Files
percona-toolkit/lib/RowDiff.pm
2013-01-02 17:19:16 -07:00

256 lines
10 KiB
Perl

# This program is copyright 2007-2011 Baron Schwartz, 2011 Percona Ireland Ltd.
# Feedback and improvements are welcome.
#
# THIS PROGRAM IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED
# WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
# MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
#
# 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 the Free Software
# Foundation, version 2; OR the Perl Artistic License. On UNIX and similar
# systems, you can issue `man perlgpl' or `man perlartistic' to read these
# licenses.
#
# You should have received a copy of the GNU General Public License along with
# this program; if not, write to the Free Software Foundation, Inc., 59 Temple
# Place, Suite 330, Boston, MA 02111-1307 USA.
# ###########################################################################
# RowDiff package
# ###########################################################################
{
# Package: RowDiff
# RowDiff compares two sets of rows to find ones that are different.
package RowDiff;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
# Required args:
# * dbh obj: dbh used for collation-specific string comparisons
# Optional args:
# * same_row Callback when rows are identical
# * not_in_left Callback when right row is not in the left
# * not_in_right Callback when left row is not in the right
# * key_cmp Callback when a column value differs
# * done Callback that stops compare_sets() if it returns true
# * trf Callback to transform numeric values before comparison
sub new {
my ( $class, %args ) = @_;
die "I need a dbh" unless $args{dbh};
my $self = { %args };
return bless $self, $class;
}
# Arguments:
# * left_sth obj: sth
# * right_sth obj: sth
# * syncer obj: TableSync* module
# * tbl_struct hashref: table struct from TableParser::parser()
# Iterates through two sets of rows and finds differences. Calls various
# methods on the $syncer object when it finds differences, passing these
# args and hashrefs to the differing rows ($lr and $rr).
sub compare_sets {
my ( $self, %args ) = @_;
my @required_args = qw(left_sth right_sth syncer tbl_struct);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless defined $args{$arg};
}
my $left_sth = $args{left_sth};
my $right_sth = $args{right_sth};
my $syncer = $args{syncer};
my $tbl_struct = $args{tbl_struct};
my ($lr, $rr); # Current row from the left/right sths.
$args{key_cols} = $syncer->key_cols(); # for key_cmp()
# We have to manually track if the left or right sth is done
# fetching rows because sth->{Active} is always true with
# DBD::mysql v3. And we cannot simply while ( $lr || $rr )
# because in the case where left and right have the same key,
# we do this:
# $lr = $rr = undef; # Fetch another row from each side.
# Unsetting both $lr and $rr there would cause while () to
# terminate. (And while ( $lr && $rr ) is not what we want
# either.) Furthermore, we need to avoid trying to fetch more
# rows if there are none to fetch because doing this would
# cause a DBI error ("fetch without execute"). That's why we
# make these checks:
# if ( !$lr && !$left_done )
# if ( !$rr && !$right_done )
# If you make changes here, be sure to test both RowDiff.t
# and RowDiff-custom.t. Look inside the later to see what
# is custom about it.
my $left_done = 0;
my $right_done = 0;
my $done = $self->{done};
do {
if ( !$lr && !$left_done ) {
PTDEBUG && _d('Fetching row from left');
eval { $lr = $left_sth->fetchrow_hashref(); };
PTDEBUG && $EVAL_ERROR && _d($EVAL_ERROR);
$left_done = !$lr || $EVAL_ERROR ? 1 : 0;
}
elsif ( PTDEBUG ) {
_d('Left still has rows');
}
if ( !$rr && !$right_done ) {
PTDEBUG && _d('Fetching row from right');
eval { $rr = $right_sth->fetchrow_hashref(); };
PTDEBUG && $EVAL_ERROR && _d($EVAL_ERROR);
$right_done = !$rr || $EVAL_ERROR ? 1 : 0;
}
elsif ( PTDEBUG ) {
_d('Right still has rows');
}
my $cmp;
if ( $lr && $rr ) {
$cmp = $self->key_cmp(%args, lr => $lr, rr => $rr);
PTDEBUG && _d('Key comparison on left and right:', $cmp);
}
if ( $lr || $rr ) {
# If the current row is the "same row" on both sides, meaning the two
# rows have the same key, check the contents of the row to see if
# they're the same.
if ( $lr && $rr && defined $cmp && $cmp == 0 ) {
PTDEBUG && _d('Left and right have the same key');
$syncer->same_row(%args, lr => $lr, rr => $rr);
$self->{same_row}->(%args, lr => $lr, rr => $rr)
if $self->{same_row};
$lr = $rr = undef; # Fetch another row from each side.
}
# The row in the left doesn't exist in the right.
elsif ( !$rr || ( defined $cmp && $cmp < 0 ) ) {
PTDEBUG && _d('Left is not in right');
$syncer->not_in_right(%args, lr => $lr, rr => $rr);
$self->{not_in_right}->(%args, lr => $lr, rr => $rr)
if $self->{not_in_right};
$lr = undef;
}
# Symmetric to the above.
else {
PTDEBUG && _d('Right is not in left');
$syncer->not_in_left(%args, lr => $lr, rr => $rr);
$self->{not_in_left}->(%args, lr => $lr, rr => $rr)
if $self->{not_in_left};
$rr = undef;
}
}
$left_done = $right_done = 1 if $done && $done->(%args);
} while ( !($left_done && $right_done) );
PTDEBUG && _d('No more rows');
$syncer->done_with_rows();
}
# Compare two rows to determine how they should be ordered. NULL sorts before
# defined values in MySQL, so I consider undef "less than." Numbers are easy to
# compare. Otherwise string comparison is tricky. This function must match
# MySQL exactly or the merge algorithm runs off the rails, so when in doubt I
# ask MySQL to compare strings for me. I can handle numbers and "normal" latin1
# characters without asking MySQL. See
# http://dev.mysql.com/doc/refman/5.0/en/charset-literal.html. $r1 and $r2 are
# row hashrefs. $key_cols is an arrayref of the key columns to compare. $tbl is the
# structure returned by TableParser. The result matches Perl's cmp or <=>
# operators:
# 1 cmp 0 => 1
# 1 cmp 1 => 0
# 1 cmp 2 => -1
# TODO: must generate the comparator function dynamically for speed, so we don't
# have to check the type of columns constantly
sub key_cmp {
my ( $self, %args ) = @_;
my @required_args = qw(lr rr key_cols tbl_struct);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless exists $args{$arg};
}
my ($lr, $rr, $key_cols, $tbl_struct) = @args{@required_args};
PTDEBUG && _d('Comparing keys using columns:', join(',', @$key_cols));
# Optional callbacks.
my $callback = $self->{key_cmp};
my $trf = $self->{trf};
foreach my $col ( @$key_cols ) {
my $l = $lr->{$col};
my $r = $rr->{$col};
if ( !defined $l || !defined $r ) {
PTDEBUG && _d($col, 'is not defined in both rows');
return defined $l ? 1 : defined $r ? -1 : 0;
}
else {
if ( $tbl_struct->{is_numeric}->{$col} ) { # Numeric column
PTDEBUG && _d($col, 'is numeric');
($l, $r) = $trf->($l, $r, $tbl_struct, $col) if $trf;
my $cmp = $l <=> $r;
if ( $cmp ) {
PTDEBUG && _d('Column', $col, 'differs:', $l, '!=', $r);
$callback->($col, $l, $r) if $callback;
return $cmp;
}
}
# Do case-sensitive cmp, expecting most will be eq. If that fails, try
# a case-insensitive cmp if possible; otherwise ask MySQL how to sort.
elsif ( $l ne $r ) {
my $cmp;
my $coll = $tbl_struct->{collation_for}->{$col};
if ( $coll && ( $coll ne 'latin1_swedish_ci'
|| $l =~ m/[^\040-\177]/ || $r =~ m/[^\040-\177]/) )
{
PTDEBUG && _d('Comparing', $col, 'via MySQL');
$cmp = $self->db_cmp($coll, $l, $r);
}
else {
PTDEBUG && _d('Comparing', $col, 'in lowercase');
$cmp = lc $l cmp lc $r;
}
if ( $cmp ) {
PTDEBUG && _d('Column', $col, 'differs:', $l, 'ne', $r);
$callback->($col, $l, $r) if $callback;
return $cmp;
}
}
}
}
return 0;
}
sub db_cmp {
my ( $self, $collation, $l, $r ) = @_;
if ( !$self->{sth}->{$collation} ) {
if ( !$self->{charset_for} ) {
PTDEBUG && _d('Fetching collations from MySQL');
my @collations = @{$self->{dbh}->selectall_arrayref(
'SHOW COLLATION', {Slice => { collation => 1, charset => 1 }})};
foreach my $collation ( @collations ) {
$self->{charset_for}->{$collation->{collation}}
= $collation->{charset};
}
}
my $sql = "SELECT STRCMP(_$self->{charset_for}->{$collation}? COLLATE $collation, "
. "_$self->{charset_for}->{$collation}? COLLATE $collation) AS res";
PTDEBUG && _d($sql);
$self->{sth}->{$collation} = $self->{dbh}->prepare($sql);
}
my $sth = $self->{sth}->{$collation};
$sth->execute($l, $r);
return $sth->fetchall_arrayref()->[0]->[0];
}
sub _d {
my ($package, undef, $line) = caller 0;
@_ = map { (my $temp = $_) =~ s/\n/\n# /g; $temp; }
map { defined $_ ? $_ : 'undef' }
@_;
print STDERR "# $package:$line $PID ", join(' ', @_), "\n";
}
1;
}
# ###########################################################################
# End RowDiff package
# ###########################################################################