Files
percona-toolkit/lib/NibbleIterator.pm

262 lines
8.9 KiB
Perl

# This program is copyright 2011 Percona Inc.
# 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.
# ###########################################################################
# NibbleIterator package
# ###########################################################################
{
# Package: NibbleIterator
# NibbleIterator nibbles tables.
package NibbleIterator;
use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant MKDEBUG => $ENV{MKDEBUG} || 0;
use Data::Dumper;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
sub new {
my ( $class, %args ) = @_;
my @required_args = qw(dbh tbl OptionParser Quoter TableNibbler TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($dbh, $tbl, $o, $q) = @args{@required_args};
# Get an index to nibble by. We'll order rows by the index's columns.
my $index = $args{TableParser}->find_best_index(
$tbl->{tbl_struct},
$o->get('chunk-index'),
);
die "No index to nibble table $tbl->{db}.$tbl->{tbl}" unless $index;
my $index_cols = $tbl->{tbl_struct}->{keys}->{$index}->{cols};
# Figure out how to nibble the table with the index.
my $asc = $args{TableNibbler}->generate_asc_stmt(
%args,
tbl_struct => $tbl->{tbl_struct},
index => $index,
asc_only => 1,
);
# Make SQL statements, prepared on first call to next(). The preamble
# and ORDER BY are the same for all statements. FORCE IDNEX and ORDER BY
# are needed to ensure deterministic nibbling.
my $nibble_sql_preamble
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{cols}})
. " FROM " . $q->quote(@{$tbl}{qw(db tbl)})
. " FORCE INDEX(`$index`)";
my $order_by = "ORDER BY " . join(', ', map {$q->quote($_)} @{$index_cols});
# This statement is only executed once, so it doesn't use a sth.
my $first_lb_sql
= $nibble_sql_preamble
. ($args{where} ? " WHERE $args{where}" : '')
. " $order_by "
. " LIMIT 1"
. " /*first lower boundary*/";
MKDEBUG && _d('First lower boundary statement:', $first_lb_sql);
# Nibbles are inclusive, so for a..z, the nibbles are: a-e, f-j, k-o, p-t,
# u-y, and z. This complicates getting the next upper boundary because
# if we use either (col >= lb AND col < ub) or (col > lb AND col <= ub)
# in nibble_sql (below), then that fails for either the last or first
# nibble respectively. E.g. (col >= z AND col < z) doesn't work, nor
# does (col > a AND col <= e). Hence the fancy LIMIT 2 which returns
# the upper boundary for the current nibble *and* the lower boundary
# for the next nibble. See _next_boundaries().
my $ub_sql
= $nibble_sql_preamble
. " WHERE (" . $asc->{boundaries}->{'>='} . ")" # lower boundary
. ($args{where} ? " AND ($args{where})" : '')
. " $order_by "
. " LIMIT 2 OFFSET " . (($o->get('chunk-size') || 1) - 1)
. " /*upper boundary*/";
MKDEBUG && _d('Next upper boundary statement:', $ub_sql);
my $nibble_sql
= $nibble_sql_preamble
. " WHERE (" . $asc->{boundaries}->{'>='} . ")" # lower boundary
. " AND (" . $asc->{boundaries}->{'<='} . ")" # upper boundary
. ($args{where} ? " AND ($args{where})" : '')
. " $order_by"
. " /*nibble*/";
MKDEBUG && _d('Nibble statement:', $nibble_sql);
my $self = {
%args,
asc => $asc,
first_lb_sql => $first_lb_sql,
ub_sql => $ub_sql,
nibble_sql => $nibble_sql,
nibbleno => 0,
have_rows => 0,
rowno => 0,
};
return bless $self, $class;
}
sub next {
my ($self) = @_;
# First call, init everything. This could be done in new(), but
# all work is delayed until actually needed.
if ($self->{nibbleno} == 0) {
$self->_prepare_sths();
$self->_get_first_lb();
# $self->_check_index_usage();
}
# Return rows in nibble. sth->{Active} is always true with DBD::mysql v3,
# so we track the status manually. have_rows will be true if a previous
# call got a nibble with rows. When there's no more rows in this nibble,
# try to get the next nibble.
if ( $self->{have_rows} ) {
my $row = $self->{nibble_sth}->fetchrow_arrayref();
if ( $row ) {
$self->{rowno}++;
MKDEBUG && _d('Row', $self->{rowno}, 'in nibble', $self->{nibbleno});
# fetchrow_arraryref re-uses its internal arrayref, so we must copy.
return [ @$row ];
}
MKDEBUG && _d('No more rowso in nibble', $self->{nibbleno});
$self->{rowno} = 0;
$self->{have_rows} = 0;
}
# If there's another boundary, fetch the rows within it.
if ( $self->_next_boundaries() ) {
MKDEBUG && _d($self->{nibble_sth}->{Statement}, 'params:',
join(', ', (@{$self->{lb}}, @{$self->{ub}})));
$self->{nibble_sth}->execute(@{$self->{lb}}, @{$self->{ub}});
$self->{have_rows} = $self->{nibble_sth}->rows();
if ( $self->{have_rows} ) {
$self->{nibbleno}++;
MKDEBUG && _d($self->{have_rows}, 'rows in nibble', $self->{nibbleno});
return $self->next();
}
}
MKDEBUG && _d('Done nibbling');
return;
}
sub _prepare_sths {
my ($self) = @_;
MKDEBUG && _d('Preparing statement handles');
$self->{ub_sth} = $self->{dbh}->prepare($self->{ub_sql});
$self->{nibble_sth} = $self->{dbh}->prepare($self->{nibble_sql});
}
sub _get_first_lb {
my ($self) = @_;
$self->{next_lb} = $self->{dbh}->selectrow_arrayref($self->{first_lb_sql});
MKDEBUG && _d('First lower boundary:', Dumper($self->{lb}));
return;
}
sub _check_index_usage {
my ($self) = @_;
my ($dbh, $tbl, $q) = @{$self}{qw(dbh tbl Quoter)};
my $table_status;
eval {
my $sql = "SHOW TABLE STATUS FROM " . $q->quote($tbl->{db})
. " LIKE " . $q->literal_like($tbl->{tbl});
MKDEBUG && _d($sql);
$table_status = $dbh->selectrow_hashref($sql);
};
MKDEBUG && $EVAL_ERROR && _d($EVAL_ERROR);
my $small_table;
if ( $table_status ) {
my $n_rows = defined $table_status->{Rows} ? $table_status->{Rows}
: defined $table_status->{rows} ? $table_status->{rows}
: undef;
$small_table = 1 if defined $n_rows && $n_rows <= 100;
}
MKDEBUG && _d('Small table:', $small_table);
if ( !$small_table ) {
my $explain;
eval {
$explain = $dbh->selectall_arrayref("", {Slice => {}});
};
if ( $EVAL_ERROR ) {
MKDEBUG && _d($EVAL_ERROR);
return;
}
MKDEBUG && _d('EXPLAIN key:', $explain->[0]->{key});
my $explain_index = lc($explain->[0]->{key} || '');
if ( $explain_index ne lc($self->{asc}->{index}) ) {
die "Cannot nibble table $tbl->{db}.$tbl->{tbl} because MySQL chose "
. ($explain_index ? "the `$explain_index`" : 'no') . ' index'
. " instead of the `$self->{asc}->{index}` index";
}
}
return;
}
sub _next_boundaries {
my ($self) = @_;
if ( $self->{no_more_boundaries} ) {
MKDEBUG && _d('No more boundaries');
return;
}
$self->{lb} = $self->{next_lb};
MKDEBUG && _d($self->{ub_sth}->{Statement}, 'params:',
join(', ', @{$self->{lb}}));
$self->{ub_sth}->execute(@{$self->{lb}});
my $boundary = $self->{ub_sth}->fetchall_arrayref();
if ( $boundary && @$boundary ) {
$self->{ub} = $boundary->[0]; # this nibble
$self->{next_lb} = $boundary->[1]; # next nibble
$self->{ub_sth}->finish();
MKDEBUG && _d('Next upper boundary:', Dumper($self->{ub}));
}
else {
$self->{no_more_boundaries} = 1; # for next call
$self->{ub} = $self->{lb};
MKDEBUG && _d('Last upper boundary:', Dumper($self->{ub}));
}
return 1; # have boundaries
}
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 NibbleIterator package
# ###########################################################################