| 1 |
joko |
1.5 |
## $Id: Checksum.pm,v 1.4 2004/05/06 12:53:07 jonen Exp $ |
| 2 |
joko |
1.1 |
## |
| 3 |
|
|
## Copyright (c) 2002 Andreas Motl <andreas.motl@ilo.de> |
| 4 |
|
|
## |
| 5 |
|
|
## See COPYRIGHT section in pod text below for usage and distribution rights. |
| 6 |
|
|
## |
| 7 |
|
|
## ---------------------------------------------------------------------------------------- |
| 8 |
joko |
1.2 |
## $Log: Checksum.pm,v $ |
| 9 |
joko |
1.5 |
## Revision 1.4 2004/05/06 12:53:07 jonen |
| 10 |
|
|
## + added use of File::RsycP::digest |
| 11 |
|
|
## |
| 12 |
jonen |
1.4 |
## Revision 1.3 2003/05/13 08:19:00 joko |
| 13 |
|
|
## switched to crc32 |
| 14 |
|
|
## |
| 15 |
joko |
1.3 |
## Revision 1.2 2003/02/11 09:53:07 joko |
| 16 |
|
|
## + metadata-structure-change, fixed some code here |
| 17 |
|
|
## |
| 18 |
joko |
1.2 |
## Revision 1.1 2003/02/09 05:10:13 joko |
| 19 |
|
|
## + initial commit |
| 20 |
|
|
## |
| 21 |
joko |
1.1 |
## ---------------------------------------------------------------------------------------- |
| 22 |
|
|
|
| 23 |
|
|
|
| 24 |
joko |
1.3 |
=pod |
| 25 |
|
|
|
| 26 |
|
|
=head1 Todo |
| 27 |
|
|
|
| 28 |
|
|
o Data::Transfer::Sync::Compare::Slot using Compare::Struct |
| 29 |
joko |
1.5 |
o Load checksum algorithm on demand, do some negotiation |
| 30 |
joko |
1.3 |
|
| 31 |
|
|
=cut |
| 32 |
|
|
|
| 33 |
joko |
1.1 |
package Data::Transfer::Sync::Compare::Checksum; |
| 34 |
|
|
|
| 35 |
|
|
use strict; |
| 36 |
|
|
use warnings; |
| 37 |
|
|
|
| 38 |
|
|
use mixin::with qw( Data::Transfer::Sync ); |
| 39 |
|
|
|
| 40 |
joko |
1.5 |
use shortcuts qw( RUNNING_IN_HELL ); |
| 41 |
joko |
1.1 |
use Data::Dumper; |
| 42 |
|
|
|
| 43 |
joko |
1.3 |
# TODO: Load these appropriatly at runtime. |
| 44 |
|
|
use Digest::MD5 qw( md5 md5_hex md5_base64 ); |
| 45 |
|
|
use String::CRC32; |
| 46 |
joko |
1.5 |
|
| 47 |
|
|
# don't use File::RsyncP::Digest on win32 |
| 48 |
|
|
# TODO: enhance here! (e.g. negotiate proper checksum-algorithm first, apply afterwards) |
| 49 |
|
|
if (not RUNNING_IN_HELL()) { |
| 50 |
|
|
eval "use File::RsyncP::Digest;"; |
| 51 |
|
|
} |
| 52 |
joko |
1.1 |
|
| 53 |
|
|
# get logger instance |
| 54 |
|
|
my $logger = Log::Dispatch::Config->instance; |
| 55 |
|
|
|
| 56 |
|
|
|
| 57 |
joko |
1.3 |
# Maybe refactor to shortcuts::checksum? |
| 58 |
joko |
1.1 |
sub _calcChecksum { |
| 59 |
|
|
|
| 60 |
|
|
my $self = shift; |
| 61 |
|
|
my $descent = shift; |
| 62 |
|
|
my $specifier = shift; |
| 63 |
|
|
|
| 64 |
|
|
# calculate checksum for current object |
| 65 |
|
|
my $ident = $self->{node}->{$descent}->{ident}; |
| 66 |
|
|
|
| 67 |
|
|
# build dump of this node |
| 68 |
|
|
my $payload = $self->{node}->{$descent}->{payload}; |
| 69 |
|
|
#my $dump = $ident . "\n" . $item->quickdump(); |
| 70 |
|
|
#my $dump = $ident . "\n" . Dumper($item); |
| 71 |
|
|
my $dump = $ident . "\n" . $self->_dumpCompact($payload); |
| 72 |
|
|
|
| 73 |
|
|
# TODO: $logger->dump( ... ); |
| 74 |
|
|
#$logger->debug( __PACKAGE__ . ": " . $dump ); |
| 75 |
|
|
#$logger->dump( __PACKAGE__ . ": " . $dump ); |
| 76 |
|
|
|
| 77 |
|
|
# calculate checksum from dump |
| 78 |
joko |
1.3 |
|
| 79 |
|
|
# 1. md5-based fingerprint, base64 encoded (from Digest::MD5) |
| 80 |
|
|
#$self->{node}->{$descent}->{checksum} = md5_base64($dump) . '=='; |
| 81 |
|
|
|
| 82 |
|
|
# 2. 32-bit integer "hash" value (maybe faster?) (from DBI) |
| 83 |
|
|
# Note: The 32-bit integer hash from DBI seems to generate duplicates |
| 84 |
|
|
# with small payloads already in ranges of hundreds of items/rows!!! |
| 85 |
|
|
# Try to avoid it or try to use it only for payloads greater than, hmmm, let's say 30 chars? |
| 86 |
|
|
# (we had about 15 chars average per item (row)) |
| 87 |
joko |
1.5 |
# Possible (generic) solution: Just generate checksum, if length(checksum(payload)) < length(payload) |
| 88 |
joko |
1.1 |
#$self->{node}->{$descent}->{checksum} = DBI::hash($dump, 1); |
| 89 |
joko |
1.3 |
|
| 90 |
joko |
1.5 |
# 3. good old crc32??? |
| 91 |
jonen |
1.4 |
#$self->{node}->{$descent}->{checksum} = crc32($dump); |
| 92 |
joko |
1.3 |
|
| 93 |
jonen |
1.4 |
# 4. File::RsyncP::Digest - Perl interface to rsync message digest algorithms |
| 94 |
joko |
1.5 |
if (not RUNNING_IN_HELL()) { |
| 95 |
jonen |
1.4 |
my $rsDigest = new File::RsyncP::Digest; |
| 96 |
|
|
$rsDigest->add($dump); |
| 97 |
|
|
my $digest = $rsDigest->digest(); |
| 98 |
|
|
$self->{node}->{$descent}->{checksum} = unpack("H*", $digest); |
| 99 |
joko |
1.5 |
|
| 100 |
|
|
} else { # fallback to Digest::MD5 on win32 |
| 101 |
|
|
$self->{node}->{$descent}->{checksum} = md5_base64($dump) . '=='; |
| 102 |
|
|
} |
| 103 |
jonen |
1.4 |
|
| 104 |
|
|
# 5. some more modern Digest::SHA1 or similar? |
| 105 |
joko |
1.1 |
|
| 106 |
|
|
# signal good |
| 107 |
|
|
return 1; |
| 108 |
|
|
|
| 109 |
|
|
} |
| 110 |
|
|
|
| 111 |
|
|
|
| 112 |
|
|
sub _readChecksum { |
| 113 |
|
|
my $self = shift; |
| 114 |
|
|
|
| 115 |
|
|
my $descent = shift; |
| 116 |
|
|
|
| 117 |
|
|
# signal checksum bad |
| 118 |
|
|
if (!$self->{node}->{$descent}) { |
| 119 |
|
|
return; |
| 120 |
|
|
} |
| 121 |
|
|
|
| 122 |
|
|
# trace |
| 123 |
|
|
#print "desc: $descent", "\n"; |
| 124 |
|
|
#print Dumper($self); |
| 125 |
joko |
1.2 |
#print Dumper($self->{meta}->{$descent}); |
| 126 |
joko |
1.1 |
#exit; |
| 127 |
|
|
|
| 128 |
|
|
# get checksum for current entry |
| 129 |
|
|
# TODO: don't have the checksum column/property hardcoded as "cs" here, make this configurable somehow |
| 130 |
joko |
1.2 |
if ($self->{meta}->{$descent}->{isChecksumAuthority}) { |
| 131 |
joko |
1.1 |
$self->_calcChecksum($descent); |
| 132 |
|
|
} else { |
| 133 |
|
|
$self->{node}->{$descent}->{checksum} = $self->{node}->{$descent}->{payload}->{cs}; |
| 134 |
|
|
} |
| 135 |
|
|
|
| 136 |
|
|
# signal checksum good |
| 137 |
|
|
return 1; |
| 138 |
|
|
|
| 139 |
|
|
} |
| 140 |
|
|
|
| 141 |
|
|
|
| 142 |
|
|
1; |
| 143 |
joko |
1.3 |
__END__ |