search.in

harvest是一个下载html网页得机器人

#
# Copyright & Disclaimer.
#	This set of routines may be freely distributed, modified and
#	used, provided this copyright & disclaimer remains intact.
#	This package is used at your own risk, if it does what you
#	want, good; if it doesn't, modify it or use something else--but
#	don't blame me. Support level = negligable (i.e. mail bugs but
#	not requests for extensions)

# Usage:
#	&get_request;    will get the request and decode it into an
#			 indexed array %rqpairs, the raw request is in
#			 $request
#
#	... = &url_decode(LIST); will return a URL decoded version of
#			         the contents of LIST
#

sub get_request {

    # Subroutine get_request reads the POST or GET form request from STDIN
    # into the variable  $request, and then splits it into its
    # name=value pairs in the associative array %rqpairs.
    # The number of bytes is given in the environment variable
    # CONTENT_LENGTH which is automatically set by the request generator.

    # Encoded HEX values and spaces are decoded in the values at this
    # stage.

    # $request will contain the RAW request. N.B. spaces and other
    # special characters are not handler in the name field.

    if ($ENV{'REQUEST_METHOD'} eq "POST") {
	read(STDIN, $request, $ENV{'CONTENT_LENGTH'});
    } elsif ($ENV{'REQUEST_METHOD'} eq "GET" ) {
	$request = $ENV{'QUERY_STRING'};
    }

    @F = split(/[&=]/, $request);
    &url_decode(@F);
}

sub url_decode {
#	Decode a URL encoded string or array of strings
#		+ -> space
#		%xx -> character xx

    foreach (@_) {
	tr/+/ /;
	s/%(..)/pack("c",hex($1))/ge;
    }

    # This gross stuff handles multiply defined attributes.  For example:
    #    foo=abc&foo=xyz
    # comes back as
    #    $RQ{'foo'} eq 'abc\0xyz'
    # Otherwise, the previous method just gave us one or the other.    -DW&HW
    #
    local ($k, $v);
    local (%Y);
    while (($k=shift @_) ne '' && ($#_ >= 0)) {
	$v=shift @_;
	$Y{$k} = defined $Y{$k} ? join ("\0", $Y{$k}, $v) : $v;
    }
    %Y;
}
######################################################################

# encode special characters for urls - h.weinreich
sub url_encode {
	$_ = shift;

	s/([^a-zA-Z0-9\%\ \-\+])/uc('%'.unpack("H*", $1))/eg;
	s/ /+/g;

	return $_;
}


# sort the objects by number of matched lines
#
#sub bynml {
#	split(/\n/, $b) <=> split(/\n/, $a);	# number of lines in object
#}

# sort the objects by "rank". This is based on code from Wesley Alan Wright
# modified and optimized by h.weinreich

sub rank_objects {
    local (@objects) = @_;
    local ($objnum) = 0;
    local (%ratings);
    local (@object_index);
    foreach $object (@objects) {
	$rank = 1;
	$lastline = ""; # don't count repeated lines
	split(/\n/,$object);
	foreach $line (@_) {

	    # $rawline=$line;

	    $line =~ s/\s{2,}/ /g;                # Remove multiple whitespaces
	    $line =~ s/(Matched line: )(.*)$/$2/;
	    $line =~ s/^(.*)\{(\d+)\}(.*)$/$1$3/; # Remove "{12}:" etc.

	    # if (open(BQLOG, ">>$BQLOG")) {
	    #	printf BQLOG ($rawline."\n");
	    #	printf BQLOG ($line."\n");
	    #	printf BQLOG ($rank."\n");
	    #	close BQLOG;
	    # }

	    if ($line ne $lastline) {

		#the domain name is often highly significant...
		if (index($line,"url#") > 0) {
		    $temphost = $line;
		    $temphost =~ s/^(.*)\/\/(([a-z]|\.|-)*)\/(.*)$/$2/;
		    $rank += 700 if (index($temphost,$query) > 0 );
		}

		#title is most important
		$rank += 600 - length($line)*3
		    if (index($line,"title#") > 0 && length($line) < 200);

		#url is also important
		$rank += 150 - length($line)/1.5
		    if (index($line,"url#") > 0 && length($line) < 200);

		#there are some very long headings...
		$rank += 110 - length($line)/2
		    if (index($line,"headings#") > 0 && length($line)<200);

		# used for PDF, PS etc.
		$rank += 80 - length($line)/3
		    if (index($line,"description#") > 0 && length($line)<200);

		$rank += 60 if (index($line,"subject#") > 0);
		$rank += 30 if (index($line,"images#") > 0);

		# these are sometimes too long.
		$rank += 25 - length($line)/8
		    if (index($line,"url-references#") > 0 );

		# somehow redundant, not as important as it may seem...
		$rank += 20 if (index($line,"keywords#") > 0);

		$rank += 10 if (index($line,"body#") > 0);

		# Consider length of URL of every Page
		$rank += 400 - length($line)*2
		    if (index($line,"120 - ") == 0 && length($line) < 200);
	    }
	    $lastline = $line;
	}
	$ratings{$objnum} = $rank;
	$object_index[$objnum] = $objnum;
	$objnum++;
    }
#   pop off the "126" and "103" results (they're at the end)
    pop(@object_index);
    pop(@object_index);
    @sorted_index= sort byrate @object_index;
    $objnum=0;

    # the first element has the highest weight
    $maxWeight = $ratings{$sorted_index[0]};

    foreach $index (@sorted_index) {
	$OBJ[$objnum] = $objects[$index];
	$aWeight[$objnum] = $ratings{$index}; # save weight of each object
	$objnum++;
    }
}

sub byrate {
    $ratings{$b} <=> $ratings{$a};
}


# translate SGML entities
# produced from
# perl -ne 'if (/^ *<.ENTITY *([\S]*) *CDATA *"&#([\d]+);".*>/)
#              { if ($2 <0x100) {printf "s/&$1;/\\x%x/g;\n", $2;}
# 		 else {printf "s/&$1;/\\\\#$2/g;\n";}}'

sub entities {
    $_ = $_[0];

    # ISO Latin 1
s/&nbsp;/\xa0/g;
s/&iexcl;/\xa1/g;
s/&cent;/\xa2/g;
s/&pound;/\xa3/g;
s/&curren;/\xa4/g;
s/&yen;/\xa5/g;
s/&brvbar;/\xa6/g;
s/&sect;/\xa7/g;
s/&uml;/\xa8/g;
s/&copy;/\xa9/g;
s/&ordf;/\xaa/g;
s/&laquo;/\xab/g;
s/&not;/\xac/g;
s/&shy;/\xad/g;
s/&reg;/\xae/g;
s/&macr;/\xaf/g;
s/&deg;/\xb0/g;
s/&plusmn;/\xb1/g;
s/&sup2;/\xb2/g;
s/&sup3;/\xb3/g;
s/&acute;/\xb4/g;
s/&micro;/\xb5/g;
s/&para;/\xb6/g;
s/&middot;/\xb7/g;
s/&cedil;/\xb8/g;
s/&sup1;/\xb9/g;
s/&ordm;/\xba/g;
s/&raquo;/\xbb/g;
s/&frac14;/\xbc/g;
s/&frac12;/\xbd/g;
s/&frac34;/\xbe/g;
s/&iquest;/\xbf/g;
s/&Agrave;/\xc0/g;
s/&Aacute;/\xc1/g;
s/&Acirc;/\xc2/g;
s/&Atilde;/\xc3/g;
s/&Auml;/\xc4/g;
s/&Aring;/\xc5/g;
s/&AElig;/\xc6/g;
s/&Ccedil;/\xc7/g;
s/&Egrave;/\xc8/g;
s/&Eacute;/\xc9/g;
s/&Ecirc;/\xca/g;
s/&Euml;/\xcb/g;
s/&Igrave;/\xcc/g;
s/&Iacute;/\xcd/g;
s/&Icirc;/\xce/g;
s/&Iuml;/\xcf/g;
s/&ETH;/\xd0/g;
s/&Ntilde;/\xd1/g;
s/&Ograve;/\xd2/g;
s/&Oacute;/\xd3/g;
s/&Ocirc;/\xd4/g;
s/&Otilde;/\xd5/g;
s/&Ouml;/\xd6/g;
s/&times;/\xd7/g;
s/&Oslash;/\xd8/g;
s/&Ugrave;/\xd9/g;
s/&Uacute;/\xda/g;
s/&Ucirc;/\xdb/g;
s/&Uuml;/\xdc/g;
s/&Yacute;/\xdd/g;
s/&THORN;/\xde/g;
s/&szlig;/\xdf/g;
s/&agrave;/\xe0/g;
s/&aacute;/\xe1/g;
s/&acirc;/\xe2/g;
s/&atilde;/\xe3/g;
s/&auml;/\xe4/g;
s/&aring;/\xe5/g;
s/&aelig;/\xe6/g;
s/&ccedil;/\xe7/g;
s/&egrave;/\xe8/g;
s/&eacute;/\xe9/g;
s/&ecirc;/\xea/g;
s/&euml;/\xeb/g;
s/&igrave;/\xec/g;
s/&iacute;/\xed/g;
s/&icirc;/\xee/g;
s/&iuml;/\xef/g;
s/&eth;/\xf0/g;
s/&ntilde;/\xf1/g;
s/&ograve;/\xf2/g;
s/&oacute;/\xf3/g;
s/&ocirc;/\xf4/g;
s/&otilde;/\xf5/g;
s/&ouml;/\xf6/g;
s/&divide;/\xf7/g;
s/&oslash;/\xf8/g;
s/&ugrave;/\xf9/g;
s/&uacute;/\xfa/g;
s/&ucirc;/\xfb/g;
s/&uuml;/\xfc/g;
s/&yacute;/\xfd/g;
s/&thorn;/\xfe/g;
s/&yuml;/\xff/g;

    # Cougar HTMLsym
s/&fnof;/\\#402/g;
s/&Alpha;/\\#913/g;
s/&Beta;/\\#914/g;
s/&Gamma;/\\#915/g;
s/&Delta;/\\#916/g;
s/&Epsilon;/\\#917/g;
s/&Zeta;/\\#918/g;
s/&Eta;/\\#919/g;
s/&Theta;/\\#920/g;
s/&Iota;/\\#921/g;
s/&Kappa;/\\#922/g;
s/&Lambda;/\\#923/g;
s/&Mu;/\\#924/g;
s/&Nu;/\\#925/g;
s/&Xi;/\\#926/g;
s/&Omicron;/\\#927/g;
s/&Pi;/\\#928/g;
s/&Rho;/\\#929/g;
s/&Sigma;/\\#931/g;
s/&Tau;/\\#932/g;
s/&Upsilon;/\\#933/g;
s/&Phi;/\\#934/g;
s/&Chi;/\\#935/g;
s/&Psi;/\\#936/g;
s/&Omega;/\\#937/g;
s/&alpha;/\\#945/g;
s/&beta;/\\#946/g;
s/&gamma;/\\#947/g;
s/&delta;/\\#948/g;
s/&epsilon;/\\#949/g;
s/&zeta;/\\#950/g;
s/&eta;/\\#951/g;
s/&theta;/\\#952/g;
s/&iota;/\\#953/g;
s/&kappa;/\\#954/g;
s/&lambda;/\\#955/g;
s/&mu;/\\#956/g;
s/&nu;/\\#957/g;
s/&xi;/\\#958/g;
s/&omicron;/\\#959/g;
s/&pi;/\\#960/g;
s/&rho;/\\#961/g;
s/&sigmaf;/\\#962/g;
s/&sigma;/\\#963/g;
s/&tau;/\\#964/g;
s/&upsilon;/\\#965/g;
s/&phi;/\\#966/g;
s/&chi;/\\#967/g;
s/&psi;/\\#968/g;
s/&omega;/\\#969/g;
s/&thetasym;/\\#977/g;
s/&upsih;/\\#978/g;
s/&piv;/\\#982/g;
s/&bull;/\\#8226/g;
s/&hellip;/\\#8230/g;
s/&prime;/\\#8242/g;
s/&Prime;/\\#8243/g;
s/&oline;/\\#8254/g;
s/&frasl;/\\#8260/g;
s/&weierp;/\\#8472/g;
s/&image;/\\#8465/g;
s/&real;/\\#8476/g;
s/&trade;/\\#8482/g;
s/&alefsym;/\\#8501/g;
s/&larr;/\\#8592/g;
s/&uarr;/\\#8593/g;
s/&rarr;/\\#8594/g;
s/&darr;/\\#8595/g;
s/&harr;/\\#8596/g;
s/&crarr;/\\#8629/g;
s/&lArr;/\\#8656/g;
s/&uArr;/\\#8657/g;
s/&rArr;/\\#8658/g;
s/&dArr;/\\#8659/g;
s/&hArr;/\\#8660/g;
s/&forall;/\\#8704/g;
s/&part;/\\#8706/g;
s/&exist;/\\#8707/g;
s/&empty;/\\#8709/g;
s/&nabla;/\\#8711/g;
s/&isin;/\\#8712/g;
s/&notin;/\\#8713/g;
s/&ni;/\\#8715/g;
s/&prod;/\\#8719/g;
s/&sum;/\\#8721/g;
s/&minus;/\\#8722/g;
s/&lowast;/\\#8727/g;
s/&radic;/\\#8730/g;
s/&prop;/\\#8733/g;
s/&infin;/\\#8734/g;
s/&ang;/\\#8736/g;
s/&and;/\\#8869/g;
s/&or;/\\#8870/g;
s/&cap;/\\#8745/g;
s/&cup;/\\#8746/g;
s/&int;/\\#8747/g;
s/&there4;/\\#8756/g;
s/&sim;/\\#8764/g;
s/&cong;/\\#8773/g;
s/&asymp;/\\#8776/g;
s/&ne;/\\#8800/g;
s/&equiv;/\\#8801/g;
s/&le;/\\#8804/g;
s/&ge;/\\#8805/g;
s/&sub;/\\#8834/g;
s/&sup;/\\#8835/g;
s/&nsub;/\\#8836/g;
s/&sube;/\\#8838/g;
s/&supe;/\\#8839/g;
s/&oplus;/\\#8853/g;
s/&otimes;/\\#8855/g;
s/&perp;/\\#8869/g;
s/&sdot;/\\#8901/g;
s/&lceil;/\\#8968/g;
s/&rceil;/\\#8969/g;
s/&lfloor;/\\#8970/g;
s/&rfloor;/\\#8971/g;
s/&lang;/\\#9001/g;
s/&rang;/\\#9002/g;
s/&loz;/\\#9674/g;
s/&spades;/\\#9824/g;
s/&clubs;/\\#9827/g;
s/&hearts;/\\#9829/g;
s/&diams;/\\#9830/g;

    # Cougar HTMLmisc
s/&quot;/\x22/g;
s/&amp;/\x26/g;
s/&lt;/\x3c/g;
s/&gt;/\x3e/g;
s/&OElig;/\\#338/g;
s/&oelig;/\\#339/g;
s/&Scaron;/\\#352/g;
s/&scaron;/\\#353/g;
s/&Yuml;/\\#376/g;
s/&circ;/\\#710/g;
s/&tilde;/\\#732/g;
s/&ensp;/\\#8194/g;
s/&emsp;/\\#8195/g;
s/&thinsp;/\\#8201/g;
s/&zwnj;/\\#8204/g;
s/&zwj;/\\#8205/g;
s/&lrm;/\\#8206/g;
s/&rlm;/\\#8207/g;
s/&ndash;/\\#8211/g;
s/&mdash;/\\#8212/g;
s/&lsquo;/\\#8216/g;
s/&rsquo;/\\#8217/g;
s/&sbquo;/\\#8218/g;
s/&ldquo;/\\#8220/g;
s/&rdquo;/\\#8221/g;
s/&bdquo;/\\#8222/g;
s/&dagger;/\\#8224/g;
s/&Dagger;/\\#8225/g;
s/&permil;/\\#8240/g;
s/&lsaquo;/\\#8249/g;
s/&rsaquo;/\\#8250/g;

    return($_);
}

sub ceil {
	my $x = shift;
	$x += 1.0 unless ($x == int($x));
	return int($x);
}