--- pandia/indexer	2025/06/27 02:14:47	1.2
+++ pandia/indexer	2025/06/28 05:40:11	1.5
@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 
 # 
-# $Id: indexer,v 1.2 2025/06/27 02:14:47 snw Exp $
+# $Id: indexer,v 1.5 2025/06/28 05:40:11 snw Exp $
 #  Copyright (C) 2025 Coherent Logic Development LLC
 #
 # Author: Serena Willis <snw@coherent-logic.com>
@@ -9,6 +9,15 @@
 # Licensed AGPL-3.0
 #
 # $Log: indexer,v $
+# Revision 1.5  2025/06/28 05:40:11  snw
+# Exclude non-textual MIME types
+#
+# Revision 1.4  2025/06/28 00:33:32  snw
+# Update locking
+#
+# Revision 1.3  2025/06/27 16:20:30  snw
+# Add blacklist
+#
 # Revision 1.2  2025/06/27 02:14:47  snw
 # Initial operational capability
 #
@@ -25,6 +34,7 @@ use Lingua::Stem;
 use DBI;
 use Data::Dumper;
 use Try::Tiny;
+use Fcntl qw(:flock);
 
 my $dbh = "";
 my $dsn = "";
@@ -33,6 +43,9 @@ $| = 1;
 print "pandia indexer v0.0.1\n";
 print " Copyright (C) 2025 Coherent Logic Development LLC\n\n";
 
+open my $file, ">", "/tmp/pandia_indexer.lock" or die $!; 
+flock $file, LOCK_EX|LOCK_NB or die "Unable to lock file $!";
+
 GetOptions("dbhost=s" => \$dbhost,
            "dbname=s" => \$dbname,
            "dbusername=s" => \$dbusername,
@@ -68,17 +81,23 @@ while (my $hashref = $sth->fetchrow_hash
     print "pandia:  retrieving $url...\n";
     try {
 	my $del_queue = 0;
+	my $head = $http->head($hashref->{url});
+	my $headers = $head->{headers};
+	my $content_type = $headers->{'content-type'};
+
+	if($content_type ne 'text/plain' && substr($content_type, 0, 9) ne 'text/html') {
+	    print "pandia:  content type $content_type not indexable; skipping $url\n";
+	    my $sth = $dbh->prepare("DELETE FROM crawl_queue WHERE url=?");
+	    $sth->execute($url);
+	    next;
+	}
+	
 	my $response = $http->get($hashref->{url});
 
 	if(not $response->{success}) {
 	    print "pandia:  http failure; skipping $url\n";
 	    $del_queue = 1;
 	}
-	
-	#if(exists $response->{redirects}) {
-	#    print "pandia:  redirects detected; skipping $url\n";
-	#    $del_queue = 1;
-	#}
 
 	if($del_queue == 1) {
 	    my $sth = $dbh->prepare("DELETE FROM crawl_queue WHERE url=?");
@@ -92,6 +111,7 @@ while (my $hashref = $sth->fetchrow_hash
 	if($response) {
 	    $tree->parse($pagedata);   
 	    $title = $tree->look_down('_tag', 'title')->as_text;
+	    $title =~ s/[^\x00-\x7F]//g;
 	    
 	    print "pandia:  processing $url [$title]\n";
 
@@ -113,11 +133,11 @@ while (my $hashref = $sth->fetchrow_hash
     $sthd->execute($url);
     
     my $sth = $dbh->prepare("INSERT INTO keywords (word, url, url_domain, word_count) VALUES (?, ?, ?, ?)");
-    my $sths = $dbh->prepare("SELECT word_count FROM keywords WHERE word=?");
+    my $sths = $dbh->prepare("SELECT word_count FROM keywords WHERE word=? AND url=?");
     my $sthu = $dbh->prepare("UPDATE keywords SET word_count=word_count + 1 WHERE word=? AND url=?");
     foreach my $word (@words) {
 	$word =~ s/[^\x00-\x7F]//g;
-	$sths->execute($word);
+	$sths->execute($word, $url);
 
 	if($sths->rows > 0) {
 	    $sthu->execute($word, $url);