Implemented Allow/deny rules. Better Tcl autoconfig.

[tclrobot.git] / robot.tcl

diff --git a/robot.tcl b/robot.tcl
index c9388bc..7c3dbe6 100755 (executable)
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,5 +1,5 @@
 #!/usr/bin/tclsh 
-# $Id: robot.tcl,v 1.20 2001/06/29 22:25:55 adam Exp $
+# $Id: robot.tcl,v 1.21 2001/10/26 13:26:11 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
@@ -344,14 +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)]} {
@@ -417,10 +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) \
@@ -446,7 +451,7 @@ 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
                
@@ -496,7 +501,7 @@ 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
                
@@ -599,6 +604,11 @@ proc Robot200 {url} {
     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)]
     puts $out "<zmbot>"
 
@@ -615,19 +625,19 @@ 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>"
     RobotFileClose $out
@@ -654,9 +664,11 @@ 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 $sock
@@ -685,7 +697,7 @@ proc RobotReadHeader {url sock} {
                    set URL($url,head,[string tolower $name]) [string trim $value]
                }
            }
-           puts "code = $code"
+           puts "HTTP CODE $code"
            set URL($url,state) skip
            switch $code {
                301 {
@@ -747,7 +759,7 @@ proc RobotNop {} {
 proc RobotGetUrl {url phost} {
     global URL robotsRunning
     flush stdout
-    puts "RobotGetUrl --------- robotsRunning=$robotsRunning url=$url"
+    puts "Retrieve $robotsRunning url=$url"
     if {![regexp {([^:]+)://([^/]+)(.*)} $url x method hostport path]} {
         return -1
     }
@@ -803,7 +815,7 @@ 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)"
 }
@@ -819,29 +831,91 @@ 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
+            if {![string match [lindex $l 2] $this]} continue
+            # OK, we have a match
+            if {[lindex $l 0] == "allow"} {
+                if {$debuglevel > 3} {
+                    puts "CHECKRULE MATH OK"
+                }
+                return 1
+            } else {
+                if {$debuglevel > 3} {
+                    puts "CHECKFULE MATCH FAIL"
+                }
+                return 0
+            }
+        }
+    }
+    if {$debuglevel > 3} {
+        puts "CHECKRULE MATH 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 -- $arg {
@@ -852,9 +926,9 @@ while  {$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 $argv [incr i]]
            }
        }
        -d* {
@@ -865,9 +939,9 @@ while  {$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 $argv [incr i]]
            }
        }
         -l* {
@@ -876,6 +950,13 @@ while  {$i < $l} {
                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 $arg
            if {[RobotHref http://www.indexdata.dk/ href host path]} {
@@ -893,15 +974,15 @@ while  {$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