X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=robot.tcl;h=bfe875f339e32169de55f564fdd4aec58d213643;hb=bd463f7d1f1610a3c7a3d9e678f5c4ff27f9d546;hp=5bd9f82b3231c2dae6861bbef2438c82198545d5;hpb=bb21c575e5bf08f8daf9d33218f6358e63d36c62;p=tclrobot.git
diff --git a/robot.tcl b/robot.tcl
index 5bd9f82..bfe875f 100755
--- a/robot.tcl
+++ b/robot.tcl
@@ -1,8 +1,8 @@
#!/usr/bin/tclsh
-# $Id: robot.tcl,v 1.15 2001/06/05 08:44:50 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"
+ # puts "RobotFileNext1 area=$area lead=$lead"
if {[catch {set ns [glob ${area}/*]}]} {
return {}
}
@@ -43,16 +43,18 @@ proc RobotReadRecord {inf fromurlx distancex} {
gets $inf
gets $inf
set distance [string trim [gets $inf]]
- puts "got distance = $distance"
+ # puts "got distance = $distance"
gets $inf
gets $inf
set fromurl [string trim [gets $inf]]
}
proc RobotFileNext {area} {
- global robotSeq global idleTime ns
+ global robotSeq
+ global idletime ns
+ global status
- puts "RobotFileNext robotSeq=$robotSeq"
+ # puts "RobotFileNext robotSeq=$robotSeq"
if {$robotSeq < 0} {
return {}
}
@@ -67,7 +69,9 @@ proc RobotFileNext {area} {
if {![string length $n]} {
set robotSeq -1
flush stdout
- puts "------------ N E X T R O U N D --------"
+ set statusfile [open status w]
+ puts $statusfile "$status(unvisited) $status(bad) $status(visited)"
+ close $statusfile
return wait
}
incr robotSeq
@@ -87,38 +91,47 @@ proc RobotFileNext {area} {
proc RobotFileExist {area host path} {
- puts "RobotFileExist begin area=$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 $npath]
}
proc RobotFileUnlink {area host path} {
- puts "RobotFileUnlink begin"
- puts "area=$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} {
@@ -130,11 +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"
@@ -154,22 +171,24 @@ 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
+ 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 $mode]
}
+ if {$mode == "w"} {
+ incr status($area)
+ }
cd $orgPwd
return $out
}
@@ -201,9 +220,9 @@ proc RobotRestart {url sock} {
proc RobotStart {} {
global URL
- global robotsRunning robotsMax idleTime
+ global robotsRunning robotsMax idletime
- puts "RobotStart"
+ # puts "RobotStart"
while {1} {
set url [RobotFileNext unvisited]
if {![string length $url]} {
@@ -211,7 +230,7 @@ proc RobotStart {} {
}
incr robotsRunning
if {[string compare $url wait] == 0} {
- after $idleTime RobotRR
+ after $idletime RobotRR
return
}
set r [RobotGetUrl $url {}]
@@ -254,12 +273,14 @@ 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
@@ -267,9 +288,12 @@ proc RobotHref {url hrefx hostx pathx} {
if {[string length $href] > 256} {
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
@@ -305,7 +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
@@ -313,38 +341,47 @@ 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 $c
}
}
}
+ if {$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 = $href"
- return 1
+
+ if {$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)]} {
@@ -409,81 +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 "