sock
+                        if {![checkrule $task mime $URL($task,$url,head,content-type)]} {
+                            RobotError $task $url mimedeny
+                            RobotRestart $task $url $sock
                             return
                         }
                     }
-		    fileevent $sock readable [list RobotReadContent $url $sock $binary]
+		    fileevent $sock readable [list RobotReadContent $task $url $sock $binary]
 		}
 		default {
-		    RobotError $url $code
-		    RobotRestart $url $sock
+		    RobotError $task $url $code
+		    RobotRestart $task $url $sock
 		}
 	    }
 	}
     }
 }
 
-proc RobotSockCancel {url sock} {
+proc RobotSockCancel {task url sock} {
 
     puts "RobotSockCancel sock=$sock url=$url"
-    RobotError $url 401
-    RobotRestart $url $sock
+    RobotError $task $url 401
+    RobotRestart $task $url $sock
 }
 
-proc RobotConnect {url sock} {
+proc RobotConnect {task url sock} {
     global URL agent acceptLanguage
 
     fconfigure $sock -translation {lf crlf} -blocking 0
-    fileevent $sock readable [list RobotReadHeader $url $sock]
-    puts $sock "GET $URL($url,path) HTTP/1.0"
-    puts $sock "Host: $URL($url,host)"
+    fileevent $sock readable [list RobotReadHeader $task $url $sock]
+    puts $sock "GET $URL($task,$url,path) HTTP/1.0"
+    puts $sock "Host: $URL($task,$url,host)"
     puts $sock "User-Agent: $agent"
     if {[string length $acceptLanguage]} {
         puts $sock "Accept-Language: $acceptLanguage"
     }
     puts $sock ""
-    flush $sock
-    set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
+    set URL($sock,cancel) [after 30000 [list RobotSockCancel $task $url $sock]]
+    if {[catch {flush $sock}]} {
+        RobotError $task $url 404
+	RobotRestart $task $url $sock
+    }
 }
 
 proc RobotNop {} {
 
 }
 
-proc RobotGetUrl {url phost} {
+proc RobotGetUrl {task url phost} {
     global URL robotsRunning
     flush stdout
-    puts "Retrieve $robotsRunning url=$url"
+    puts "Retrieve running=$robotsRunning url=$url task=$task"
     if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
         return -1
     }
@@ -774,25 +989,25 @@ proc RobotGetUrl {url phost} {
 	set port 80
 	set host $hostport
     }
-    set URL($url,method) $method
-    set URL($url,host) $host
-    set URL($url,hostport) $hostport
-    set URL($url,path) $path
-    set URL($url,state) head
-    set URL($url,buf) {}
+    set URL($task,$url,method) $method
+    set URL($task,$url,host) $host
+    set URL($task,$url,hostport) $hostport
+    set URL($task,$url,path) $path
+    set URL($task,$url,state) head
+    set URL($task,$url,buf) {}
 
     if {[string compare $path /robots.txt]} {
 	set ok 1
 	if {![info exists URL($hostport,robots)]} {
 	    puts "READING robots.txt for host $hostport"
-	    if {[RobotFileExist visited $hostport /robots.txt]} {
-		set inf [RobotFileOpen visited $hostport /robots.txt r]
+	    if {[RobotFileExist $task visited $hostport /robots.txt]} {
+		set inf [RobotFileOpen $task visited $hostport /robots.txt r]
 		set buf [read $inf 32768]
 		close $inf
 	    } else {
 		set buf "User-agent: *\nAllow: /\n"
 	    }
-	    RobotsTxt0 URL($hostport,robots) $buf
+	    RobotsTxt0 $task URL($hostport,robots) $buf
 	}
 	if {[info exists URL($hostport,robots)]} {
 	    foreach l $URL($hostport,robots) {
@@ -810,7 +1025,7 @@ proc RobotGetUrl {url phost} {
     if [catch {set sock [socket -async $host $port]}] {
         return -1
     }
-    RobotConnect $url $sock
+    RobotConnect $task $url $sock
 
     return 0
 }
@@ -822,7 +1037,7 @@ if {![llength [info commands htmlSwitch]]} {
     }
 }
 
-set agent "zmbot/0.1"
+set agent "zmbot/0.2"
 if {![catch {set os [exec uname -s -r]}]} {
     set agent "$agent ($os)"
 }
@@ -836,28 +1051,24 @@ proc bgerror {m} {
 }
 
 set robotsRunning 0
-set robotSeq 0
 set workdir [pwd]
-set idletime 60000
+set idletime 30000
 set acceptLanguage {}
 set debuglevel 0
-set status(unvisited) 0
-set status(visited) 0
-set status(bad) 0
-set status(raw) 0
-
 
 # Rules: allow, deny, url
 
-proc checkrule {type this} {
-    global alrules
+proc checkrule {task type this} {
+    global control
     global debuglevel
 
+    set default_ret 1
+
     if {$debuglevel > 3} {
         puts "CHECKRULE $type $this"
     }
-    if {[info exist alrules]} {
-        foreach l $alrules {
+    if {[info exist control($task,alrules)]} {
+        foreach l $control($task,alrules) {
             if {$debuglevel > 3} {
                 puts "consider $l"
             }
@@ -866,6 +1077,7 @@ proc checkrule {type this} {
             # consider mask (! negates)
             set masks [lindex $l 2]
 	    set ok 0
+	    set default_ret 0
 	    foreach mask $masks {	
                 if {$debuglevel > 4} {
                     puts "consider single mask $mask"
@@ -897,18 +1109,18 @@ proc checkrule {type this} {
         }
     }
     if {$debuglevel > 3} {
-        puts "CHECKRULE MATCH OK"
+        puts "CHECKRULE MATCH DEFAULT $default_ret"
     }
-    return 1
+    return $default_ret
 }
 
 
 proc url {href} {
-    global debuglevel
+    global debuglevel task
 
-    if {[RobotHref http://www.indexdata.dk/ href host path]} {
-        if {![RobotFileExist visited $host $path]} {
-            set outf [RobotFileOpen unvisited $host $path]
+    if {[RobotHref $task http://www.indexdata.dk/ href host path]} {
+        if {![RobotFileExist $task visited $host $path]} {
+            set outf [RobotFileOpen $task unvisited $host $path]
             RobotWriteRecord $outf href 0
             RobotFileClose $outf
         }
@@ -916,15 +1128,15 @@ proc url {href} {
 }
 
 proc deny {type stuff} {
-    global alrules
+    global control task
 
-    lappend alrules [list deny $type $stuff]
+    lappend control($task,alrules) [list deny $type $stuff]
 }
 
 proc allow {type stuff} {
-    global alrules
+    global control task
 
-    lappend alrules [list allow $type $stuff]
+    lappend control($task,alrules) [list allow $type $stuff]
 }
 
 proc debug {level} {
@@ -933,6 +1145,37 @@ proc debug {level} {
     set debuglevel $level
 }
 
+proc task {t} {
+    global tasks task status control
+
+    set task $t
+
+    if {[info exists tasks]} {
+        if {[lsearch -exact $tasks $t] >= 0} {
+            return 0
+        }
+    }
+
+    lappend tasks $t
+    set status($t,unvisited) 0
+    set status($t,visited) 0
+    set status($t,bad) 0
+    set status($t,raw) 0
+    set status($t,active) 1
+    set control($t,seq) 0
+    set control($t,distance) 10
+    return 1
+}
+
+# Little utility that ensures that at least one task is present (main).
+proc chktask {} {
+    global tasks
+    if {![info exist tasks]} {
+        task main
+    }
+}
+
+
 # Parse options
 
 set i 0
@@ -940,14 +1183,30 @@ set l [llength $argv]
 
 if {$l < 2} {
     puts {tclrobot: usage:}
-    puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-r rules] [url ..]}
+    puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-D dir] [-r rules] [url ..]}
     puts " Example: -c 3 -d '*.dk' http://www.indexdata.dk/"
 
     exit 1
 }
+
+
 while  {$i < $l} {
     set arg [lindex $argv $i]
     switch -glob -- $arg {
+        -t* {
+	    set t [string range $arg 2 end]
+	    if {![string length $t]} {
+		set t [lindex $argv [incr i]]
+	    }
+            task $t
+	}
+        -D* {
+	    set dir [string range $arg 2 end]
+	    if {![string length $dir]} {
+		set dir [lindex $argv [incr i]]
+	    }
+            lappend daemon_dir $dir
+        }
 	-j* {
 	    set robotsMax [string range $arg 2 end]
 	    if {![string length $robotsMax]} {
@@ -955,17 +1214,19 @@ while  {$i < $l} {
 	    }
 	}
 	-c* {
-	    set maxdistance [string range $arg 2 end]
-	    if {![string length $maxdistance]} {
-		set maxdistance [lindex $argv [incr i]]
+            chktask
+	    set control($task,distance) [string range $arg 2 end]
+	    if {![string length $control($task,distance)]} {
+		set control($task,distance) [lindex $argv [incr i]]
 	    }
 	}
 	-d* {
+            chktask
 	    set dom [string range $arg 2 end]
 	    if {![string length $dom]} {
 		set dom [lindex $argv [incr i]]
 	    }
-	    lappend domains $dom
+	    lappend control($task,domains) $dom
 	}
 	-i* {
 	    set idletime [string range $arg 2 end]
@@ -974,23 +1235,30 @@ while  {$i < $l} {
 	    }
 	}
         -l* {
+            chktask
 	    set acceptLanguage [string range $arg 2 end]
 	    if {![string length $acceptLanguage]} {
 		set acceptLanguage [lindex $argv [incr i]]
  	    }
 	}
         -r* {
+            chktask
 	    set rfile [string range $arg 2 end]
 	    if {![string length $rfile]} {
 		set rfile [lindex $argv [incr i]]
  	    }
+            catch {unset maxdistance}
             source $rfile
+            if {[info exists maxdistance]} {
+                set control($task,distance) $maxdistance
+            }
         }
 	default {
+            chktask
 	    set href $arg
-	    if {[RobotHref http://www.indexdata.dk/ href host path]} {
-		if {![RobotFileExist visited $host $path]} {
-		    set outf [RobotFileOpen unvisited $host $path]
+	    if {[RobotHref $task http://www.indexdata.dk/ href host path]} {
+		if {![RobotFileExist $task visited $host $path]} {
+		    set outf [RobotFileOpen $task unvisited $host $path]
 		    RobotWriteRecord $outf href 0
 		    RobotFileClose $outf
 		}
@@ -1000,29 +1268,36 @@ while  {$i < $l} {
     incr i
 }
 
-if {![info exist domains]} {
-    set domains {*}
-}
-if {![info exist maxdistance]} {
-    set maxdistance 50
-}
 if {![info exist robotsMax]} {
     set robotsMax 5
 }
 
-puts "domains=$domains"
-puts "max distance=$maxdistance"
-puts "max jobs=$robotsMax"
-
-
-RobotStart
-
-
-while {$robotsRunning} {
-    vwait robotsRunning
+if {[info exist daemon_dir]} {
+    RobotDaemonLoop
+} else {
+    foreach t $tasks {
+	puts "task $t"
+	puts "max distance=$control($t,distance)"
+	if {[info exists control($t,domains)]} {
+	    puts "domains=$control($t,domains)"
+	}
+    }
+    puts "max jobs=$robotsMax"
+    
+    foreach t $tasks {
+	RobotStart $t
+    }
+    
+    while {$robotsRunning} {
+	vwait robotsRunning
+    }
+    
+    if {[info exists tasks]} {
+	foreach t $tasks {
+	    set statusfile [open $t/status w]
+	    puts $statusfile "$status($t,unvisited) $status($t,bad) $status($t,visited)"
+	    close $statusfile
+	}
+    }
 }
 
-set statusfile [open status w]
-puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
-close $statusfile
-