X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=2e175d623cb41f9c62a06a7b0d318a29dca11b45;hb=833faf12d797d629cae34abc8e84e88a6044eb7f;hp=b8db6c45d17b4e570a217921833f1c5a49b931b7;hpb=9b7fbb227055fc0ad4eeb10d482fd9e8ada7ce4e;p=tclrobot.git

diff --git a/robot.tcl b/robot.tcl
index b8db6c4..2e175d6 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,5 +1,5 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.18 2001/06/07 08:17:00 adam Exp $
+# $Id: robot.tcl,v 1.25 2001/11/07 11:50:07 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
     # puts "RobotFileNext1 area=$area lead=$lead"
@@ -50,7 +50,7 @@ proc RobotReadRecord {inf fromurlx distancex} {
 }
 
 proc RobotFileNext {area} {
-    global robotSeq global idleTime ns
+    global robotSeq global idletime ns
 
     # puts "RobotFileNext robotSeq=$robotSeq"
     if {$robotSeq < 0} {
@@ -67,7 +67,7 @@ proc RobotFileNext {area} {
     if {![string length $n]} {
 	set robotSeq -1
 	flush stdout
-	puts "------------ N E X T  R O U N D --------"
+	puts "Round robin"
 	return wait
     }
     incr robotSeq
@@ -201,7 +201,7 @@ proc RobotRestart {url sock} {
 
 proc RobotStart {} {
     global URL
-    global robotsRunning robotsMax idleTime
+    global robotsRunning robotsMax idletime
   
     # puts "RobotStart"
     while {1} {
@@ -211,7 +211,7 @@ proc RobotStart {} {
 	}
 	incr robotsRunning
 	if {[string compare $url wait] == 0} {
-	    after $idleTime RobotRR
+	    after $idletime RobotRR
 	    return
 	}
         set r [RobotGetUrl $url {}]
@@ -254,12 +254,14 @@ 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
@@ -267,6 +269,9 @@ 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
     }
@@ -320,8 +325,9 @@ proc RobotHref {url hrefx hostx pathx} {
         switch -- $c {
 	    .. {
  		if {$pathl > 0} {
-		    incr pathl -1
+		    incr pathl -2
 		    set path [lrange $path 0 $pathl]
+		    incr pathl
 		}
 	    }
             . {
@@ -340,14 +346,17 @@ proc RobotHref {url hrefx hostx pathx} {
     }
     regsub -all {~} $path {%7E} path
     set href "$method://$host$path"
-    puts "Ref href = $href"
-    return 1
+
+    if {$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 {}
     set distance -1
     if {[RobotFileExist unvisited $URL($url,hostport) $URL($url,path)]} {
@@ -413,10 +422,10 @@ proc RobotRedirect {url tourl code} {
 }
 
 proc RobotTextHtml {url out} {
-    global URL maxDistance
+    global URL maxdistance
 
     set distance 0
-    if {$maxDistance < 1000 && [info exists URL($url,dist)]} {
+    if {$maxdistance < 1000 && [info exists URL($url,dist)]} {
 	set distance [expr $URL($url,dist) + 1]
     }
     htmlSwitch $URL($url,buf) \
@@ -432,8 +441,9 @@ 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>"
@@ -442,7 +452,7 @@ proc RobotTextHtml {url out} {
 	        puts "no href"
 	        continue
             }
-	    if {[expr $distance <= $maxDistance]} {
+	    if {[expr $distance <= $maxdistance]} {
 		set href [string trim $parm(href)]
 		if {![RobotHref $url href host path]} continue
 		
@@ -492,7 +502,7 @@ proc RobotTextHtml {url out} {
 	        puts "no href"
 	        continue
             }
-	    if {[expr $distance <= $maxDistance]} {
+	    if {[expr $distance <= $maxdistance]} {
 		set href [string trim $parm(href)]
 		if {![RobotHref $url href host path]} continue
 		
@@ -588,14 +598,9 @@ proc RobotTextPlain {url out} {
     }
 }
 
-proc Robot200 {url} {
+proc RobotWriteMetadata {url out} {
     global URL domains
-    
-    set out [RobotFileOpen raw $URL($url,hostport) $URL($url,path)]
-    puts -nonewline $out $URL($url,buf)
-    RobotFileClose $out
 
-    set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
     puts $out "<zmbot>"
 
     set distance 1000
@@ -611,30 +616,44 @@ proc Robot200 {url} {
     headSave $url $out
     puts "Parsing $url distance=$distance"
     switch $URL($url,head,content-type) {
-	text/html {
-	    if {[string length $distance]} {
-		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
+        }
+        application/pdf {
+            set pdff [open test.pdf w]
+            puts -nonewline $pdff $URL($url,buf)
+            close $pdff
+        }
     }
     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
+
+    if {![checkrule mime $URL($url,head,content-type)]} {
+        RobotError $url mimedeny
+        return
+    }
+    set out [RobotFileOpen visited $URL($url,hostport) $URL($url,path)]
+    RobotWriteMetadata $url $out
     RobotFileClose $out
-    # puts "Parsing done"
+
     RobotFileUnlink unvisited $URL($url,hostport) $URL($url,path)
 }
 
 proc RobotReadContent {url sock binary} {
     global URL
 
-    puts "RobotReadContent $url"
     set buffer [read $sock 16384]
     set readCount [string length $buffer]
 
@@ -651,9 +670,11 @@ 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 $url $sock
@@ -682,7 +703,7 @@ proc RobotReadHeader {url sock} {
 		    set URL($url,head,[string tolower $name]) [string trim $value]
 		}
 	    }
-	    puts "code = $code"
+	    puts "HTTP CODE $code"
 	    set URL($url,state) skip
 	    switch $code {
 		301 {
@@ -722,13 +743,16 @@ proc RobotSockCancel {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 30000 [list RobotSockCancel $url $sock]]
@@ -741,7 +765,7 @@ proc RobotNop {} {
 proc RobotGetUrl {url phost} {
     global URL robotsRunning
     flush stdout
-    puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
+    puts "Retrieve $robotsRunning url=$url"
     if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
         return -1
     }
@@ -797,7 +821,7 @@ 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)"
 }
@@ -813,28 +837,108 @@ proc bgerror {m} {
 set robotsRunning 0
 set robotSeq 0
 set workdir [pwd]
-set idleTime 60000
+set idletime 60000
+set acceptLanguage {}
 
 set i 0
 set l [llength $argv]
 
-# For testing only
-if {0} {
-    set url "http://www.sportsfiskeren.dk/sportsfiskeren/corner/index.htm"
-    set href "../../data/../../data2/newsovs.asp?Mode=5"
-
-    set URL($url,path) /sportsfiskeren/corner/index.htm
-    set URL($url,hostport) www.sportsfiskeren.dk
-    RobotHref $url href host path
-    exit 0
-}
-
 if {$l < 2} {
-    puts {tclrobot: usage [-j jobs] [-i idle] [-c count] [-d domain] [url ..]}
+    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
 }
 
+# Rules: allow, deny, url
+set debuglevel 0
+
+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]
+}
+
+proc debug {level} {
+    global debuglevel
+
+    set debuglevel $level
+}
+
+# Parse options
+
 while  {$i < $l} {
     set arg [lindex $argv $i]
     switch -glob -- $arg {
@@ -845,9 +949,9 @@ while  {$i < $l} {
 	    }
 	}
 	-c* {
-	    set maxDistance [string range $arg 2 end]
-	    if {![string length $maxDistance]} {
-		set maxDistance [lindex $argv [incr i]]
+	    set maxdistance [string range $arg 2 end]
+	    if {![string length $maxdistance]} {
+		set maxdistance [lindex $argv [incr i]]
 	    }
 	}
 	-d* {
@@ -858,11 +962,24 @@ while  {$i < $l} {
 	    lappend domains $dom
 	}
 	-i* {
-	    set idleTime [string range $arg 2 end]
-	    if {![string length $idleTime]} {
-		set idleTime [lindex $argv [incr 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 $arg
 	    if {[RobotHref http://www.indexdata.dk/ href host path]} {
@@ -880,15 +997,15 @@ while  {$i < $l} {
 if {![info exist domains]} {
     set domains {*}
 }
-if {![info exist maxDistance]} {
-    set maxDistance 50
+if {![info exist maxdistance]} {
+    set maxdistance 50
 }
 if {![info exist robotsMax]} {
     set robotsMax 5
 }
 
 puts "domains=$domains"
-puts "max distance=$maxDistance"
+puts "max distance=$maxdistance"
 puts "max jobs=$robotsMax"
 
 RobotStart