X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=bfe875f339e32169de55f564fdd4aec58d213643;hb=bd463f7d1f1610a3c7a3d9e678f5c4ff27f9d546;hp=ee70b9afc44461c421fb18f837fb5fcc502e62af;hpb=7476a63e6732f7f51eea10bf38daaea4a31be57f;p=tclrobot.git

diff --git a/robot.tcl b/robot.tcl
index ee70b9a..bfe875f 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,5 +1,5 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.28 2001/11/13 11:17:26 adam Exp $
+# $Id: robot.tcl,v 1.32 2002/03/25 16:11:08 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
     # puts "RobotFileNext1 area=$area lead=$lead"
@@ -69,7 +69,9 @@ proc RobotFileNext {area} {
     if {![string length $n]} {
 	set robotSeq -1
 	flush stdout
-	puts "Round robin un,ba,vi=$status(unvisited),$status(bad),$status(visited)"
+        set statusfile [open status w]
+        puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+        close $statusfile
 	return wait
     }
     incr robotSeq
@@ -89,14 +91,20 @@ proc RobotFileNext {area} {
 
 
 proc RobotFileExist {area host path} {
-    # puts "RobotFileExist begin 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 "RobotFileExist end npath=$npath"
+    if {$debuglevel > 3} {
+        puts "RobotFileExist end npath=$npath"
+    }
     return [file exists $npath]
 }
 
@@ -136,11 +144,14 @@ 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"
@@ -160,20 +171,18 @@ proc RobotFileOpen {area host path {mode w}} {
             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 $path /robots.txt]} {
+		    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 $mode]
-	} else {
-	    set out [open f$d $mode]
-	}
+        set out [open f$d $mode]
     } else {
         set out [open f $mode]
     }
@@ -279,12 +288,12 @@ 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
-    }
+#    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
@@ -320,7 +329,11 @@ proc RobotHref {url hrefx hostx pathx} {
     }
     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
@@ -433,182 +446,142 @@ proc RobotRedirect {url tourl code} {
     }
 }
 
+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 maxdistance
 
+    # 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 distance [expr $URL($url,dist) + 1]
+        set fdistance $URL($url,dist)
+	set distance [expr $fdistance + 1]
     }
     htmlSwitch $URL($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 [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)
+                    }
+                }
+            }
+	    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 $out "</documentcontent>"
-        } -nonest a {
+            # don't print title of document content if noindex is used
+            if {!$noindex} {
+                puts $out "<title>$title</title>"
+                regsub -all {<!--[^-]*-->} $body { } abody
+                regsub -all -nocase {<script[^<]*</script>} $abody {} bbody
+                regsub -all {<[^\>]+>} $bbody {} nbody
+                puts $out "<documentcontent>"
+                puts $out $nbody
+                puts $out "</documentcontent>"
+            }
+        } -nonest base {
+            # <base href=.. >
             if {![info exists parm(href)]} {
-	        puts "no href"
 	        continue
             }
-	    if {[expr $distance <= $maxdistance]} {
-		set href [string trim $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]} {
-		    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
-		    }
-		}
-	    }
+            set href [string trim $parm(href)]
+            if {![RobotHref $url href host path]} continue
+            set URL($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 $distance
         } -nonest area {
+            if {$nofollow} continue
             if {![info exists parm(href)]} {
-	        puts "no href"
 	        continue
             }
-	    if {[expr $distance <= $maxdistance]} {
-		set href [string trim $parm(href)]
-		if {![RobotHref $url href host path]} continue
-		
-		puts $out "<cr>"
-		puts $out "<identifier>$href</identifier>"
-		puts $out "<description></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
-		    }
-		}
-	    }
+            link $url $out [string trim $parm(href)] $body $distance
         } -nonest frame {
             if {![info exists parm(src)]} {
-	        puts "no src"
 	        continue
             }
-	    if {[expr $distance <= $maxdistance]} {
-		set href [string trim $parm(src)]
-		if {![RobotHref $url href host path]} continue
-		
-		puts $out "<cr>"
-		puts $out "<identifier>$href</identifier>"
-		puts $out "<description></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
-		    }
-		}
-	    }
+            link $url $out [string trim $parm(src)] $body $fdistance
 	}
 }
 
@@ -1082,4 +1055,7 @@ while {$robotsRunning} {
     vwait robotsRunning
 }
 
-puts "End un,ba,vi=$status(unvisited),$status(bad),$status(visited)"
+set statusfile [open status w]
+puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+close $statusfile
+