percona/percona-toolkit
1
0
Fork 0
mirror of https://github.com/percona/percona-toolkit.git synced 2026-05-16 01:07:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #422 from percona/PT-1757_nibble_small_tables

Browse Source 
PT-1757 Implemented fallback to nibble in NibbleIterator
This commit is contained in:
Carlos Salguero
2019-10-14 14:45:43 -03:00
committed by GitHub
parent ed0b883eda 13f7afc49d
commit d45d393d8a
16 changed files with 671 additions and 474 deletions
Download Patch File Download Diff File Expand all files Collapse all files
bin/pt-archiver
+10 -6
View File
@@ -3136,6 +3136,7 @@ sub generate_asc_stmt {
cols => \@cols,
quoter => $q,
is_nullable => $tbl_struct->{is_nullable},
type_for => $tbl_struct->{type_for},
);
$asc_stmt->{boundaries}->{$cmp} = $cmp_where->{where};
}
@@ -3156,6 +3157,7 @@ sub generate_cmp_where {
my @slice = @{$args{slice}};
my @cols = @{$args{cols}};
my $is_nullable = $args{is_nullable};
my $type_for = $args{type_for};
my $type = $args{type};
my $q = $self->{Quoter};
@@ -3172,13 +3174,14 @@ sub generate_cmp_where {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @clause, "(($val IS NULL AND $quo IS NULL) OR ($quo = $val))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @clause, "$quo = ?";
push @clause, "$quo = $val";
push @r_slice, $ord;
push @r_scols, $col;
}
@@ -3188,15 +3191,16 @@ sub generate_cmp_where {
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
push @clause, "(? IS NULL OR $quo $type ?)";
push @clause, "($val IS NULL OR $quo $type $val)";
}
elsif ( $type =~ m/>/ ) {
push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
push @clause, "(($val IS NULL AND $quo IS NOT NULL) OR ($quo $cmp $val)";
}
else { # If $type =~ m/</ ) {
push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
push @clauses, "(($val IS NOT NULL AND $quo IS NULL) OR ($quo $cmp $val))";
}
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
@@ -3204,7 +3208,7 @@ sub generate_cmp_where {
else {
push @r_slice, $ord;
push @r_scols, $col;
push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
push @clause, ($type =~ m/=/ && $end ? "$quo $type $val" : "$quo $cmp $val");
}
push @clauses, '(' . join(' AND ', @clause) . ')';
bin/pt-online-schema-change
+133 -100
View File
@@ -3057,15 +3057,13 @@ sub generate_cmp_where {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $val = $type_for->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
#push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @clause, "(($val IS NULL AND $quo IS NULL) OR ($quo = $val))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
#push @clause, "$quo = ?";
push @clause, "$quo = $val";
push @r_slice, $ord;
push @r_scols, $col;
@@ -3076,18 +3074,15 @@ sub generate_cmp_where {
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
my $val = $type_for->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
#push @clause, "(? IS NULL OR $quo $type ?)";
push @clause, "($val IS NULL OR $quo $type $val)";
}
elsif ( $type =~ m/>/ ) {
#push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
push @clause, "(($val IS NULL AND $quo IS NOT NULL) OR ($quo $cmp $val)";
}
else { # If $type =~ m/</ ) {
#push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
push @clauses, "(($val IS NOT NULL AND $quo IS NULL) OR ($quo $cmp $val))";
}
push @r_slice, $ord, $ord;
@@ -3096,7 +3091,6 @@ sub generate_cmp_where {
else {
push @r_slice, $ord;
push @r_scols, $col;
#push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
push @clause, ($type =~ m/=/ && $end ? "$quo $type $val" : "$quo $cmp $val");
}
@@ -5494,88 +5488,130 @@ sub new {
my @cols = grep { !$ignore_col->{$_} } @$all_cols;
my $self;
if ( $nibble_params->{one_nibble} ) {
my $params = _one_nibble(\%args, \@cols, $where, $tbl, \%comments);
$self = {
%args,
one_nibble => 1,
limit => 0,
nibble_sql => $params->{nibble_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
};
} else {
my $params = _nibble_params($nibble_params, $tbl, \%args, \@cols, $chunk_size, $where, \%comments, $q);
$self = {
%args,
index => $params->{index},
limit => $params->{limit},
first_lb_sql => $params->{first_lb_sql},
last_ub_sql => $params->{last_ub_sql},
ub_sql => $params->{ub_sql},
nibble_sql => $params->{nibble_sql},
explain_first_lb_sql => $params->{explain_first_lb_sql},
explain_ub_sql => $params->{explain_ub_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
resume_lb_sql => $params->{resume_lb_sql},
sql => $params->{sql},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
$self->{nibble_params} = $nibble_params;
$self->{tbl} = $tbl;
$self->{args} = \%args;
$self->{cols} = \@cols;
$self->{chunk_size} = $chunk_size;
$self->{where} = $where;
$self->{comments} = \%comments;
return bless $self, $class;
}
sub switch_to_nibble {
my $self = shift;
my $params = _nibble_params($self->{nibble_params}, $self->{tbl}, $self->{args}, $self->{cols},
$self->{chunk_size}, $self->{where}, $self->{comments}, $self->{Quoter});
$self->{one_nibble} = 0;
$self->{index} = $params->{index};
$self->{limit} = $params->{limit};
$self->{first_lb_sql} = $params->{first_lb_sql};
$self->{last_ub_sql} = $params->{last_ub_sql};
$self->{ub_sql} = $params->{ub_sql};
$self->{nibble_sql} = $params->{nibble_sql};
$self->{explain_first_lb_sql} = $params->{explain_first_lb_sql};
$self->{explain_ub_sql} = $params->{explain_ub_sql};
$self->{explain_nibble_sql} = $params->{explain_nibble_sql};
$self->{resume_lb_sql} = $params->{resume_lb_sql};
$self->{sql} = $params->{sql};
$self->_get_bounds();
$self->_prepare_sths();
}
sub _one_nibble {
my ($args, $cols, $where, $tbl, $comments) = @_;
my $q = new Quoter();
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
# : join(', ', map { $q->quote($_) } @cols))
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ?
"CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @cols))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ?
"CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{bite}*/";
PTDEBUG && _d('One nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
# : join(', ', map { $q->quote($_) } @cols))
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ?
"CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @cols))
. ($args->{select} ? $args->{select}
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum'
? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{bite}*/";
PTDEBUG && _d('Explain one nibble statement:', $explain_nibble_sql);
$self = {
%args,
return {
one_nibble => 1,
limit => 0,
nibble_sql => $nibble_sql,
explain_nibble_sql => $explain_nibble_sql,
};
}
else {
}
sub _nibble_params {
my ($nibble_params, $tbl, $args, $cols, $chunk_size, $where, $comments, $q) = @_;
my $index = $nibble_params->{index}; # brevity
my $index_cols = $tbl->{tbl_struct}->{keys}->{$index}->{cols};
my $asc = $args{TableNibbler}->generate_asc_stmt(
%args,
my $asc = $args->{TableNibbler}->generate_asc_stmt(
%$args,
tbl_struct => $tbl->{tbl_struct},
index => $index,
n_index_cols => $args{n_chunk_index_cols},
cols => \@cols,
n_index_cols => $args->{n_chunk_index_cols},
cols => $cols,
asc_only => 1,
);
PTDEBUG && _d('Ascend params:', Dumper($asc));
# my $force_concat_enums = $o->has('force-concat-enums') && $o->get('force-concat-enums');
# my $i=0;
# for my $index (@{$index_cols}) {
# last if $args{n_chunk_index_cols} && $i >= $args{n_chunk_index_cols};
# $i++;
# if ($tbl->{tbl_struct}->{type_for}->{$index} eq 'enum') {
# if ($tbl->{tbl_struct}->{defs}->{$index} =~ m/enum\s*\((.*?)\)/) {
# my @items = split(/,\s*/, $1);
# my $sorted = 1; # Asume the items list is sorted to later check if this is true
# for (my $i=1; $i < scalar(@items); $i++) {
# if ($items[$i-1] gt $items[$i]) {
# $sorted = 0;
# last;
# }
# }
# if (!$force_concat_enums && !$sorted) {
# die "The index " . $index . " in table " . $tbl->{name} .
# " has unsorted enum items.\nPlease read the documentation for the --force-concat-enums parameter\n";
# }
# }
# }
# }
my $force_concat_enums;
my $from = "$tbl->{name} FORCE INDEX(`$index`)";
#my $order_by = join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
# ? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by = join(', ', map {$q->quote($_)} @{$index_cols});
#my $order_by_dec = join(' DESC,', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
# ? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by_dec = join(' DESC,', map {$q->quote($_)} @{$index_cols});
my $first_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
#. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
@@ -5585,10 +5621,9 @@ sub new {
PTDEBUG && _d('First lower boundary statement:', $first_lb_sql);
my $resume_lb_sql;
if ( $args{resume} ) {
if ( $args->{resume} ) {
$resume_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
#. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>'}
@@ -5601,7 +5636,6 @@ sub new {
my $last_ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
#. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
@@ -5613,7 +5647,6 @@ sub new {
my $ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
#. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='}
@@ -5624,37 +5657,36 @@ sub new {
PTDEBUG && _d('Upper boundary statement:', $ub_sql);
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
# : join(', ', map { $q->quote($_) } @{$asc->{cols}}))
: join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{cols}}))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
: join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{nibble}*/";
PTDEBUG && _d('Nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
. ($args->{select} ? $args->{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{nibble}*/";
PTDEBUG && _d('Explain nibble statement:', $explain_nibble_sql);
my $limit = $chunk_size - 1;
PTDEBUG && _d('Initial chunk size (LIMIT):', $limit);
$self = {
%args,
my $params = {
one_nibble => 0,
index => $index,
limit => $limit,
first_lb_sql => $first_lb_sql,
@@ -5673,17 +5705,7 @@ sub new {
order_by => $order_by,
},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
return bless $self, $class;
return $params;
}
sub next {
@@ -5720,7 +5742,6 @@ sub next {
NIBBLE:
while ( $self->{have_rows} || $self->_next_boundaries() ) {
if ($self->{pause_file}) {
while(-f $self->{pause_file}) {
print "Sleeping $self->{sleep} seconds because $self->{pause_file} exists\n";
@@ -5741,7 +5762,7 @@ sub next {
if ( !$self->{have_rows} ) {
$self->{nibbleno}++;
PTDEBUG && _d('Nibble:', $self->{nibble_sth}->{Statement}, 'params:',
join(', ', (@{$self->{lower}} || [], @{$self->{upper} }||[])));
join(', ', (@{$self->{lower} || []}, @{$self->{upper} || []})));
if ( my $callback = $self->{callbacks}->{exec_nibble} ) {
$self->{have_rows} = $callback->(%callback_args);
}
@@ -6140,7 +6161,7 @@ sub _next_boundaries {
PTDEBUG && _d($self->{ub_sth}->{Statement}, 'params:',
join(', ', @{$self->{lower}} || []), $self->{limit});
join(', ', @{$self->{lower}}), $self->{limit});
$self->{ub_sth}->execute(@{$self->{lower}}, $self->{limit});
my $boundary = $self->{ub_sth}->fetchall_arrayref();
PTDEBUG && _d('Next boundary:', Dumper($boundary));
@@ -6181,10 +6202,7 @@ sub identical_boundaries {
if scalar @$b1 != scalar @$b2; # shouldn't happen
my $n_vals = scalar @$b1;
for my $i ( 0..($n_vals-1) ) {
next if (!defined($b1->[$i]) && !defined($b2->[$i]));
return 0 if (!defined($b1->[$i]) && defined($b2->[$i])); # diff
return 0 if (defined($b1->[$i]) && !defined($b2->[$i])); # diff
return 0 if $b1->[$i] ne $b2->[$i]; # diff
return 0 if ($b1->[$i] || '') ne ($b2->[$i] || ''); # diff
}
return 1;
}
@@ -6647,6 +6665,7 @@ use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Data::Dumper;
use Carp;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
@@ -6659,7 +6678,7 @@ sub new {
}
my $self = {
Quoter => $args{Quoter},
Quoter => $args{Quoter},
};
return bless $self, $class;
@@ -6716,8 +6735,16 @@ sub _get_first_values {
my $index_struct = $tbl->{tbl_struct}->{keys}->{$index};
my $index_cols = $index_struct->{cols};
my $index_columns = join (', ',
my $index_columns;
eval {
$index_columns = join (', ',
map { $q->quote($_) } @{$index_cols}[0..($n_index_cols - 1)]);
};
if ($EVAL_ERROR) {
confess "$EVAL_ERROR";
}
my @where;
foreach my $col ( @{$index_cols}[0..($n_index_cols - 1)] ) {
@@ -6751,8 +6778,7 @@ sub _make_range_query {
if ( $n_index_cols > 1 ) {
foreach my $n ( 0..($n_index_cols - 2) ) {
my $col = $index_cols->[$n];
my $val = $vals->[$n];
$val = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
my $val = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
push @where, $q->quote($col) . " = " . $val;
}
}
@@ -9077,7 +9103,7 @@ sub main {
# (*) Frank: commented them out because it caused infinite loop
# and the mentioned test error doesn't arise
#my $original_error = $EVAL_ERROR;
my $original_error = $EVAL_ERROR;
foreach my $task ( reverse @cleanup_tasks ) {
eval {
@@ -9087,7 +9113,7 @@ sub main {
warn "Error cleaning up: $EVAL_ERROR\n";
}
}
#die $original_error if $original_error; # rethrow original error
die $original_error if $original_error; # rethrow original error
return;
}
);
@@ -9579,6 +9605,7 @@ sub main {
# a single chunk on all slaves. E.g. if a slave is out of sync
# and has a lot more rows than the master, single chunking on the
# master could cause the slave to choke.
my $failed_to_nibble;
if ( $nibble_iter->one_nibble() ) {
PTDEBUG && _d('Getting table row estimate on replicas');
my @too_large;
@@ -9605,10 +9632,16 @@ sub main {
. ($tbl->{chunk_size} * $limit)
. " rows (chunk size=$tbl->{chunk_size}"
. " * chunk size limit=$limit).\n";
die ts($msg);
warn $msg;
$failed_to_nibble = 1;
warn "Switching to nibble\n";
$nibble_iter->switch_to_nibble();
# die ts($msg);
} else {
return 1;
}
}
else { # chunking the table
if (!$nibble_iter->one_nibble()) { # chunking the table
if ( $o->get('check-plan') ) {
my $idx_len = new IndexLength(Quoter => $q);
my ($key_len, $key) = $idx_len->index_length(
bin/pt-table-checksum
+152 -109
View File
@@ -4892,6 +4892,7 @@ sub generate_asc_stmt {
cols => \@cols,
quoter => $q,
is_nullable => $tbl_struct->{is_nullable},
type_for => $tbl_struct->{type_for},
);
$asc_stmt->{boundaries}->{$cmp} = $cmp_where->{where};
}
@@ -4912,6 +4913,7 @@ sub generate_cmp_where {
my @slice = @{$args{slice}};
my @cols = @{$args{cols}};
my $is_nullable = $args{is_nullable};
my $type_for = $args{type_for};
my $type = $args{type};
my $q = $self->{Quoter};
@@ -4928,13 +4930,14 @@ sub generate_cmp_where {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @clause, "(($val IS NULL AND $quo IS NULL) OR ($quo = $val))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @clause, "$quo = ?";
push @clause, "$quo = $val";
push @r_slice, $ord;
push @r_scols, $col;
}
@@ -4944,15 +4947,16 @@ sub generate_cmp_where {
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
push @clause, "(? IS NULL OR $quo $type ?)";
push @clause, "($val IS NULL OR $quo $type $val)";
}
elsif ( $type =~ m/>/ ) {
push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
push @clause, "(($val IS NULL AND $quo IS NOT NULL) OR ($quo $cmp $val)";
}
else { # If $type =~ m/</ ) {
push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
push @clauses, "(($val IS NOT NULL AND $quo IS NULL) OR ($quo $cmp $val))";
}
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
@@ -4960,7 +4964,7 @@ sub generate_cmp_where {
else {
push @r_slice, $ord;
push @r_scols, $col;
push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
push @clause, ($type =~ m/=/ && $end ? "$quo $type $val" : "$quo $cmp $val");
}
push @clauses, '(' . join(' AND ', @clause) . ')';
@@ -6293,7 +6297,6 @@ use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use IndexLength;
use Data::Dumper;
$Data::Dumper::Indent = 1;
@@ -6331,82 +6334,131 @@ sub new {
my @cols = grep { !$ignore_col->{$_} } @$all_cols;
my $self;
if ( $nibble_params->{one_nibble} ) {
my $params = _one_nibble(\%args, \@cols, $where, $tbl, \%comments);
$self = {
%args,
one_nibble => 1,
limit => 0,
nibble_sql => $params->{nibble_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
};
} else {
my $params = _nibble_params($nibble_params, $tbl, \%args, \@cols, $chunk_size, $where, \%comments, $q);
$self = {
%args,
index => $params->{index},
limit => $params->{limit},
first_lb_sql => $params->{first_lb_sql},
last_ub_sql => $params->{last_ub_sql},
ub_sql => $params->{ub_sql},
nibble_sql => $params->{nibble_sql},
explain_first_lb_sql => $params->{explain_first_lb_sql},
explain_ub_sql => $params->{explain_ub_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
resume_lb_sql => $params->{resume_lb_sql},
sql => $params->{sql},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
$self->{nibble_params} = $nibble_params;
$self->{tbl} = $tbl;
$self->{args} = \%args;
$self->{cols} = \@cols;
$self->{chunk_size} = $chunk_size;
$self->{where} = $where;
$self->{comments} = \%comments;
return bless $self, $class;
}
sub switch_to_nibble {
my $self = shift;
my $params = _nibble_params($self->{nibble_params}, $self->{tbl}, $self->{args}, $self->{cols},
$self->{chunk_size}, $self->{where}, $self->{comments}, $self->{Quoter});
$self->{one_nibble} = 0;
$self->{index} = $params->{index};
$self->{limit} = $params->{limit};
$self->{first_lb_sql} = $params->{first_lb_sql};
$self->{last_ub_sql} = $params->{last_ub_sql};
$self->{ub_sql} = $params->{ub_sql};
$self->{nibble_sql} = $params->{nibble_sql};
$self->{explain_first_lb_sql} = $params->{explain_first_lb_sql};
$self->{explain_ub_sql} = $params->{explain_ub_sql};
$self->{explain_nibble_sql} = $params->{explain_nibble_sql};
$self->{resume_lb_sql} = $params->{resume_lb_sql};
$self->{sql} = $params->{sql};
$self->_get_bounds();
$self->_prepare_sths();
}
sub _one_nibble {
my ($args, $cols, $where, $tbl, $comments) = @_;
my $q = new Quoter();
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ?
"CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{bite}*/";
PTDEBUG && _d('One nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
. ($args->{select} ? $args->{select}
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum'
? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{bite}*/";
PTDEBUG && _d('Explain one nibble statement:', $explain_nibble_sql);
$self = {
%args,
return {
one_nibble => 1,
limit => 0,
nibble_sql => $nibble_sql,
explain_nibble_sql => $explain_nibble_sql,
};
}
else {
}
sub _nibble_params {
my ($nibble_params, $tbl, $args, $cols, $chunk_size, $where, $comments, $q) = @_;
my $index = $nibble_params->{index}; # brevity
my $index_cols = $tbl->{tbl_struct}->{keys}->{$index}->{cols};
my $asc = $args{TableNibbler}->generate_asc_stmt(
%args,
my $asc = $args->{TableNibbler}->generate_asc_stmt(
%$args,
tbl_struct => $tbl->{tbl_struct},
index => $index,
n_index_cols => $args{n_chunk_index_cols},
cols => \@cols,
n_index_cols => $args->{n_chunk_index_cols},
cols => $cols,
asc_only => 1,
);
PTDEBUG && _d('Ascend params:', Dumper($asc));
my $force_concat_enums = $o->has('force-concat-enums') && $o->get('force-concat-enums');
my $i=0;
for my $index (@{$index_cols}) {
last if $args{n_chunk_index_cols} && $i >= $args{n_chunk_index_cols};
$i++;
if ($tbl->{tbl_struct}->{type_for}->{$index} eq 'enum') {
if ($tbl->{tbl_struct}->{defs}->{$index} =~ m/enum\s*\((.*?)\)/) {
my @items = split(/,\s*/, $1);
my $sorted = 1; # Asume the items list is sorted to later check if this is true
for (my $i=1; $i < scalar(@items); $i++) {
if ($items[$i-1] gt $items[$i]) {
$sorted = 0;
last;
}
}
if (!$force_concat_enums && !$sorted) {
die "The index " . $index . " in table " . $tbl->{name} .
" has unsorted enum items.\nPlease read the documentation for the --force-concat-enums parameter\n";
}
}
}
}
my $force_concat_enums;
my $from = "$tbl->{name} FORCE INDEX(`$index`)";
my $order_by = join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by_dec = join(' DESC,', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by = join(', ', map {$q->quote($_)} @{$index_cols});
my $order_by_dec = join(' DESC,', map {$q->quote($_)} @{$index_cols});
my $first_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY $order_by"
@@ -6415,10 +6467,10 @@ sub new {
PTDEBUG && _d('First lower boundary statement:', $first_lb_sql);
my $resume_lb_sql;
if ( $args{resume} ) {
if ( $args->{resume} ) {
$resume_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>'}
. ($where ? " AND ($where)" : '')
@@ -6430,7 +6482,7 @@ sub new {
my $last_ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY "
@@ -6441,7 +6493,7 @@ sub new {
my $ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='}
. ($where ? " AND ($where)" : '')
@@ -6451,36 +6503,36 @@ sub new {
PTDEBUG && _d('Upper boundary statement:', $ub_sql);
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
: join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{nibble}*/";
PTDEBUG && _d('Nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
. ($args->{select} ? $args->{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{nibble}*/";
PTDEBUG && _d('Explain nibble statement:', $explain_nibble_sql);
my $limit = $chunk_size - 1;
PTDEBUG && _d('Initial chunk size (LIMIT):', $limit);
$self = {
%args,
my $params = {
one_nibble => 0,
index => $index,
limit => $limit,
first_lb_sql => $first_lb_sql,
@@ -6499,17 +6551,7 @@ sub new {
order_by => $order_by,
},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
return bless $self, $class;
return $params;
}
sub next {
@@ -6546,7 +6588,6 @@ sub next {
NIBBLE:
while ( $self->{have_rows} || $self->_next_boundaries() ) {
if ($self->{pause_file}) {
while(-f $self->{pause_file}) {
print "Sleeping $self->{sleep} seconds because $self->{pause_file} exists\n";
@@ -6593,6 +6634,7 @@ sub next {
}
$self->{rowno} = 0;
$self->{have_rows} = 0;
}
PTDEBUG && _d('Done nibbling');
@@ -6692,11 +6734,11 @@ sub row_estimate {
sub can_nibble {
my (%args) = @_;
my @required_args = qw(Cxn tbl chunk_size OptionParser TableParser Quoter);
my @required_args = qw(Cxn tbl chunk_size OptionParser TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $chunk_size, $o, $q) = @args{@required_args};
my ($cxn, $tbl, $chunk_size, $o) = @args{@required_args};
my $where = $o->has('where') ? $o->get('where') : '';
@@ -6707,28 +6749,13 @@ sub can_nibble {
);
if ( !$where ) {
$mysql_index = undef;
}
$mysql_index = undef;
}
my $chunk_size_limit = $o->get('chunk-size-limit') || 1;
my $one_nibble = !defined $args{one_nibble} || $args{one_nibble}
? $row_est <= $chunk_size * $chunk_size_limit
: 0;
if ($mysql_index) {
my $idx_len = IndexLength->new(Quoter => $q);
my ($key_len, $key) = $idx_len->index_length(
Cxn => $args{Cxn},
tbl => $tbl,
index => $mysql_index,
n_index_cols => $o->get('chunk-index-columns'),
);
if ( !$key || !$key_len || lc($key) ne lc($mysql_index)) {
$one_nibble = 1;
}
}
PTDEBUG && _d('One nibble:', $one_nibble ? 'yes' : 'no');
if ( $args{resume}
@@ -6906,7 +6933,7 @@ sub _get_bounds {
if ( defined $nibble->{lower_boundary}
&& defined $nibble->{upper_boundary} ) {
my $sth = $dbh->prepare($self->{resume_lb_sql});
my @ub = $self->{Quoter}->deserialize_list($nibble->{upper_boundary});
my @ub = split ',', $nibble->{upper_boundary};
PTDEBUG && _d($sth->{Statement}, 'params:', @ub);
$sth->execute(@ub);
$self->{next_lower} = $sth->fetchrow_arrayref();
@@ -7021,7 +7048,7 @@ sub identical_boundaries {
if scalar @$b1 != scalar @$b2; # shouldn't happen
my $n_vals = scalar @$b1;
for my $i ( 0..($n_vals-1) ) {
return 0 if $b1->[$i] ne $b2->[$i]; # diff
return 0 if ($b1->[$i] || '') ne ($b2->[$i] || ''); # diff
}
return 1;
}
@@ -9595,6 +9622,7 @@ use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use Data::Dumper;
use Carp;
$Data::Dumper::Indent = 1;
$Data::Dumper::Sortkeys = 1;
$Data::Dumper::Quotekeys = 0;
@@ -9607,7 +9635,7 @@ sub new {
}
my $self = {
Quoter => $args{Quoter},
Quoter => $args{Quoter},
};
return bless $self, $class;
@@ -9659,11 +9687,21 @@ sub _get_first_values {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $index, $n_index_cols) = @args{@required_args};
my $q = new Quoter();
my $q = $self->{Quoter};
my $index_struct = $tbl->{tbl_struct}->{keys}->{$index};
my $index_cols = $index_struct->{cols};
my $index_columns = join (', ',
my $index_columns;
eval {
$index_columns = join (', ',
map { $q->quote($_) } @{$index_cols}[0..($n_index_cols - 1)]);
};
if ($EVAL_ERROR) {
confess "$EVAL_ERROR";
}
my @where;
foreach my $col ( @{$index_cols}[0..($n_index_cols - 1)] ) {
@@ -9675,19 +9713,21 @@ sub _get_first_values {
. "WHERE " . join(' AND ', @where)
. " ORDER BY $index_columns "
. "LIMIT 1 /*key_len*/"; # only need 1 row
PTDEBUG && _d("_get_first_values: $sql");
PTDEBUG && _d($sql);
my $vals = $cxn->dbh()->selectrow_arrayref($sql);
return $vals;
}
sub _make_range_query {
my ($self, %args) = @_;
my @required_args = qw(tbl index n_index_cols);
my @required_args = qw(tbl index n_index_cols vals);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($tbl, $index, $n_index_cols, $vals) = @args{@required_args};
my $q = new Quoter();
my $q = $self->{Quoter};
my $index_struct = $tbl->{tbl_struct}->{keys}->{$index};
my $index_cols = $index_struct->{cols};
@@ -9695,12 +9735,15 @@ sub _make_range_query {
if ( $n_index_cols > 1 ) {
foreach my $n ( 0..($n_index_cols - 2) ) {
my $col = $index_cols->[$n];
push @where, $q->quote($col) . " = ?";
my $val = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
push @where, $q->quote($col) . " = " . $val;
}
}
my $col = $index_cols->[$n_index_cols - 1];
push @where, $q->quote($col) . " >= ?";
my $val = $vals->[-1]; # should only be as many vals as cols
my $condition = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
push @where, $q->quote($col) . " >= " . $condition;
my $sql = "EXPLAIN SELECT /*!40001 SQL_NO_CACHE */ * "
. "FROM $tbl->{name} FORCE INDEX (" . $q->quote($index) . ") "
bin/pt-table-sync
+10 -6
View File
@@ -6437,6 +6437,7 @@ sub generate_asc_stmt {
cols => \@cols,
quoter => $q,
is_nullable => $tbl_struct->{is_nullable},
type_for => $tbl_struct->{type_for},
);
$asc_stmt->{boundaries}->{$cmp} = $cmp_where->{where};
}
@@ -6457,6 +6458,7 @@ sub generate_cmp_where {
my @slice = @{$args{slice}};
my @cols = @{$args{cols}};
my $is_nullable = $args{is_nullable};
my $type_for = $args{type_for};
my $type = $args{type};
my $q = $self->{Quoter};
@@ -6473,13 +6475,14 @@ sub generate_cmp_where {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @clause, "(($val IS NULL AND $quo IS NULL) OR ($quo = $val))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @clause, "$quo = ?";
push @clause, "$quo = $val";
push @r_slice, $ord;
push @r_scols, $col;
}
@@ -6489,15 +6492,16 @@ sub generate_cmp_where {
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
push @clause, "(? IS NULL OR $quo $type ?)";
push @clause, "($val IS NULL OR $quo $type $val)";
}
elsif ( $type =~ m/>/ ) {
push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
push @clause, "(($val IS NULL AND $quo IS NOT NULL) OR ($quo $cmp $val)";
}
else { # If $type =~ m/</ ) {
push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
push @clauses, "(($val IS NOT NULL AND $quo IS NULL) OR ($quo $cmp $val))";
}
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
@@ -6505,7 +6509,7 @@ sub generate_cmp_where {
else {
push @r_slice, $ord;
push @r_scols, $col;
push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
push @clause, ($type =~ m/=/ && $end ? "$quo $type $val" : "$quo $cmp $val");
}
push @clauses, '(' . join(' AND ', @clause) . ')';
lib/IndexLength.pm
+5 -4
View File
@@ -99,7 +99,7 @@ sub _get_first_values {
}
my ($cxn, $tbl, $index, $n_index_cols) = @args{@required_args};
my $q = $self->{quoter};
my $q = $self->{Quoter};
# Select just the index columns.
my $index_struct = $tbl->{tbl_struct}->{keys}->{$index};
@@ -151,8 +151,8 @@ sub _make_range_query {
# we don't want the last column; that's added below.
foreach my $n ( 0..($n_index_cols - 2) ) {
my $col = $index_cols->[$n];
my $val = $vals->[$n];
push @where, $q->quote($col) . " = ?";
my $val = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
push @where, $q->quote($col) . " = " . $val;
}
}
@@ -160,7 +160,8 @@ sub _make_range_query {
# the N left-most index columns.
my $col = $index_cols->[$n_index_cols - 1];
my $val = $vals->[-1]; # should only be as many vals as cols
push @where, $q->quote($col) . " >= ?";
my $condition = $tbl->{tbl_struct}->{type_for}->{$col} eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
push @where, $q->quote($col) . " >= " . $condition;
my $sql = "EXPLAIN SELECT /*!40001 SQL_NO_CACHE */ * "
. "FROM $tbl->{name} FORCE INDEX (" . $q->quote($index) . ") "
lib/NibbleIterator.pm
+115 -98
View File
@@ -26,7 +26,6 @@ use strict;
use warnings FATAL => 'all';
use English qw(-no_match_vars);
use constant PTDEBUG => $ENV{PTDEBUG} || 0;
use IndexLength;
use Data::Dumper;
$Data::Dumper::Indent = 1;
@@ -90,92 +89,142 @@ sub new {
my @cols = grep { !$ignore_col->{$_} } @$all_cols;
my $self;
if ( $nibble_params->{one_nibble} ) {
my $params = _one_nibble(\%args, \@cols, $where, $tbl, \%comments);
$self = {
%args,
one_nibble => 1,
limit => 0,
nibble_sql => $params->{nibble_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
};
} else {
my $params = _nibble_params($nibble_params, $tbl, \%args, \@cols, $chunk_size, $where, \%comments, $q);
$self = {
%args,
index => $params->{index},
limit => $params->{limit},
first_lb_sql => $params->{first_lb_sql},
last_ub_sql => $params->{last_ub_sql},
ub_sql => $params->{ub_sql},
nibble_sql => $params->{nibble_sql},
explain_first_lb_sql => $params->{explain_first_lb_sql},
explain_ub_sql => $params->{explain_ub_sql},
explain_nibble_sql => $params->{explain_nibble_sql},
resume_lb_sql => $params->{resume_lb_sql},
sql => $params->{sql},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
$self->{nibble_params} = $nibble_params;
$self->{tbl} = $tbl;
$self->{args} = \%args;
$self->{cols} = \@cols;
$self->{chunk_size} = $chunk_size;
$self->{where} = $where;
$self->{comments} = \%comments;
return bless $self, $class;
}
sub switch_to_nibble {
my $self = shift;
my $params = _nibble_params($self->{nibble_params}, $self->{tbl}, $self->{args}, $self->{cols},
$self->{chunk_size}, $self->{where}, $self->{comments}, $self->{Quoter});
$self->{one_nibble} = 0;
$self->{index} = $params->{index};
$self->{limit} = $params->{limit};
$self->{first_lb_sql} = $params->{first_lb_sql};
$self->{last_ub_sql} = $params->{last_ub_sql};
$self->{ub_sql} = $params->{ub_sql};
$self->{nibble_sql} = $params->{nibble_sql};
$self->{explain_first_lb_sql} = $params->{explain_first_lb_sql};
$self->{explain_ub_sql} = $params->{explain_ub_sql};
$self->{explain_nibble_sql} = $params->{explain_nibble_sql};
$self->{resume_lb_sql} = $params->{resume_lb_sql};
$self->{sql} = $params->{sql};
$self->_get_bounds();
$self->_prepare_sths();
}
sub _one_nibble {
my ($args, $cols, $where, $tbl, $comments) = @_;
my $q = new Quoter();
# If the chunk size is >= number of rows in table, then we don't
# need to chunk; we can just select all rows, in order, at once.
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
# : join(', ', map { $q->quote($_) } @$cols))
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ?
"CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{bite}*/";
PTDEBUG && _d('One nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @cols))
. ($args->{select} ? $args->{select}
: join(', ', map{ $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum'
? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_) } @$cols))
. " FROM $tbl->{name}"
. ($where ? " WHERE $where" : '')
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{bite}*/";
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{bite}*/";
PTDEBUG && _d('Explain one nibble statement:', $explain_nibble_sql);
$self = {
%args,
return {
one_nibble => 1,
limit => 0,
nibble_sql => $nibble_sql,
explain_nibble_sql => $explain_nibble_sql,
};
}
else {
}
sub _nibble_params {
my ($nibble_params, $tbl, $args, $cols, $chunk_size, $where, $comments, $q) = @_;
my $index = $nibble_params->{index}; # brevity
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,
my $asc = $args->{TableNibbler}->generate_asc_stmt(
%$args,
tbl_struct => $tbl->{tbl_struct},
index => $index,
n_index_cols => $args{n_chunk_index_cols},
cols => \@cols,
n_index_cols => $args->{n_chunk_index_cols},
cols => $cols,
asc_only => 1,
);
PTDEBUG && _d('Ascend params:', Dumper($asc));
# Check if enum fields items are sorted or not.
# If they are sorted we can skip adding CONCAT to improve the queries eficiency.
my $force_concat_enums = $o->has('force-concat-enums') && $o->get('force-concat-enums');
my $i=0;
for my $index (@{$index_cols}) {
last if $args{n_chunk_index_cols} && $i >= $args{n_chunk_index_cols};
$i++;
if ($tbl->{tbl_struct}->{type_for}->{$index} eq 'enum') {
if ($tbl->{tbl_struct}->{defs}->{$index} =~ m/enum\s*\((.*?)\)/) {
my @items = split(/,\s*/, $1);
my $sorted = 1; # Asume the items list is sorted to later check if this is true
for (my $i=1; $i < scalar(@items); $i++) {
if ($items[$i-1] gt $items[$i]) {
$sorted = 0;
last;
}
}
if (!$force_concat_enums && !$sorted) {
die "The index " . $index . " in table " . $tbl->{name} .
" has unsorted enum items.\nPlease read the documentation for the --force-concat-enums parameter\n";
}
}
}
}
my $force_concat_enums;
# Make SQL statements, prepared on first call to next(). FROM and
# ORDER BY are the same for all statements. FORCE IDNEX and ORDER BY
# are needed to ensure deterministic nibbling.
my $from = "$tbl->{name} FORCE INDEX(`$index`)";
my $order_by = join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by_dec = join(' DESC,', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' && $force_concat_enums
? "CONCAT(".$q->quote($_).")" : $q->quote($_)} @{$index_cols});
my $order_by = join(', ', map {$q->quote($_)} @{$index_cols});
my $order_by_dec = join(' DESC,', map {$q->quote($_)} @{$index_cols});
# The real first row in the table. Usually we start nibbling from
# this row. Called once in _get_bounds().
my $first_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY $order_by"
@@ -186,10 +235,10 @@ sub new {
# If we're resuming, this fetches the effective first row, which
# should differ from the real first row. Called once in _get_bounds().
my $resume_lb_sql;
if ( $args{resume} ) {
if ( $args->{resume} ) {
$resume_lb_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>'}
. ($where ? " AND ($where)" : '')
@@ -204,7 +253,7 @@ sub new {
# upper in some cases. Called once in _get_bounds().
my $last_ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. ($where ? " WHERE $where" : '')
. " ORDER BY "
@@ -223,7 +272,7 @@ sub new {
# for the next nibble. See _next_boundaries().
my $ub_sql
= "SELECT /*!40001 SQL_NO_CACHE */ "
. join(', ', map { $q->quote($_) } @{$asc->{scols}})
. join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{scols}})
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='}
. ($where ? " AND ($where)" : '')
@@ -235,36 +284,36 @@ sub new {
# This statement does the actual nibbling work; its rows are returned
# to the caller via next().
my $nibble_sql
= ($args{dml} ? "$args{dml} " : "SELECT ")
. ($args{select} ? $args{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
= ($args->{dml} ? "$args->{dml} " : "SELECT ")
. ($args->{select} ? $args->{select}
: join(', ', map { $tbl->{tbl_struct}->{type_for}->{$_} eq 'enum' ? "CAST(".$q->quote($_)." AS UNSIGNED)" : $q->quote($_)} @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*$comments->{nibble}*/";
PTDEBUG && _d('Nibble statement:', $nibble_sql);
my $explain_nibble_sql
= "EXPLAIN SELECT "
. ($args{select} ? $args{select}
. ($args->{select} ? $args->{select}
: join(', ', map { $q->quote($_) } @{$asc->{cols}}))
. " FROM $from"
. " WHERE " . $asc->{boundaries}->{'>='} # lower boundary
. " AND " . $asc->{boundaries}->{'<='} # upper boundary
. ($where ? " AND ($where)" : '')
. ($args{order_by} ? " ORDER BY $order_by" : "")
. ($args{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments{nibble}*/";
. ($args->{order_by} ? " ORDER BY $order_by" : "")
. ($args->{lock_in_share_mode} ? " LOCK IN SHARE MODE" : "")
. " /*explain $comments->{nibble}*/";
PTDEBUG && _d('Explain nibble statement:', $explain_nibble_sql);
my $limit = $chunk_size - 1;
PTDEBUG && _d('Initial chunk size (LIMIT):', $limit);
$self = {
%args,
my $params = {
one_nibble => 0,
index => $index,
limit => $limit,
first_lb_sql => $first_lb_sql,
@@ -283,17 +332,7 @@ sub new {
order_by => $order_by,
},
};
}
$self->{row_est} = $nibble_params->{row_est},
$self->{nibbleno} = 0;
$self->{have_rows} = 0;
$self->{rowno} = 0;
$self->{oktonibble} = 1;
$self->{pause_file} = $nibble_params->{pause_file};
$self->{sleep} = $args{sleep} || 60;
return bless $self, $class;
return $params;
}
sub next {
@@ -333,7 +372,6 @@ sub next {
# If there's another nibble, fetch the rows within it.
NIBBLE:
while ( $self->{have_rows} || $self->_next_boundaries() ) {
if ($self->{pause_file}) {
while(-f $self->{pause_file}) {
print "Sleeping $self->{sleep} seconds because $self->{pause_file} exists\n";
@@ -488,11 +526,11 @@ sub row_estimate {
sub can_nibble {
my (%args) = @_;
my @required_args = qw(Cxn tbl chunk_size OptionParser TableParser Quoter);
my @required_args = qw(Cxn tbl chunk_size OptionParser TableParser);
foreach my $arg ( @required_args ) {
die "I need a $arg argument" unless $args{$arg};
}
my ($cxn, $tbl, $chunk_size, $o, $q) = @args{@required_args};
my ($cxn, $tbl, $chunk_size, $o) = @args{@required_args};
my $where = $o->has('where') ? $o->get('where') : '';
@@ -503,23 +541,6 @@ sub can_nibble {
where => $where,
);
my $can_get_keys;
if ($mysql_index) {
my $idx_len = IndexLength->new(Quoter => $q);
my ($key_len, $key) = $idx_len->index_length(
Cxn => $args{Cxn},
tbl => $tbl,
index => $mysql_index,
n_index_cols => $o->get('chunk-index-columns'),
);
if ( !$key || !$key_len || lc($key) ne lc($mysql_index)) {
$can_get_keys = 0;
} else {
$can_get_keys = 1;
}
}
# MySQL's chosen index is only something we should prefer
# if --where is used. Else, we can chose our own index
# and disregard the MySQL index from the row estimate.
@@ -539,10 +560,6 @@ sub can_nibble {
my $one_nibble = !defined $args{one_nibble} || $args{one_nibble}
? $row_est <= $chunk_size * $chunk_size_limit
: 0;
if (!$can_get_keys) {
$one_nibble = 1;
}
PTDEBUG && _d('One nibble:', $one_nibble ? 'yes' : 'no');
# Special case: we're resuming and there's no boundaries, so the table
@@ -922,7 +939,7 @@ sub identical_boundaries {
if scalar @$b1 != scalar @$b2; # shouldn't happen
my $n_vals = scalar @$b1;
for my $i ( 0..($n_vals-1) ) {
return 0 if $b1->[$i] ne $b2->[$i]; # diff
return 0 if ($b1->[$i] || '') ne ($b2->[$i] || ''); # diff
}
return 1;
}
lib/TableNibbler.pm
+10 -6
View File
@@ -125,6 +125,7 @@ sub generate_asc_stmt {
cols => \@cols,
quoter => $q,
is_nullable => $tbl_struct->{is_nullable},
type_for => $tbl_struct->{type_for},
);
$asc_stmt->{boundaries}->{$cmp} = $cmp_where->{where};
}
@@ -152,6 +153,7 @@ sub generate_cmp_where {
my @slice = @{$args{slice}};
my @cols = @{$args{cols}};
my $is_nullable = $args{is_nullable};
my $type_for = $args{type_for};
my $type = $args{type};
my $q = $self->{Quoter};
@@ -169,13 +171,14 @@ sub generate_cmp_where {
my $ord = $slice[$j];
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
push @clause, "((? IS NULL AND $quo IS NULL) OR ($quo = ?))";
push @clause, "(($val IS NULL AND $quo IS NULL) OR ($quo = $val))";
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
}
else {
push @clause, "$quo = ?";
push @clause, "$quo = $val";
push @r_slice, $ord;
push @r_scols, $col;
}
@@ -188,15 +191,16 @@ sub generate_cmp_where {
my $col = $cols[$ord];
my $quo = $q->quote($col);
my $end = $i == $#slice; # Last clause of the whole group.
my $val = ($col && ($type_for->{$col} || '')) eq 'enum' ? "CAST(? AS UNSIGNED)" : "?";
if ( $is_nullable->{$col} ) {
if ( $type =~ m/=/ && $end ) {
push @clause, "(? IS NULL OR $quo $type ?)";
push @clause, "($val IS NULL OR $quo $type $val)";
}
elsif ( $type =~ m/>/ ) {
push @clause, "((? IS NULL AND $quo IS NOT NULL) OR ($quo $cmp ?))";
push @clause, "(($val IS NULL AND $quo IS NOT NULL) OR ($quo $cmp $val)";
}
else { # If $type =~ m/</ ) {
push @clause, "((? IS NOT NULL AND $quo IS NULL) OR ($quo $cmp ?))";
push @clauses, "(($val IS NOT NULL AND $quo IS NULL) OR ($quo $cmp $val))";
}
push @r_slice, $ord, $ord;
push @r_scols, $col, $col;
@@ -204,7 +208,7 @@ sub generate_cmp_where {
else {
push @r_slice, $ord;
push @r_scols, $col;
push @clause, ($type =~ m/=/ && $end ? "$quo $type ?" : "$quo $cmp ?");
push @clause, ($type =~ m/=/ && $end ? "$quo $type $val" : "$quo $cmp $val");
}
# Add the clause to the larger WHERE clause.
t/lib/NibbleIterator.t
+1 -72
View File
@@ -36,11 +36,10 @@ my $dp = new DSNParser(opts=>$dsn_opts);
my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
my $dbh = $sb->get_dbh_for('master');
plan skip_all => 'Cannot connect to sandbox master';
if ( !$dbh ) {
plan skip_all => 'Cannot connect to sandbox master';
} else {
plan tests => 62;
plan tests => 56;
}
my $q = new Quoter();
@@ -897,76 +896,6 @@ is(
"Use non-unique index with highest cardinality (bug 1199591)"
);
$sb->load_file('master', "t/lib/samples/NibbleIterator/enum_keys.sql");
$ni = undef;
eval {
$ni = make_nibble_iter(
db => 'test',
tbl => 't1',
argv => [qw(--databases test --chunk-size 3)],
);
};
like(
$EVAL_ERROR,
qr/The index f3 in table `test`.`t1` has unsorted enum items/,
"PT-1572 Die on unsorted enum items in index",
);
eval {
$ni = make_nibble_iter(
db => 'test',
tbl => 't1',
argv => [qw(--databases test --force-concat-enums --chunk-size 3)],
);
};
like(
$ni->{explain_first_lb_sql},
qr/ORDER BY `f1`, `f2`, CONCAT\(`f3`\)/,
"PT-1572 Use of CONCAT for unsorted ENUM field items without --",
);
eval {
$ni = make_nibble_iter(
db => 'test',
tbl => 't2',
argv => [qw(--databases test --chunk-size 3)],
);
};
is(
$EVAL_ERROR,
'',
"PT-1572 No errors on sorted enum items in index",
);
like(
$ni->{explain_first_lb_sql},
qr/ORDER BY `f1`, `f2`, `f3`/,
"PT-1572 Don't use CONCAT for sorted ENUM field items without --force-concat-enums",
);
eval {
$ni = make_nibble_iter(
db => 'test',
tbl => 't1',
argv => [qw(--databases test --chunk-size 3 --chunk-index-columns 2)],
);
};
is(
$EVAL_ERROR,
'',
"PT-1572 No errors on unsorted enum items in index and --chunk-index-columns",
);
like(
$ni->{explain_first_lb_sql},
qr/ORDER BY `f1`, `f2`, `f3`/,
"PT-1572 Don't use CONCAT for sorted ENUM field items without --force-concat-enums & --chunk-index-columns",
);
# #############################################################################
# Done.
# #############################################################################
t/lib/TableNibbler.t
+105 -59
View File
@@ -417,17 +417,24 @@ is_deeply(
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND (? IS NULL OR `customer_id` >= ?)))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
. 'OR (`customer_id` > ?))))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND `inventory_id` > ?) OR '
. '(`rental_date` = ? AND `inventory_id` = ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
. 'OR (`customer_id` > ?)))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND '
#. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
# . '= ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
# . 'OR (`customer_id` > ?))))',
'<=' => '((`rental_date` < ?) OR (`rental_date` = ? AND '
. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND (? IS NULL OR `customer_id` <= ?)))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND '
. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND ((? IS NOT NULL AND `customer_id` IS NULL) '
. 'OR (`customer_id` < ?))))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND `inventory_id` < ?) OR '
. '((? IS NOT NULL AND `customer_id` IS NULL) OR (`customer_id` < ?)) OR '
. '(`rental_date` = ? AND `inventory_id` = ?))',
# '((`rental_date` < ?) OR (`rental_date` = ? AND '
#. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
#. '= ? AND ((? IS NOT NULL AND `customer_id` IS NULL) '
#. 'OR (`customer_id` < ?))))',
},
},
'Alternate index on sakila.rental with nullable customer_id',
@@ -460,26 +467,35 @@ is_deeply(
cols => [qw(rental_id rental_date inventory_id customer_id
return_date staff_id last_update)],
index => 'rental_date',
where => '((`rental_date` > ?) OR (`rental_date` = ? AND `inventory_id` > ?)'
. ' OR (`rental_date` = ? AND `inventory_id` = ? AND '
. '((? IS NULL AND `customer_id` IS NOT NULL) OR (`customer_id` > ?))))',
where => '((`rental_date` > ?) OR (`rental_date` = ? AND `inventory_id` > ?) OR '
. '(`rental_date` = ? AND `inventory_id` = ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
. 'OR (`customer_id` > ?)))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND `inventory_id` > ?)'
#. ' OR (`rental_date` = ? AND `inventory_id` = ? AND '
#. '((? IS NULL AND `customer_id` IS NOT NULL) OR (`customer_id` > ?))))',
slice => [1, 1, 2, 1, 2, 3, 3],
scols => [qw(rental_date rental_date inventory_id rental_date inventory_id customer_id customer_id)],
boundaries => {
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND (? IS NULL OR `customer_id` >= ?)))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
. 'OR (`customer_id` > ?))))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND `inventory_id` > ?) OR '
. '(`rental_date` = ? AND `inventory_id` = ? AND ((? IS NULL AND `customer_id` IS NOT NULL) OR '
. '(`customer_id` > ?)))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND '
#. '`inventory_id` > ?) OR (`rental_date` = ? AND `inventory_id` '
#. '= ? AND ((? IS NULL AND `customer_id` IS NOT NULL) '
#. 'OR (`customer_id` > ?))))',
'<=' => '((`rental_date` < ?) OR (`rental_date` = ? AND '
. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND (? IS NULL OR `customer_id` <= ?)))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND '
. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
. '= ? AND ((? IS NOT NULL AND `customer_id` IS NULL) '
. 'OR (`customer_id` < ?))))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND `inventory_id` < ?) OR '
. '((? IS NOT NULL AND `customer_id` IS NULL) OR (`customer_id` < ?)) OR '
. '(`rental_date` = ? AND `inventory_id` = ?))',
# '((`rental_date` < ?) OR (`rental_date` = ? AND '
#. '`inventory_id` < ?) OR (`rental_date` = ? AND `inventory_id` '
#. '= ? AND ((? IS NOT NULL AND `customer_id` IS NULL) '
#. 'OR (`customer_id` < ?))))',
},
},
'Alternate index on sakila.rental with nullable customer_id and strict ascending',
@@ -500,30 +516,45 @@ is_deeply(
cols => [qw(rental_id rental_date inventory_id customer_id
return_date staff_id last_update)],
index => 'rental_date',
where => '((`rental_date` > ?) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?)))'
. ' OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
where => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL)'
. ' OR (`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` >= ?))',
# '((`rental_date` > ?) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?)))'
#. ' OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
slice => [1, 1, 2, 2, 1, 2, 2, 3],
scols => [qw(rental_date rental_date inventory_id inventory_id
rental_date inventory_id inventory_id customer_id)],
boundaries => {
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` '
. '> ?))) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` '
. 'IS NULL) OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL '
. 'AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?))) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
'<=' => '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` <= ?))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) '
. 'OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS '
. 'NULL) OR (`inventory_id` = ?)) AND `customer_id` < ?))',
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR '
. '(`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` >= ?))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND '
#. '((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` '
#. '> ?))) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` '
#. 'IS NULL) OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR '
. '(`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` > ?))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL '
#. 'AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?))) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
'<=' => '((`rental_date` < ?) OR ((? IS NOT NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?)) '
. 'OR (`rental_date` = ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` <= ?))',
# '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
#. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` <= ?))',
'<' => '((`rental_date` < ?) OR ((? IS NOT NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?)) '
. 'OR (`rental_date` = ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` < ?))',
# '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
#. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) '
#. 'OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS '
#. 'NULL) OR (`inventory_id` = ?)) AND `customer_id` < ?))',
},
},
'Alternate index on sakila.rental with nullable inventory_id',
@@ -540,30 +571,45 @@ is_deeply(
cols => [qw(rental_id rental_date inventory_id customer_id
return_date staff_id last_update)],
index => 'rental_date',
where => '((`rental_date` > ?) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?)))'
. ' OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
where => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR '
. '(`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` > ?))',
# '((`rental_date` > ?) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?)))'
#. ' OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
slice => [1, 1, 2, 2, 1, 2, 2, 3],
scols => [qw(rental_date rental_date inventory_id inventory_id
rental_date inventory_id inventory_id customer_id)],
boundaries => {
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND '
. '((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` '
. '> ?))) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` '
. 'IS NULL) OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL '
. 'AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?))) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
'<=' => '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) OR '
. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
. 'OR (`inventory_id` = ?)) AND `customer_id` <= ?))',
'<' => '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) '
. 'OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS '
. 'NULL) OR (`inventory_id` = ?)) AND `customer_id` < ?))',
'>=' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR '
. '(`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` >= ?))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND '
#. '((? IS NULL AND `inventory_id` IS NOT NULL) OR (`inventory_id` '
#. '> ?))) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` '
#. 'IS NULL) OR (`inventory_id` = ?)) AND `customer_id` >= ?))',
'>' => '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NOT NULL) OR '
. '(`inventory_id` > ?)) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` > ?))',
# '((`rental_date` > ?) OR (`rental_date` = ? AND ((? IS NULL '
#. 'AND `inventory_id` IS NOT NULL) OR (`inventory_id` > ?))) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` > ?))',
'<=' => '((`rental_date` < ?) OR ((? IS NOT NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?)) OR '
. '(`rental_date` = ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` <= ?))',
# '((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
#. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) OR '
#. '(`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) '
#. 'OR (`inventory_id` = ?)) AND `customer_id` <= ?))',
'<' => '((`rental_date` < ?) OR ((? IS NOT NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?)) OR '
. '(`rental_date` = ?) OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS NULL) OR '
. '(`inventory_id` = ?)) AND `customer_id` < ?))',
#'((`rental_date` < ?) OR (`rental_date` = ? AND ((? IS NOT '
#. 'NULL AND `inventory_id` IS NULL) OR (`inventory_id` < ?))) '
#. 'OR (`rental_date` = ? AND ((? IS NULL AND `inventory_id` IS '
#. 'NULL) OR (`inventory_id` = ?)) AND `customer_id` < ?))',
},
},
'Alternate index on sakila.rental with nullable inventory_id and strict ascending',
t/pt-online-schema-change/basics.t
+5 -1
View File
@@ -177,12 +177,14 @@ sub test_alter_table {
) or $fail = 1;
# Rows in the original and new table should be identical.
my $query = "SELECT $cols FROM $table ORDER BY `$pk_col`";
my $new_rows = $master_dbh->selectall_arrayref("SELECT $cols FROM $table ORDER BY `$pk_col`");
my $should_diag;
is_deeply(
$new_rows,
$orig_rows,
"$name rows"
) or $fail = 1;
) or $fail=1;
if ( grep { $_ eq '--preserve-triggers' } @$cmds ) {
my $new_triggers = $master_dbh->selectall_arrayref($triggers_sql);
@@ -341,6 +343,8 @@ sub test_alter_table {
my $db_flavor = VersionParser->new($master_dbh)->flavor();
if ( $db_flavor =~ m/XtraDB Cluster/ ) {
diag('====================================================================================================');
diag($db_flavor);
test_alter_table(
name => "Basic no fks --dry-run",
table => "pt_osc.t",
t/pt-online-schema-change/long_fk_constraints.t
+4 -3
View File
@@ -50,8 +50,6 @@ $sb->load_file('master', "$sample/long_fk_constraints.sql");
qw(--execute)) },
);
warn $output;
my $query = <<_SQL;
SELECT TABLE_NAME, CONSTRAINT_NAME
FROM information_schema.KEY_COLUMN_USAGE
@@ -62,8 +60,11 @@ ORDER BY TABLE_NAME, CONSTRAINT_NAME
_SQL
my $constraints = $master_dbh->selectall_arrayref($query);
# why we need to sort? Depending on the MySQL version and the characters set, the ORDER BY clause
# in the query will return different values so, it is better to rely on our own sorted results.
my @sorted_constraints = sort { @$a[0].@$a[1] cmp @$b[0].@$b[1] } @$constraints;
is_deeply(
$constraints,
\@sorted_constraints,
[
[ 'Table1', '__fkey1a' ],
[ 'Table1', '__fkey_SALES_RECURRING_PROFILE_CUSTOMER_CUSTOMER_ENTITY_ENTITY_I' ],
t/pt-online-schema-change/pt_1757.t
+96
View File
@@ -0,0 +1,96 @@
#!/usr/bin/env perl
BEGIN {
die "The PERCONA_TOOLKIT_BRANCH environment variable is not set.\n"
unless $ENV{PERCONA_TOOLKIT_BRANCH} && -d $ENV{PERCONA_TOOLKIT_BRANCH};
unshift @INC, "$ENV{PERCONA_TOOLKIT_BRANCH}/lib";
};
use strict;
use warnings FATAL => 'all';
use threads;
use threads::shared;
use Thread::Semaphore;
use English qw(-no_match_vars);
use Test::More;
use Data::Dumper;
use PerconaTest;
use Sandbox;
use SqlModes;
use File::Temp qw/ tempdir /;
if ($sandbox_version lt '5.7') {
plan skip_all => 'This test needs MySQL 5.7+';
} else {
plan tests => 3;
}
require "$trunk/bin/pt-online-schema-change";
my $dp = new DSNParser(opts=>$dsn_opts);
my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
my $dbh = $sb->get_dbh_for('master');
my $dsn = $sb->dsn_for("master");
my $slave_dbh = $sb->get_dbh_for('slave1');
# The sandbox servers run with lock_wait_timeout=3 and it's not dynamic
# so we need to specify --set-vars innodb_lock_wait_timeout=3 else the
# tool will die.
my @args = (qw(--set-vars innodb_lock_wait_timeout=3));
my $output;
my $exit_status;
my $master_port = $sb->port_for('master');
my $num_rows = 5000;
$sb->load_file('master', "t/pt-online-schema-change/samples/pt-1757.sql");
diag(`util/mysql_random_data_load --host=127.0.0.1 --port=$master_port --user=msandbox --password=msandbox test t1 $num_rows`);
# Let's alter the stats to force this scenario:
# 1) On master, we are going to put 100 as the number of rows in the table. This will make osc to try to run in one chunk
# 2) On the slave, we are going to put the real number of rows. This will cause a fallback to nibble and pt-osc should call
# NibbleIterator->switch_to_nibble()
$dbh->do('SET @@SESSION.sql_log_bin=0');
$dbh->do('update mysql.innodb_table_stats set n_rows=100 where table_name="t1"');
$dbh->do('update mysql.innodb_index_stats set stat_value=100 where stat_description in("id") and table_name="t1"');
$dbh->do('SET @@SESSION.sql_log_bin=1');
$slave_dbh->do('SET @@SESSION.sql_log_bin=0');
$slave_dbh->do("update mysql.innodb_table_stats set n_rows=$num_rows where table_name='t1'");
$slave_dbh->do("update mysql.innodb_index_stats set stat_value=$num_rows where stat_description in('id') and table_name='t1'");
$slave_dbh->do('SET @@SESSION.sql_log_bin=1');
($output, $exit_status) = full_output(
sub { pt_online_schema_change::main(@args, "$dsn,D=test,t=t1",
'--execute', '--alter', "ADD COLUMN new_col INT NOT NULL DEFAULT 1",
'--chunk-size', '25',
),
},
stderr => 1,
);
is(
$exit_status,
0,
"Altered OK status",
);
# The WHERE clause here is important as a double check that the table was altered and new_col exists
my $rows = $dbh->selectrow_arrayref("SELECT COUNT(*) FROM test.t1 WHERE new_col = 1");
is(
$rows->[0],
$num_rows,
"Correct rows count"
) or diag(Dumper($rows));
# #############################################################################
# Done.
# #############################################################################
$sb->wipe_clean($dbh);
ok($sb->ok(), "Sandbox servers") or BAIL_OUT(__FILE__ . " broke the sandbox");
done_testing;
t/pt-online-schema-change/rename_fk_constraints.t
+9 -9
View File
@@ -22,7 +22,6 @@ my $sb = new Sandbox(basedir => '/tmp', DSNParser => $dp);
my $master_dbh = $sb->get_dbh_for('master');
my $vp = VersionParser->new($master_dbh);
warn Data::Dumper::Dumper($vp);
if ($vp->cmp('8.0.14') > -1 && $vp->flavor() !~ m/maria/i) {
plan skip_all => 'Cannot run this test under the current MySQL version';
@@ -65,11 +64,14 @@ my $query = <<__SQL;
and CONSTRAINT_NAME LIKE '%fkey%'
ORDER BY TABLE_NAME, CONSTRAINT_NAME
__SQL
my $constraints = $master_dbh->selectall_arrayref($query);
my $constraints = $master_dbh->selectall_arrayref($query);
# why we need to sort? Depending on the MySQL version and the characters set, the ORDER BY clause
# in the query will return different values so, it is better to rely on our own sorted results.
my @sorted_constraints = sort { @$a[0].@$a[1] cmp @$b[0].@$b[1] } @$constraints;
is_deeply(
$constraints,
\@sorted_constraints,
[
['Table1', '__fkey1a'],
['Table1', '_fkey1b'],
@@ -79,8 +81,6 @@ is_deeply(
"First run adds or removes underscore from constraint names, accordingly"
);
# run second time: we expect constraint names to be prefixed with one underscore
# if they havre't one, and to remove 2 if they have 2
($output, $exit_status) = full_output(
@@ -91,10 +91,10 @@ is_deeply(
);
$constraints = $master_dbh->selectall_arrayref("SELECT TABLE_NAME, CONSTRAINT_NAME FROM information_schema.KEY_COLUMN_USAGE WHERE table_schema='bug1215587' and (TABLE_NAME='Table1' OR TABLE_NAME='Table2') and CONSTRAINT_NAME LIKE '%fkey%' ORDER BY TABLE_NAME, CONSTRAINT_NAME");
@sorted_constraints = sort { @$a[0].@$a[1] cmp @$b[0].@$b[1] } @$constraints; # read above why we need to sort
is_deeply(
$constraints,
\@sorted_constraints,
[
['Table1', '__fkey1b'],
['Table1', 'fkey1a'],
@@ -113,10 +113,10 @@ is_deeply(
);
$constraints = $master_dbh->selectall_arrayref("SELECT TABLE_NAME, CONSTRAINT_NAME FROM information_schema.KEY_COLUMN_USAGE WHERE table_schema='bug1215587' and (TABLE_NAME='Table1' OR TABLE_NAME='Table2') and CONSTRAINT_NAME LIKE '%fkey%' ORDER BY TABLE_NAME, CONSTRAINT_NAME");
@sorted_constraints = sort { @$a[0].@$a[1] cmp @$b[0].@$b[1] } @$constraints; # read above why we need to sort
is_deeply(
$constraints,
\@sorted_constraints,
[
['Table1', '_fkey1a'],
['Table1', 'fkey1b'],
t/pt-online-schema-change/samples/pt-1757.sql
+15
View File
@@ -0,0 +1,15 @@
DROP DATABASE IF EXISTS `test`;
CREATE DATABASE `test`;
CREATE TABLE `test`.`t1` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`f2` char(32) DEFAULT NULL,
`f3` ENUM('red','green','blue'),
PRIMARY KEY (`id`),
UNIQUE KEY `c` (`f2`)
) ENGINE=InnoDB
AUTO_INCREMENT=0
DEFAULT CHARSET=utf8
STATS_AUTO_RECALC = 0
STATS_PERSISTENT = 1;
t/pt-online-schema-change/sanity_checks.t
+1 -1
View File
@@ -116,7 +116,7 @@ like(
$output,
qr/Failed to find a unique new table name/,
"Doesn't try forever to find a new table name"
);
) or diag($output);
# #############################################################################
# Done.
util/mysql_random_data_load
BIN
View File
Binary file not shown.