--- find3.pl	2002/01/31 16:54:32	1.1.1.1
+++ find3.pl	2002/06/19 10:58:59	1.5
@@ -1,7 +1,13 @@
 #!/usr/bin/perl -w
 
+# indexer, Dobrica Pavlinusic <dpavlin@rot13.org> 2002-01-28
+# options:	-q quiet
+# 		-d debug
+# 		-v verbose
+
 use strict;
 use DBI;
+use Getopt::Std;
 
 my $sadrzaj=0;
 my $nr=0;
@@ -13,6 +19,11 @@
 
 my $nn_dir=".";		# dir u kojem su wget-ani fileovi
 
+my %opts;
+getopts("vqd", \%opts);
+
+my $brojeva=0;
+my $zakona=0;
 
 #--------------------------------------------------------------------
 
@@ -92,7 +103,7 @@
 				$tmp_word = $word.$add;
 			}
 			if ($tmp_word =~ m/$regexp/ix) {
-#				print "$word -> $tmp_word\t-$sub, +$add, regexp: $regexp\n";
+				print "$word -> $tmp_word\t-$sub, +$add, regexp: $regexp\n" if ($opts{d});
 				push @out,lc($tmp_word);
 			}
 		}
@@ -114,12 +125,16 @@
 my @files = grep { /^CijeliBrojS/ && -f "$nn_dir/$_" } readdir(DIR);
 closedir(DIR);
 
+my $sth=$dbh->prepare("insert into nn (br,god,nr,aname,title,title_czs) values (?,?,?,?,?,?)") || die $dbh->errstr();
+
+
 foreach my $file (@files) {
 	open(IN,$file) || die "can't open $file: $!";
 
 	if ($file=~m/god=(\d+)\&br=(\d+)/) {
 		($br,$god) = ($2,$1);
-		print "$file -- $2 -- $1\n";
+		print "$file -- $2 -- $1\n" if (! $opts{q});
+		$brojeva++;
 	}
 
 	while(<IN>) {
@@ -127,34 +142,46 @@
 		s/\015//g;	# kill cr
 		tr/šðžèæŠÐŽÈÆ/¹ð¾èæ©Ð®ÈÆ/;	# 1250 -> iso8859-2
 
-		$sadrzaj++ if (m,<div class=sadrzaj>,);
-	
-		if ($sadrzaj) {
-			if (s/<a href="#([^"]+)">\s*(\d+)\.\s*<[^>]+>//i) {
-				($aname,$nr) = ($1,$2);
-				$aname = $1 if ($aname =~ m/Javascript:Mojdok\(\d+,\d+,\d+,(\d+)\);/);
-			}
-			$naslov.=$_;
+		if (m,<div class=sadrzaj>,) {
+			$sadrzaj++;
+			next;
 		}
-		
+
 		if ($sadrzaj && m,</div>,) {
 			$sadrzaj--;
 			$naslov=~s/\s+/ /g;
 			$naslov=~s/<[^>]+>//g;
 			$naslov=~s/^\s+//g;
 			$naslov=~s/\s+$//g;
-			print "$god $br $nr: $naslov\n";
+			print "$god $br $nr: $naslov\n" if ($opts{v});
 			my $naslov_czs = lc($naslov);
 			$naslov_czs =~ tr/¹©ðÐèÈæƾ®/sSdDcCcCzZ/;
 			$naslov_czs =~ tr/a-zA-Z/ /cs;	# non a-z  -> space
 			$naslov_czs = join(" ",normalize_word(split(/ /,$naslov_czs)));
-			$dbh->do("insert into nn (br,god,nr,aname,title,title_czs) values ($br,$god,$nr,'$aname','$naslov','$naslov_czs')") || die $dbh->errstr();
+			$sth->execute($br,$god,$nr,$aname,$naslov,$naslov_czs) || die $dbh->errstr();
 			$naslov="";
 			$nr=0;
+			$zakona++;
+		}
+
+		if ($sadrzaj) {
+			if (s/<a href="#([^"]+)">\s*(\d+)\.\s*<[^>]+>//i) {
+				($aname,$nr) = ($1,$2);
+			} elsif (s/<a href="Javascript:Mojdok\((\d+),(\d+),(\w+),(\d+)\)[^>]*>//i) {
+				($nr,$aname) = ($3,$4);
+				die "conflict in godina: $1 != $god"  if ($god != $1);
+				die "conflict in broj: $2 != $br"  if ($br != $2);
+			} else {
+				die "can't find nr in line: $_";
+			}
+			$naslov.=$_;
+			$naslov=~s/^\s*$nr\.*\s*//g;
 		}
+		
 	}
 
 	close(IN);
 }
 
 $dbh->do("vacuum") || die $dbh->errstr();
+print "Ukupno $brojeva brojeva NN, sa $zakona zakona...\n" if (! $opts{q});