X-Git-Url: http://git.hungrycats.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=faster-dupemerge;h=8df7885878b2085bb2cfc8b61e05566c239225b0;hb=7b251f6413738a3fef1a97ff6e74137460e03292;hp=6371fd037747c419cde6efce2c2a33e8b6788d6a;hpb=2f607d26931ff8ed86b7f41d19f62679f8ac916d;p=dupemerge

diff --git a/faster-dupemerge b/faster-dupemerge
index 6371fd0..8df7885 100755
--- a/faster-dupemerge
+++ b/faster-dupemerge
@@ -4,7 +4,7 @@ use Fcntl qw(:DEFAULT :flock);
 use File::Compare;
 use File::Temp;
 
-# Copyright (C) 2003-2006 Zygo Blaxell <zblaxell@feedme.hungrycats.org>
+# Copyright (C) 2003-2010 Zygo Blaxell <faster-dupemerge@mailtoo.hungrycats.org>
 
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -305,14 +305,14 @@ sub merge_files {
 	}
 
 	print STDERR "Merging...\n" if $debug;
-	foreach my $candidate (@candidate_list) {
+	foreach my $candidate (sort @candidate_list) {
 		print STDERR "\tDigesting candidate $candidate\n" if $debug;
 		my $ok = 0;
 		my $digest;
 
 hash_file:
 
-		foreach my $filename (keys(%{$inode_to_file_name{$candidate}})) {
+		foreach my $filename (sort keys(%{$inode_to_file_name{$candidate}})) {
 			print STDERR "\t\tDigesting file $filename\n" if $debug;
 			if ((-l $filename) || ! -f _) {
 				warn "Bogon file " . tick_quote($filename);
@@ -333,8 +333,9 @@ hash_file:
 		if ($ok) {
 			print STDERR "\t\tDigest is $digest\n" if $debug;
 
-			my $incumbent = $hash_to_inode{$digest};
-			if (defined($incumbent)) {
+			my $incumbent_list = ($hash_to_inode{$digest} ||= []);
+			my $incumbent_matched = 0;
+			for my $incumbent (sort @$incumbent_list) {
 				print STDERR "\t\tInodes $incumbent and $candidate have same hash\n" if $debug;
 
 				my $finished = 0;
@@ -342,9 +343,9 @@ hash_file:
 link_start:
 
 				until ($finished) {
-					my @incumbent_names = keys(%{$inode_to_file_name{$incumbent}});
-					my @candidate_names = keys(%{$inode_to_file_name{$candidate}});
-					print STDERR "\t\tLinks to $incumbent:",   join("\n\t\t\t", '', @incumbent_names),   "\n" if $debug;
+					my @incumbent_names = sort keys(%{$inode_to_file_name{$incumbent}});
+					my @candidate_names = sort keys(%{$inode_to_file_name{$candidate}});
+					print STDERR "\t\tLinks to $incumbent:", join("\n\t\t\t", '', @incumbent_names), "\n" if $debug;
 					print STDERR "\t\tLinks to $candidate:", join("\n\t\t\t", '', @candidate_names), "\n" if $debug;
 
 incumbent_file:
@@ -411,6 +412,7 @@ candidate_file:
 										print STDERR "$quoted_incumbent_file and $quoted_candidate_file have same hash but do not compare equal!\n" unless $skip_hashes;
 									} else {
 										$identical = 1;
+										$incumbent_matched = 1;
 									}
 									$compare_count++;
 									$compare_bytes += $incumbent_size;
@@ -441,6 +443,12 @@ candidate_file:
 									my $link_done = 0;
 
 									my ($from_file, $to_file, $from_inode, $to_inode, $from_nlink, $to_nlink);
+
+									# If the candidate has more links than incumbent, replace incumbent with candidate.
+									# If the incumbent has more links than candidate, replace candidate with incumbent.
+									# If the link counts are equal, we saw incumbent first, so keep the incumbent.
+									# "We saw incumbent first" is significant because we explicitly sort the inodes.
+									# Thank Johannes Niess for this idea.
 									if ($candidate_nlink > $incumbent_nlink) {
 										$from_file = $candidate_file;
 										$to_file = $incumbent_file;
@@ -487,13 +495,9 @@ candidate_file:
 									# My random number generator chooses the incumbent's size.
 
 									if ($link_done) {
-										# Since we're in a dry run, the filesystem doesn't change.
-										# Our notion of what the filesystem should look like should not change either.
 										delete $inode_to_file_name{$to_inode}->{$to_file};
-										unless ($dry_run) {
-											$inode_to_file_name{$from_inode}->{$to_file} = undef;
-											$hash_to_inode{$digest} = $from_inode;
-										}
+										$inode_to_file_name{$from_inode}->{$to_file} = undef;
+										$hash_to_inode{$digest} = [ $from_inode ];
 
 										$hard_links++;
 										if ($to_nlink == 1) {
@@ -521,9 +525,10 @@ candidate_file:
 					}
 					$finished = 1;
 				}
-			} else {
+			}
+			unless ($incumbent_matched) {
 				print STDERR "\t\tNew hash entered\n" if $debug;
-				$hash_to_inode{$digest} = $candidate;
+				push(@$incumbent_list, $candidate);
 			}
 		} else {
 			warn "No digests found for inode $candidate\n";
@@ -538,7 +543,7 @@ end_merge:
 }
 
 while (<FIND>) {
-	my ($weak_key, $inode, $name) = m/^(\d+ \d+ \d+ \d+ -?\d+) (\d+:\d+) (.+)\0$/so;
+	my ($weak_key, $inode, $name) = m/^(\d+ \d+ \d+ \d+ -?[\d.]+) (\d+:\d+) (.+)\0$/so;
 	die "read error: $!\nLast input line was '$_'" unless defined($name);
 
 	print STDERR "weak_key=$weak_key inode=$inode name=$name\n" if $debug;