X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=793298c6ff9f955b149d72ca8e653e59ce27eb43;hb=2e203397567386e2806359f5ac678491d184b047;hp=b6466c347186d9954fbdda74c78ed7054fab76c5;hpb=9d3f82cd1140362487d8fa6372cac1b24a49d21e;p=tclrobot.git

diff --git a/robot.tcl b/robot.tcl
index b6466c3..793298c 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 1.33 2002/03/25 16:13:21 adam Exp $
 #
 proc RobotFileNext1 {area lead} {
     # puts "RobotFileNext1 area=$area lead=$lead"
@@ -171,23 +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 $d]} {
         set out [open f$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]
     }
@@ -293,12 +288,11 @@ 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
@@ -502,6 +496,13 @@ proc link {url out href body distance} {
 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)]} {
@@ -510,36 +511,67 @@ proc RobotTextHtml {url out} {
     }
     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 $out "</documentcontent>"
+            # don't print title of document content if noindex is used
+            if {!$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)]} {
 	        continue
             }
             set href [string trim $parm(href)]
             if {![RobotHref $url href host path]} continue
             set URL($url,bpath) $path
-        } -nonest a {
+        } a {
+            # <a href="...."> .. </a> 
+            # we're not using nonest - otherwise body isn't set
+            if {$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)]} {
 	        continue
             }