X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=b6466c347186d9954fbdda74c78ed7054fab76c5;hb=9d3f82cd1140362487d8fa6372cac1b24a49d21e;hp=82e5c28ce8ee9ba634c85235473b450adaa44eec;hpb=de694c0a797c955ec564b91fee7d0293ce9bd03c;p=tclrobot.git

diff --git a/robot.tcl b/robot.tcl
index 82e5c28..b6466c3 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,8 +1,8 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.9 2000/12/11 17:11:03 adam Exp $
+# $Id: robot.tcl,v 1.29 2001/11/14 09:15:23 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
-    puts "RobotFileNext1 area=$area lead=$lead"
+    # puts "RobotFileNext1 area=$area lead=$lead"
     if {[catch {set ns [glob ${area}/*]}]} {
         return {}
     }
@@ -26,35 +26,53 @@ proc RobotFileNext1 {area lead} {
     return {}
 }
 
-proc RobotFileWait {} {
-    global robotSeq
-    set robotSeq 0
+proc RobotWriteRecord {outf fromurl distance} {
+    puts $outf "<zmbot>"
+    puts $outf "<distance>"
+    puts $outf $distance
+    puts $outf "</distance>"
+    puts $outf "<fromurl>"
+    puts $outf $fromurl
+    puts $outf "</fromurl>"
+    puts $outf "</zmbot>"
+}
+
+proc RobotReadRecord {inf fromurlx distancex} {
+    upvar $fromurlx fromurl
+    upvar $distancex distance
+    gets $inf
+    gets $inf
+    set distance [string trim [gets $inf]]
+    # puts "got distance = $distance"
+    gets $inf
+    gets $inf
+    set fromurl [string trim [gets $inf]]
 }
 
 proc RobotFileNext {area} {
     global robotSeq
-    puts "RobotFileNext robotSeq=$robotSeq"
-    if {[catch {set ns [glob ${area}/*]}]} {
-        return {}
+    global idletime ns
+    global status
+
+    # puts "RobotFileNext robotSeq=$robotSeq"
+    if {$robotSeq < 0} {
+	return {}
+    }
+    if {$robotSeq == 0} {
+	if {[catch {set ns [glob ${area}/*]}]} {
+	    return {}
+	}
     }
     set off [string length $area]
     incr off
-
     set n [lindex $ns $robotSeq]
     if {![string length $n]} {
-	flush stdout
-	puts "------------ N E X T  R O U N D --------"
 	set robotSeq -1
-	after 60000 RobotFileWait
-	vwait robotSeq
-
-	set n [lindex $ns $robotSeq]
-	if {![string length $n]} {
-	    puts "robotSeq = $robotSeq"
-	    puts "ns=$ns"
-	    puts "no more work at index"
-	    return {}
-	}
+	flush stdout
+        set statusfile [open status w]
+        puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+        close $statusfile
+	return wait
     }
     incr robotSeq
     if {[file isfile $n/frobots.txt]} {
@@ -73,40 +91,47 @@ proc RobotFileNext {area} {
 
 
 proc RobotFileExist {area host path} {
-    puts "RobotFileExist begin"
-    puts "area=$area host=$host path=$path"
+    global debuglevel
+
+    if {$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
-    puts "npath=$npath"
-    puts "RobotFileExist end"
+    if {$debuglevel > 3} {
+        puts "RobotFileExist end npath=$npath"
+    }
     return [file exists $npath]
 }
 
 proc RobotFileUnlink {area host path} {
-    puts "RobotFileUnlink begin"
-    puts "area=$area host=$host path=$path"
+    global status
+    # puts "RobotFileUnlink begin"
+    # puts "area=$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
-    puts "npath=$npath"
+    # puts "npath=$npath"
     set comp [split $npath /]
+    if {[catch {exec rm [join $comp /]}]} return
+
     set l [llength $comp]
     incr l -1
-    if {[catch {exec rm [join $comp /]}]} return
     incr l -1
+    incr status($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
     }
-    puts "RobotFileUnlink end"
+    # puts "RobotFileUnlink end"
 }
 
 proc RobotFileClose {out} {
@@ -118,11 +143,15 @@ proc RobotFileClose {out} {
 proc RobotFileOpen {area host path {mode w}} {
     set orgPwd [pwd]
     global workdir
+    global status
+    global debuglevel
 
     if {![info exists workdir]} {
 	return stdout
     }
-    puts "RobotFileOpen orgPwd=$orgPwd area=$area host=$host path=$path mode=$mode"
+    if {$debuglevel > 3} {
+        puts "RobotFileOpen orgPwd=$orgPwd area=$area host=$host path=$path mode=$mode"
+    }
     if {[string compare $orgPwd $workdir]} {
         puts "ooops. RobotFileOpen failed"
 	puts "workdir = $workdir"
@@ -140,52 +169,87 @@ proc RobotFileOpen {area host path {mode w}} {
         }
         if {[catch {cd ./$d}]} {
             exec mkdir $d
-            puts "creating $d"
             cd ./$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)
 	    }
         }
     }
     set d [lindex $comp $len]
     if {[string length $d]} {
-	if {[file isdirectory $d]} {
-	    set out [open $d/f $mode]
-           puts "1"
-	} else {
-	    set out [open f$d $mode]
-           puts "2"
-	}
+        set out [open f$d $mode]
+        if {0} {
+            if {[file isfile $d/f]} {
+                set out [open $d/f $mode]
+            } else {
+                set out [open f$d $mode]
+            }
+        }
     } else {
         set out [open f $mode]
-        puts "3"
+    }
+    if {$mode == "w"} {
+        incr status($area)
     }
     cd $orgPwd
-    #puts "RobotFileStop"
     return $out
 }
 
-proc RobotRestart {sock} {
-    global URL
-    global robotMoreWork
-  
+proc RobotRR {} {
+    global robotSeq robotsRunning
+
+    incr robotsRunning -1
+    while {$robotsRunning} {
+	vwait robotsRunning
+    }
+    set robotSeq 0
+    RobotStart
+}
+
+proc RobotRestart {url sock} {
+    global URL robotsRunning
+
     close $sock
     after cancel $URL($sock,cancel) 
-    while {1} {    
+
+    foreach v [array names URL $url,*] {
+	unset URL($v)
+    }
+
+    incr robotsRunning -1
+    RobotStart
+}
+
+proc RobotStart {} {
+    global URL
+    global robotsRunning robotsMax idletime
+  
+    # puts "RobotStart"
+    while {1} {
         set url [RobotFileNext unvisited]
         if {![string length $url]} {
-	    break
+	    return
+	}
+	incr robotsRunning
+	if {[string compare $url wait] == 0} {
+	    after $idletime RobotRR
+	    return
 	}
         set r [RobotGetUrl $url {}]
         if {!$r} {
-	    return
+	    if {$robotsRunning >= $robotsMax} return
         } else {
-            RobotFileUnlink unvisited $URL($url,host) $URL($url,path)
-        }
+	    incr robotsRunning -1
+	    if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
+		set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
+		RobotFileClose $outf
+	    }
+            RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
+	}
     }
-    incr robotMoreWork -1
 }
 
 proc headSave {url out} {
@@ -214,12 +278,27 @@ proc headSave {url out} {
 }
 
 proc RobotHref {url hrefx hostx pathx} {
-    global URL domains
+    global URL domains debuglevel
     upvar $hrefx href
     upvar $hostx host
     upvar $pathx path
 
-    puts "Ref url = $url href=$href"
+    if {$debuglevel > 1} {
+        puts "Ref input url = $url href=$href"
+    }
+
+    if {[string first { } $href] >= 0} {
+	return 0
+    }
+    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
+    }
     # get method (if any)
     if {![regexp {^([^/:]+):(.*)} $href x method hpath]} {
 	set hpath $href
@@ -234,26 +313,32 @@ proc RobotHref {url hrefx hostx pathx} {
 	if {![string length $surl]} {
 	    set surl /
 	}
-	set ok 0
-	foreach domain $domains {
-	    if {[string match $domain $host]} {
-		set ok 1
-		break
+        if {[info exist domains]} {
+	    set ok 0
+	    foreach domain $domains {
+	        if {[string match $domain $host]} {
+		    set ok 1
+		    break
+	         }
 	    }
-	}
-	if {!$ok} {
-	    return 0
-	}
+	    if {!$ok} {
+	        return 0
+	    }
+        }
     } else {
 	regexp {^([^\#]*)} $hpath x surl
-	set host $URL($url,host)
+	set host $URL($url,hostport)
     }
     if {![string length $surl]} {
 	return 0
     }
     if {[string first / $surl]} {
 	# relative path
-	regexp {^([^\#?]*)} $URL($url,path) x dpart
+        set curpath $URL($url,path)
+        if {[info exists URL($url,bpath)]} {
+            set curpath $URL($url,bpath)
+        }
+	regexp {^([^\#?]*)} $curpath x dpart
 	set l [string last / $dpart]
 	if {[expr $l >= 0]} {
 	    set surl [string range $dpart 0 $l]$surl
@@ -261,58 +346,58 @@ proc RobotHref {url hrefx hostx pathx} {
 	    set surl $dpart/$surl
 	}
     }
-    set c [split $surl /]
-    set i [llength $c]
-    incr i -1
-    set path [lindex $c $i]
-    incr i -1
-    while {$i >= 0} {
-	switch -- [lindex $c $i] {
+    set surllist [split $surl /]
+    catch {unset path}
+    set pathl 0
+    foreach c $surllist {
+        switch -- $c {
 	    .. {
-		incr i -2
-		if {$i < 0} {
-		    set i 0
+ 		if {$pathl > 1} {
+		    incr pathl -2
+		    set path [lrange $path 0 $pathl]
+		    incr pathl
 		}
 	    }
-	    . {
-		incr i -1
-	    }
-	    default {
-		set path [lindex $c $i]/$path
-		incr i -1
+            . {
+
+            }
+            default {
+		incr pathl
+                lappend path $c
 	    }
 	}
     }
-    regsub -all {~} $path {%7E} path
-    set ok 1
-    if {[info exists URL($host,robots)]} {
-	foreach l $URL($host,robots) {
-	    if {[string first [lindex $l 1] $path] == 0} {
-		set ok [lindex $l 0]
-		break
-	    }
-	}
+    if {$debuglevel > 4} {
+        puts "pathl=$pathl output path=$path"
+    }
+    set path [join $path /]
+    if {![string length $path]} {
+	set path /
     }
+    regsub -all {~} $path {%7E} path
     set href "$method://$host$path"
-    puts "Ref href = $href, ok=$ok"
-    return $ok
+
+    if {$debuglevel > 1} {
+        puts "Ref result = $href"
+    }
+    return [checkrule url $href]
 }
 
 proc RobotError {url code} {
     global URL
 
-    puts "Bad URL $url, $code"
+    puts "Bad URL $url (code $code)"
     set fromurl {}
-    if {[RobotFileExist unvisited $URL($url,host) $URL($url,path)]} {
-	set inf [RobotFileOpen unvisited $URL($url,host) $URL($url,path) r]
-	set fromurl [gets $inf]
-	close $inf
-    }
-    RobotFileUnlink unvisited $URL($url,host) $URL($url,path)
-    if {![RobotFileExist bad $URL($url,host) $URL($url,path)]} {
-	set outf [RobotFileOpen bad $URL($url,host) $URL($url,path)]
-	puts $outf "URL=$url $code"
-	puts $outf "Reference $fromurl"
+    set distance -1
+    if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
+	set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($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)]
+	RobotWriteRecord $outf $fromurl $distance
 	RobotFileClose $outf
     }
 }
@@ -322,35 +407,107 @@ proc RobotRedirect {url tourl code} {
 
     puts "Redirecting from $url to $tourl"
 
+    set distance {}
     set fromurl {}
-    if {[RobotFileExist unvisited $URL($url,host) $URL($url,path)]} {
-	set inf [RobotFileOpen unvisited $URL($url,host) $URL($url,path) r]
-	set fromurl [gets $inf]
+    if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
+	set inf [RobotFileOpen unvisited $URL($url,hostport) $URL($url,path) r]
+	RobotReadRecord $inf fromurl distance
 	RobotFileClose $inf
     }
-    if {[catch {RobotFileUnlink unvisited $URL($url,host) $URL($url,path)}]} {
-        puts "unlink failed"
-        exit 1
-    }
-    if {![RobotFileExist bad $URL($url,host) $URL($url,path)]} {
-	set outf [RobotFileOpen bad $URL($url,host) $URL($url,path)]
-	puts $outf "URL=$url to $tourl $code"
-	puts $outf "Reference $fromurl"
+    if {![RobotFileExist bad $URL($url,hostport) $URL($url,path)]} {
+	set outf [RobotFileOpen bad $URL($url,hostport) $URL($url,path)]
+	RobotWriteRecord $outf $fromurl $distance
 	RobotFileClose $outf
     }
     if {[RobotHref $url tourl host path]} {
-	if {![RobotFileExist unvisited $host $path]} {
-		puts "Mark as unvisited"
-	    set outf [RobotFileOpen unvisited $host $path]
-	    puts $outf $code
-	    RobotFileClose $outf
+	if {![RobotFileExist visited $host $path]} {
+	    if {![RobotFileExist unvisited $host $path]} {
+		set outf [RobotFileOpen unvisited $host $path]
+		RobotWriteRecord $outf $fromurl $distance
+		RobotFileClose $outf
+	    }
+	} else {
+	    set olddistance {}
+	    set inf [RobotFileOpen visited $host $path r]
+	    RobotReadRecord $inf oldurl olddistance
+	    RobotFileClose $inf
+	    if {[string length $olddistance] == 0} {
+		set olddistance 1000
+	    }
+	    if {[string length $distance] == 0} {
+		set distance 1000
+	    }
+	    puts "distance=$distance olddistance=$olddistance"
+	    if {[expr $distance < $olddistance]} {
+		set outf [RobotFileOpen unvisited $host $path]
+		RobotWriteRecord $outf $tourl $distance
+		RobotFileClose $outf
+	    }
 	}
     }
+    if {[catch {RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)}]} {
+        puts "unlink failed"
+        exit 1
+    }
+}
+
+proc link {url out href body distance} {
+    global URL maxdistance
+    if {[expr $distance > $maxdistance]} return
+    
+    if {![RobotHref $url href host path]} return
+    
+    puts $out "<cr>"
+    puts $out "<identifier>$href</identifier>"
+    puts $out "<description>$body</description>"
+    puts $out "</cr>"
+    
+    if {![RobotFileExist visited $host $path]} {
+        set olddistance 1000
+        if {![RobotFileExist bad $host $path]} {
+            if {[RobotFileExist unvisited $host $path]} {
+                set inf [RobotFileOpen unvisited $host $path r]
+                RobotReadRecord $inf oldurl olddistance
+                RobotFileClose $inf
+            }
+        } else {
+            set olddistance 0
+        }
+        if {[string length $olddistance] == 0} {
+            set olddistance 1000
+        }
+        if {[expr $distance < $olddistance]} {
+            set outf [RobotFileOpen unvisited $host $path]
+            RobotWriteRecord $outf $url $distance
+            RobotFileClose $outf
+        }
+    } elseif {[string compare $href $url]} {
+        set inf [RobotFileOpen visited $host $path r]
+        RobotReadRecord $inf xurl olddistance
+        close $inf
+        if {[string length $olddistance] == 0} {
+            set olddistance 1000
+        }
+        if {[expr $distance < $olddistance]} {
+            puts "OK remarking url=$url href=$href"
+            puts "olddistance = $olddistance"
+            puts "newdistance = $distance"
+            set outf [RobotFileOpen unvisited $host $path]
+            RobotWriteRecord $outf $url $distance
+            RobotFileClose $outf
+        }
+    }
 }
 
 proc RobotTextHtml {url out} {
-    global URL
+    global URL maxdistance
 
+    set distance 0
+    set fdistance 0
+    if {$maxdistance < 1000 && [info exists URL($url,dist)]} {
+        set fdistance $URL($url,dist)
+	set distance [expr $fdistance + 1]
+    }
     htmlSwitch $URL($url,buf) \
         title {
 	    puts $out "<title>$body</title>"
@@ -364,61 +521,62 @@ proc RobotTextHtml {url out} {
             }
 	    puts $out {></meta>}
 	} body {
-	    regsub -all -nocase {<script.*</script>} $body {} abody
-	    regsub -all {<[^\>]+>} $abody {} nbody
+	    regsub -all {<!--[^-]*->} $body { } abody
+	    regsub -all -nocase {<script[^<]*</script>} $abody {} bbody
+	    regsub -all {<[^\>]+>} $bbody {} nbody
 	    puts $out "<documentcontent>"
             puts $out $nbody
             puts $out "</documentcontent>"
-        } a {
+        } -nonest base {
             if {![info exists parm(href)]} {
-	        puts "no href"
 	        continue
             }
-	    if {1} {
-		set href $parm(href)
-		if {![RobotHref $url href host path]} continue
-		
-		puts $out "<cr>"
-		puts $out "<identifier>$href</identifier>"
-		puts $out "<description>$body</description>"
-		puts $out "</cr>"
-		
-		if {![RobotFileExist visited $host $path]} {
-		    if {![RobotFileExist bad $host $path]} {
-		        if {[catch {set outf [RobotFileOpen unvisited $host $path]} msg]} {
-			    puts "--- Error $msg"
-			    exit 1
-		        }
-			puts $outf $url
-		        RobotFileClose $outf
-		    }
-		}
-	    }
+            set href [string trim $parm(href)]
+            if {![RobotHref $url href host path]} continue
+            set URL($url,bpath) $path
+        } -nonest a {
+            if {![info exists parm(href)]} {
+	        continue
+            }
+            link $url $out [string trim $parm(href)] $body $distance
+        } -nonest area {
+            if {![info exists parm(href)]} {
+	        continue
+            }
+            link $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
 	}
 }
 
 proc RobotsTxt {url} {
     global agent URL
 
-    set v URL($URL($url,host),robots)
+    RobotsTxt0 URL(URL($url,hostport),robots) $URL($url,buf)
+}
+
+proc RobotsTxt0 {v buf} {
+    global URL agent
     set section 0
-    foreach l [split $URL($url,buf) \n] {
-	puts $l
-	if {[regexp {([-A-Za-z]+):[ \t]*([^\#]+)} $l match cmd arg]} {
+    foreach l [split $buf \n] {
+	if {[regexp {([-A-Za-z]+):[ ]*([^\# ]+)} $l match cmd arg]} {
 	    puts "cmd=$cmd arg=$arg"
-	    switch $cmd {
-		User-Agent {
+	    switch -- [string tolower $cmd] {
+		user-agent {
 		    if {$section} break
 		    set pat [string tolower $arg]*
 		    set section [string match $pat $agent]
 		}
-		Disallow {
+		disallow {
 		    if {$section} {
 			puts "rule [list 0 $arg]"
 			lappend $v [list 0 $arg]
 		    }
 		}
-		Allow {
+		allow {
 		    if {$section} {
 			puts "rule [list 1 $arg]"
 			lappend $v [list 1 $arg]
@@ -433,7 +591,8 @@ proc RobotTextPlain {url out} {
     global URL
 
     puts $out "<documentcontent>"
-    puts $out $URL($url,buf)
+    regsub -all {<} $URL($url,buf) {\&lt;} content
+    puts $out $content
     puts $out "</documentcontent>"
 
     if {![string compare $URL($url,path) /robots.txt]} {
@@ -441,30 +600,48 @@ proc RobotTextPlain {url out} {
     }
 }
 
-proc Robot200 {url} {
+proc RobotWriteMetadata {url out} {
     global URL domains
-    
-    puts "Parsing $url"
-    set out [RobotFileOpen visited $URL($url,host) $URL($url,path)]
+
     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]
+	RobotReadRecord $inf fromurl distance
+	RobotFileClose $inf
+    }
+    set URL($url,dist) $distance
+    puts $out "<distance>"
+    puts $out "  $distance"
+    puts $out "</distance>"
     headSave $url $out
+    puts "Parsing $url distance=$distance"
     switch $URL($url,head,content-type) {
-	text/html {
-	    RobotTextHtml $url $out
-	}
-	text/plain {
-	    RobotTextPlain $url $out
-	}
-	application/pdf {
-	    set pdff [open test.pdf w]
-	    puts -nonewline $pdff $URL($url,buf)
-	    close $pdff
-	}
+        text/html {
+            if {[string length $distance]} {
+                RobotTextHtml $url $out
+            }
+        }
+        text/plain {
+            RobotTextPlain $url $out
+        }
     }
     puts $out "</zmbot>"
+}
+
+proc Robot200 {url} {
+    global URL domains
+    
+    set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
+    puts -nonewline $out $URL($url,buf)
     RobotFileClose $out
-    # puts "Parsing done"
-    RobotFileUnlink unvisited $URL($url,host) $URL($url,path)
+
+    set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
+    RobotWriteMetadata $url $out
+    RobotFileClose $out
+
+    RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
 }
 
 proc RobotReadContent {url sock binary} {
@@ -475,10 +652,10 @@ proc RobotReadContent {url sock binary} {
 
     if {$readCount <= 0} {
 	Robot200 $url
-	RobotRestart $sock
+	RobotRestart $url $sock
     } elseif {!$binary && [string first \0 $buffer] >= 0} {
 	Robot200 $url
-	RobotRestart $sock
+	RobotRestart $url $sock
     } else {
 	# puts "Got $readCount bytes"
 	set URL($url,buf) $URL($url,buf)$buffer
@@ -486,18 +663,21 @@ proc RobotReadContent {url sock binary} {
 }
 
 proc RobotReadHeader {url sock} {
-    global URL
+    global URL debuglevel
 
-    puts "RobotReadHeader $url"
+    if {$debuglevel > 1} {
+        puts "HTTP head $url"
+    }
     if {[catch {set buffer [read $sock 2148]}]} {
 	RobotError $url 404
-	RobotRestart $sock
+	RobotRestart $url $sock
+        return
     }
     set readCount [string length $buffer]
     
     if {$readCount <= 0} {
 	RobotError $url 404
-	RobotRestart $sock
+	RobotRestart $url $sock
     } else {
 	# puts "Got $readCount bytes"
 	set URL($url,buf) $URL($url,buf)$buffer
@@ -513,60 +693,70 @@ proc RobotReadHeader {url sock} {
 	    regexp {^HTTP/([0-9.]+)[ ]+([0-9]+)} $headbuf x version code
 	    set lines [split $headbuf \n]
 	    foreach line $lines {
-		if {[regexp {^([^:]+):[ ]+(.*)} $line x name value]} {
+		if {[regexp {^([^:]+):[ ]+([^;]*)} $line x name value]} {
 		    set URL($url,head,[string tolower $name]) [string trim $value]
 		}
 	    }
-	    puts "code = $code"
+	    puts "HTTP CODE $code"
 	    set URL($url,state) skip
 	    switch $code {
 		301 {
 		    RobotRedirect $url $URL($url,head,location) 301
-		    RobotRestart $sock
+		    RobotRestart $url $sock
 		}
 		302 {
 		    RobotRedirect $url $URL($url,head,location) 302
-		    RobotRestart $sock
+		    RobotRestart $url $sock
 		}
 		200 {
 		    if {![info exists URL($url,head,content-type)]} {
 			set URL($url,head,content-type) {}
 		    }
-		    set binary 0
-		    switch $URL($url,head,content-type) {
-			application/pdf {
-			    set binary 1
+		    set binary 1
+		    switch -glob -- $URL($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
+                            return
+                        }
+                    }
 		    fileevent $sock readable [list RobotReadContent $url $sock $binary]
 		}
 		default {
 		    RobotError $url $code
-		    RobotRestart $sock
+		    RobotRestart $url $sock
 		}
 	    }
 	}
     }
 }
 
-proc RobotSockCancel {sock url} {
+proc RobotSockCancel {url sock} {
 
     puts "RobotSockCancel sock=$sock url=$url"
     RobotError $url 401
-    RobotRestart $sock
+    RobotRestart $url $sock
 }
 
 proc RobotConnect {url sock} {
-    global URL agent
+    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)"
     puts $sock "User-Agent: $agent"
+    if {[string length $acceptLanguage]} {
+        puts $sock "Accept-Language: $acceptLanguage"
+    }
     puts $sock ""
     flush $sock
-    set URL($sock,cancel) [after 60000 [list RobotSockCancel $sock $url]]
+    set URL($sock,cancel) [after 30000 [list RobotSockCancel $url $sock]]
 }
 
 proc RobotNop {} {
@@ -574,11 +764,10 @@ proc RobotNop {} {
 }
 
 proc RobotGetUrl {url phost} {
-    global URL
+    global URL robotsRunning
     flush stdout
-    puts "---------"
-    puts $url
-    if {![regexp {([^:]+)://([^/]+)([^ ]*)} $url x method hostport path]} {
+    puts "Retrieve $robotsRunning url=$url"
+    if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
         return -1
     }
     if {![regexp {([^:]+):([0-9]+)} $hostport x host port]} {
@@ -587,10 +776,37 @@ proc RobotGetUrl {url phost} {
     }
     set URL($url,method) $method
     set URL($url,host) $host
-    set URL($url,port) $port
+    set URL($url,hostport) $hostport
     set URL($url,path) $path
     set URL($url,state) head
     set URL($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]
+		set buf [read $inf 32768]
+		close $inf
+	    } else {
+		set buf "User-agent: *\nAllow: /\n"
+	    }
+	    RobotsTxt0 URL($hostport,robots) $buf
+	}
+	if {[info exists URL($hostport,robots)]} {
+	    foreach l $URL($hostport,robots) {
+		if {[string first [lindex $l 1] $path] == 0} {
+		    set ok [lindex $l 0]
+		    break
+		}
+	    }
+	}
+	if {!$ok} {
+	    puts "skipped due to robots.txt"
+	    return -1
+	}
+    }
     if [catch {set sock [socket -async $host $port]}] {
         return -1
     }
@@ -606,37 +822,207 @@ if {![llength [info commands htmlSwitch]]} {
     }
 }
 
-set agent "zmbot/0.0"
+set agent "zmbot/0.1"
 if {![catch {set os [exec uname -s -r]}]} {
     set agent "$agent ($os)"
-	puts "agent: $agent"
 }
 
+puts "agent: $agent"
+
 proc bgerror {m} {
     global errorInfo
     puts "BGERROR $m"
     puts $errorInfo
 }
 
-set robotMoreWork 0
+set robotsRunning 0
 set robotSeq 0
 set workdir [pwd]
+set idletime 60000
+set acceptLanguage {}
+set debuglevel 0
+set status(unvisited) 0
+set status(visited) 0
+set status(bad) 0
+set status(raw) 0
 
-if {[llength $argv] < 2} {
-    puts "Tclrobot: usage <domain> <start>"
-    puts " Example: '*.indexdata.dk' http://www.indexdata.dk/"
-    exit 1
+
+# Rules: allow, deny, url
+
+proc checkrule {type this} {
+    global alrules
+    global debuglevel
+
+    if {$debuglevel > 3} {
+        puts "CHECKRULE $type $this"
+    }
+    if {[info exist alrules]} {
+        foreach l $alrules {
+            if {$debuglevel > 3} {
+                puts "consider $l"
+            }
+            # consider type
+            if {[lindex $l 1] != $type} continue
+            # consider mask (! negates)
+            set masks [lindex $l 2]
+	    set ok 0
+	    foreach mask $masks {	
+                if {$debuglevel > 4} {
+                    puts "consider single mask $mask"
+                }
+                if {[string index $mask 0] == "!"} {
+                    set mask [string range $mask 1 end]
+                    if {[string match $mask $this]}  continue
+                } else {
+                    if {![string match $mask $this]} continue
+                }
+                set ok 1
+            }
+            if {$debuglevel > 4} {
+                puts "ok = $ok"
+            }
+            if {!$ok} continue
+            # OK, we have a match
+            if {[lindex $l 0] == "allow"} {
+                if {$debuglevel > 3} {
+                    puts "CHECKRULE MATCH OK"
+                }
+                return 1
+            } else {
+                if {$debuglevel > 3} {
+                    puts "CHECKFULE MATCH FAIL"
+                }
+                return 0
+            }
+        }
+    }
+    if {$debuglevel > 3} {
+        puts "CHECKRULE MATCH OK"
+    }
+    return 1
+}
+
+
+proc url {href} {
+    global debuglevel
+
+    if {[RobotHref http://www.indexdata.dk/ href host path]} {
+        if {![RobotFileExist visited $host $path]} {
+            set outf [RobotFileOpen unvisited $host $path]
+            RobotWriteRecord $outf href 0
+            RobotFileClose $outf
+        }
+    }
+}
+
+proc deny {type stuff} {
+    global alrules
+
+    lappend alrules [list deny $type $stuff]
+}
+
+proc allow {type stuff} {
+    global alrules
+
+    lappend alrules [list allow $type $stuff]
 }
 
-set domains [lindex $argv 0]
-foreach site [lindex $argv 1] {
-    incr robotMoreWork
-    if [RobotGetUrl $site {}] {
-	incr robotMoreWork -1
-	puts "Couldn't process $site"
+proc debug {level} {
+    global debuglevel
+
+    set debuglevel $level
+}
+
+# Parse options
+
+set i 0
+set l [llength $argv]
+
+if {$l < 2} {
+    puts {tclrobot: usage:}
+    puts {tclrobot [-j jobs] [-i idle] [-c count] [-d domain] [-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 {
+	-j* {
+	    set robotsMax [string range $arg 2 end]
+	    if {![string length $robotsMax]} {
+		set robotsMax [lindex $argv [incr i]]
+	    }
+	}
+	-c* {
+	    set maxdistance [string range $arg 2 end]
+	    if {![string length $maxdistance]} {
+		set maxdistance [lindex $argv [incr i]]
+	    }
+	}
+	-d* {
+	    set dom [string range $arg 2 end]
+	    if {![string length $dom]} {
+		set dom [lindex $argv [incr i]]
+	    }
+	    lappend domains $dom
+	}
+	-i* {
+	    set idletime [string range $arg 2 end]
+	    if {![string length $idletime]} {
+		set idletime [lindex $argv [incr i]]
+	    }
+	}
+        -l* {
+	    set acceptLanguage [string range $arg 2 end]
+	    if {![string length $acceptLanguage]} {
+		set acceptLanguage [lindex $argv [incr i]]
+ 	    }
+	}
+        -r* {
+	    set rfile [string range $arg 2 end]
+	    if {![string length $rfile]} {
+		set rfile [lindex $argv [incr i]]
+ 	    }
+            source $rfile
+        }
+	default {
+	    set href $arg
+	    if {[RobotHref http://www.indexdata.dk/ href host path]} {
+		if {![RobotFileExist visited $host $path]} {
+		    set outf [RobotFileOpen unvisited $host $path]
+		    RobotWriteRecord $outf href 0
+		    RobotFileClose $outf
+		}
+	    }
+	}
     }
+    incr i
 }
 
-while {$robotMoreWork} {
-    vwait robotMoreWork
+if {![info exist domains]} {
+    set domains {*}
+}
+if {![info exist maxdistance]} {
+    set maxdistance 50
 }
+if {![info exist robotsMax]} {
+    set robotsMax 5
+}
+
+puts "domains=$domains"
+puts "max distance=$maxdistance"
+puts "max jobs=$robotsMax"
+
+
+RobotStart
+
+
+while {$robotsRunning} {
+    vwait robotsRunning
+}
+
+set statusfile [open status w]
+puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+close $statusfile
+