1 package DBIx::Class::Storage::DBI;
2 # -*- mode: cperl; cperl-indent-level: 2 -*-
7 use base 'DBIx::Class::Storage';
10 use Carp::Clan qw/^DBIx::Class/;
12 use DBIx::Class::Storage::DBI::Cursor;
13 use DBIx::Class::Storage::Statistics;
17 __PACKAGE__->mk_group_accessors('simple' =>
18 qw/_connect_info _dbi_connect_info _dbh _sql_maker _sql_maker_opts _conn_pid
19 _conn_tid transaction_depth _dbh_autocommit _driver_determined savepoints/
22 # the values for these accessors are picked out (and deleted) from
23 # the attribute hashref passed to connect_info
24 my @storage_options = qw/
25 on_connect_call on_disconnect_call on_connect_do on_disconnect_do
26 disable_sth_caching unsafe auto_savepoint
28 __PACKAGE__->mk_group_accessors('simple' => @storage_options);
31 # default cursor class, overridable in connect_info attributes
32 __PACKAGE__->cursor_class('DBIx::Class::Storage::DBI::Cursor');
34 __PACKAGE__->mk_group_accessors('inherited' => qw/sql_maker_class/);
35 __PACKAGE__->sql_maker_class('DBIx::Class::SQLAHacks');
40 DBIx::Class::Storage::DBI - DBI storage handler
44 my $schema = MySchema->connect('dbi:SQLite:my.db');
46 $schema->storage->debug(1);
48 my @stuff = $schema->storage->dbh_do(
50 my ($storage, $dbh, @args) = @_;
51 $dbh->do("DROP TABLE authors");
56 $schema->resultset('Book')->search({
57 written_on => $schema->storage->datetime_parser(DateTime->now)
62 This class represents the connection to an RDBMS via L<DBI>. See
63 L<DBIx::Class::Storage> for general information. This pod only
64 documents DBI-specific methods and behaviors.
71 my $new = shift->next::method(@_);
73 $new->transaction_depth(0);
74 $new->_sql_maker_opts({});
75 $new->{savepoints} = [];
76 $new->{_in_dbh_do} = 0;
84 This method is normally called by L<DBIx::Class::Schema/connection>, which
85 encapsulates its argument list in an arrayref before passing them here.
87 The argument list may contain:
93 The same 4-element argument set one would normally pass to
94 L<DBI/connect>, optionally followed by
95 L<extra attributes|/DBIx::Class specific connection attributes>
96 recognized by DBIx::Class:
98 $connect_info_args = [ $dsn, $user, $password, \%dbi_attributes?, \%extra_attributes? ];
102 A single code reference which returns a connected
103 L<DBI database handle|DBI/connect> optionally followed by
104 L<extra attributes|/DBIx::Class specific connection attributes> recognized
107 $connect_info_args = [ sub { DBI->connect (...) }, \%extra_attributes? ];
111 A single hashref with all the attributes and the dsn/user/password
114 $connect_info_args = [{
122 $connect_info_args = [{
123 dbh_maker => sub { DBI->connect (...) },
128 This is particularly useful for L<Catalyst> based applications, allowing the
129 following config (L<Config::General> style):
134 dsn dbi:mysql:database=test
141 The C<dsn>/C<user>/C<password> combination can be substituted by the
142 C<dbh_maker> key whose value is a coderef that returns a connected
143 L<DBI database handle|DBI/connect>
147 Please note that the L<DBI> docs recommend that you always explicitly
148 set C<AutoCommit> to either I<0> or I<1>. L<DBIx::Class> further
149 recommends that it be set to I<1>, and that you perform transactions
150 via our L<DBIx::Class::Schema/txn_do> method. L<DBIx::Class> will set it
151 to I<1> if you do not do explicitly set it to zero. This is the default
152 for most DBDs. See L</DBIx::Class and AutoCommit> for details.
154 =head3 DBIx::Class specific connection attributes
156 In addition to the standard L<DBI|DBI/ATTRIBUTES_COMMON_TO_ALL_HANDLES>
157 L<connection|DBI/Database_Handle_Attributes> attributes, DBIx::Class recognizes
158 the following connection options. These options can be mixed in with your other
159 L<DBI> connection attributes, or placed in a seperate hashref
160 (C<\%extra_attributes>) as shown above.
162 Every time C<connect_info> is invoked, any previous settings for
163 these options will be cleared before setting the new ones, regardless of
164 whether any options are specified in the new C<connect_info>.
171 Specifies things to do immediately after connecting or re-connecting to
172 the database. Its value may contain:
178 This contains one SQL statement to execute.
180 =item an array reference
182 This contains SQL statements to execute in order. Each element contains
183 a string or a code reference that returns a string.
185 =item a code reference
187 This contains some code to execute. Unlike code references within an
188 array reference, its return value is ignored.
192 =item on_disconnect_do
194 Takes arguments in the same form as L</on_connect_do> and executes them
195 immediately before disconnecting from the database.
197 Note, this only runs if you explicitly call L</disconnect> on the
200 =item on_connect_call
202 A more generalized form of L</on_connect_do> that calls the specified
203 C<connect_call_METHOD> methods in your storage driver.
205 on_connect_do => 'select 1'
209 on_connect_call => [ [ do_sql => 'select 1' ] ]
211 Its values may contain:
217 Will call the C<connect_call_METHOD> method.
219 =item a code reference
221 Will execute C<< $code->($storage) >>
223 =item an array reference
225 Each value can be a method name or code reference.
227 =item an array of arrays
229 For each array, the first item is taken to be the C<connect_call_> method name
230 or code reference, and the rest are parameters to it.
234 Some predefined storage methods you may use:
240 Executes a SQL string or a code reference that returns a SQL string. This is
241 what L</on_connect_do> and L</on_disconnect_do> use.
249 Will execute the scalar as SQL.
253 Taken to be arguments to L<DBI/do>, the SQL string optionally followed by the
254 attributes hashref and bind values.
256 =item a code reference
258 Will execute C<< $code->($storage) >> and execute the return array refs as
265 Execute any statements necessary to initialize the database session to return
266 and accept datetime/timestamp values used with
267 L<DBIx::Class::InflateColumn::DateTime>.
269 Only necessary for some databases, see your specific storage driver for
270 implementation details.
274 =item on_disconnect_call
276 Takes arguments in the same form as L</on_connect_call> and executes them
277 immediately before disconnecting from the database.
279 Calls the C<disconnect_call_METHOD> methods as opposed to the
280 C<connect_call_METHOD> methods called by L</on_connect_call>.
282 Note, this only runs if you explicitly call L</disconnect> on the
285 =item disable_sth_caching
287 If set to a true value, this option will disable the caching of
288 statement handles via L<DBI/prepare_cached>.
292 Sets the limit dialect. This is useful for JDBC-bridge among others
293 where the remote SQL-dialect cannot be determined by the name of the
294 driver alone. See also L<SQL::Abstract::Limit>.
298 Specifies what characters to use to quote table and column names. If
299 you use this you will want to specify L</name_sep> as well.
301 C<quote_char> expects either a single character, in which case is it
302 is placed on either side of the table/column name, or an arrayref of length
303 2 in which case the table/column name is placed between the elements.
305 For example under MySQL you should use C<< quote_char => '`' >>, and for
306 SQL Server you should use C<< quote_char => [qw/[ ]/] >>.
310 This only needs to be used in conjunction with C<quote_char>, and is used to
311 specify the charecter that seperates elements (schemas, tables, columns) from
312 each other. In most cases this is simply a C<.>.
314 The consequences of not supplying this value is that L<SQL::Abstract>
315 will assume DBIx::Class' uses of aliases to be complete column
316 names. The output will look like I<"me.name"> when it should actually
321 This Storage driver normally installs its own C<HandleError>, sets
322 C<RaiseError> and C<ShowErrorStatement> on, and sets C<PrintError> off on
323 all database handles, including those supplied by a coderef. It does this
324 so that it can have consistent and useful error behavior.
326 If you set this option to a true value, Storage will not do its usual
327 modifications to the database handle's attributes, and instead relies on
328 the settings in your connect_info DBI options (or the values you set in
329 your connection coderef, in the case that you are connecting via coderef).
331 Note that your custom settings can cause Storage to malfunction,
332 especially if you set a C<HandleError> handler that suppresses exceptions
333 and/or disable C<RaiseError>.
337 If this option is true, L<DBIx::Class> will use savepoints when nesting
338 transactions, making it possible to recover from failure in the inner
339 transaction without having to abort all outer transactions.
343 Use this argument to supply a cursor class other than the default
344 L<DBIx::Class::Storage::DBI::Cursor>.
348 Some real-life examples of arguments to L</connect_info> and
349 L<DBIx::Class::Schema/connect>
351 # Simple SQLite connection
352 ->connect_info([ 'dbi:SQLite:./foo.db' ]);
355 ->connect_info([ sub { DBI->connect(...) } ]);
357 # Connect via subref in hashref
359 dbh_maker => sub { DBI->connect(...) },
360 on_connect_do => 'alter session ...',
363 # A bit more complicated
370 { quote_char => q{"}, name_sep => q{.} },
374 # Equivalent to the previous example
380 { AutoCommit => 1, quote_char => q{"}, name_sep => q{.} },
384 # Same, but with hashref as argument
385 # See parse_connect_info for explanation
388 dsn => 'dbi:Pg:dbname=foo',
390 password => 'my_pg_password',
397 # Subref + DBIx::Class-specific connection options
400 sub { DBI->connect(...) },
404 on_connect_do => ['SET search_path TO myschema,otherschema,public'],
405 disable_sth_caching => 1,
415 my ($self, $info_arg) = @_;
417 return $self->_connect_info if !$info_arg;
419 my @args = @$info_arg; # take a shallow copy for further mutilation
420 $self->_connect_info([@args]); # copy for _connect_info
423 # combine/pre-parse arguments depending on invocation style
426 if (ref $args[0] eq 'CODE') { # coderef with optional \%extra_attributes
427 %attrs = %{ $args[1] || {} };
430 elsif (ref $args[0] eq 'HASH') { # single hashref (i.e. Catalyst config)
431 %attrs = %{$args[0]};
433 if (my $code = delete $attrs{dbh_maker}) {
436 my @ignored = grep { delete $attrs{$_} } (qw/dsn user password/);
439 'Attribute(s) %s in connect_info were ignored, as they can not be applied '
440 . "to the result of 'dbh_maker'",
442 join (', ', map { "'$_'" } (@ignored) ),
447 @args = delete @attrs{qw/dsn user password/};
450 else { # otherwise assume dsn/user/password + \%attrs + \%extra_attrs
452 % { $args[3] || {} },
453 % { $args[4] || {} },
455 @args = @args[0,1,2];
458 # Kill sql_maker/_sql_maker_opts, so we get a fresh one with only
459 # the new set of options
460 $self->_sql_maker(undef);
461 $self->_sql_maker_opts({});
464 for my $storage_opt (@storage_options, 'cursor_class') { # @storage_options is declared at the top of the module
465 if(my $value = delete $attrs{$storage_opt}) {
466 $self->$storage_opt($value);
469 for my $sql_maker_opt (qw/limit_dialect quote_char name_sep/) {
470 if(my $opt_val = delete $attrs{$sql_maker_opt}) {
471 $self->_sql_maker_opts->{$sql_maker_opt} = $opt_val;
476 if (ref $args[0] eq 'CODE') {
477 # _connect() never looks past $args[0] in this case
481 %{ $self->_default_dbi_connect_attributes || {} },
486 $self->_dbi_connect_info([@args, keys %attrs ? \%attrs : ()]);
487 $self->_connect_info;
490 sub _default_dbi_connect_attributes {
500 This method is deprecated in favour of setting via L</connect_info>.
504 =head2 on_disconnect_do
506 This method is deprecated in favour of setting via L</connect_info>.
510 sub _parse_connect_do {
511 my ($self, $type) = @_;
513 my $val = $self->$type;
514 return () if not defined $val;
519 push @res, [ 'do_sql', $val ];
520 } elsif (ref($val) eq 'CODE') {
522 } elsif (ref($val) eq 'ARRAY') {
523 push @res, map { [ 'do_sql', $_ ] } @$val;
525 $self->throw_exception("Invalid type for $type: ".ref($val));
533 Arguments: ($subref | $method_name), @extra_coderef_args?
535 Execute the given $subref or $method_name using the new exception-based
536 connection management.
538 The first two arguments will be the storage object that C<dbh_do> was called
539 on and a database handle to use. Any additional arguments will be passed
540 verbatim to the called subref as arguments 2 and onwards.
542 Using this (instead of $self->_dbh or $self->dbh) ensures correct
543 exception handling and reconnection (or failover in future subclasses).
545 Your subref should have no side-effects outside of the database, as
546 there is the potential for your subref to be partially double-executed
547 if the database connection was stale/dysfunctional.
551 my @stuff = $schema->storage->dbh_do(
553 my ($storage, $dbh, @cols) = @_;
554 my $cols = join(q{, }, @cols);
555 $dbh->selectrow_array("SELECT $cols FROM foo");
566 my $dbh = $self->_get_dbh;
568 return $self->$code($dbh, @_) if $self->{_in_dbh_do}
569 || $self->{transaction_depth};
571 local $self->{_in_dbh_do} = 1;
574 my $want_array = wantarray;
579 @result = $self->$code($dbh, @_);
581 elsif(defined $want_array) {
582 $result[0] = $self->$code($dbh, @_);
585 $self->$code($dbh, @_);
589 # ->connected might unset $@ - copy
591 if(!$exception) { return $want_array ? @result : $result[0] }
593 $self->throw_exception($exception) if $self->connected;
595 # We were not connected - reconnect and retry, but let any
596 # exception fall right through this time
597 carp "Retrying $code after catching disconnected exception: $exception"
598 if $ENV{DBIC_DBIRETRY_DEBUG};
599 $self->_populate_dbh;
600 $self->$code($self->_dbh, @_);
603 # This is basically a blend of dbh_do above and DBIx::Class::Storage::txn_do.
604 # It also informs dbh_do to bypass itself while under the direction of txn_do,
605 # via $self->{_in_dbh_do} (this saves some redundant eval and errorcheck, etc)
610 ref $coderef eq 'CODE' or $self->throw_exception
611 ('$coderef must be a CODE reference');
613 return $coderef->(@_) if $self->{transaction_depth} && ! $self->auto_savepoint;
615 local $self->{_in_dbh_do} = 1;
618 my $want_array = wantarray;
627 @result = $coderef->(@_);
629 elsif(defined $want_array) {
630 $result[0] = $coderef->(@_);
638 # ->connected might unset $@ - copy
640 if(!$exception) { return $want_array ? @result : $result[0] }
642 if($tried++ || $self->connected) {
643 eval { $self->txn_rollback };
644 my $rollback_exception = $@;
645 if($rollback_exception) {
646 my $exception_class = "DBIx::Class::Storage::NESTED_ROLLBACK_EXCEPTION";
647 $self->throw_exception($exception) # propagate nested rollback
648 if $rollback_exception =~ /$exception_class/;
650 $self->throw_exception(
651 "Transaction aborted: ${exception}. "
652 . "Rollback failed: ${rollback_exception}"
655 $self->throw_exception($exception)
658 # We were not connected, and was first try - reconnect and retry
660 carp "Retrying $coderef after catching disconnected exception: $exception"
661 if $ENV{DBIC_DBIRETRY_DEBUG};
662 $self->_populate_dbh;
668 Our C<disconnect> method also performs a rollback first if the
669 database is not in C<AutoCommit> mode.
679 push @actions, ( $self->on_disconnect_call || () );
680 push @actions, $self->_parse_connect_do ('on_disconnect_do');
682 $self->_do_connection_actions(disconnect_call_ => $_) for @actions;
684 $self->_dbh_rollback unless $self->_dbh_autocommit;
686 $self->_dbh->disconnect;
692 =head2 with_deferred_fk_checks
696 =item Arguments: C<$coderef>
698 =item Return Value: The return value of $coderef
702 Storage specific method to run the code ref with FK checks deferred or
703 in MySQL's case disabled entirely.
707 # Storage subclasses should override this
708 sub with_deferred_fk_checks {
709 my ($self, $sub) = @_;
718 =item Arguments: none
720 =item Return Value: 1|0
724 Verifies that the the current database handle is active and ready to execute
725 an SQL statement (i.e. the connection did not get stale, server is still
726 answering, etc.) This method is used internally by L</dbh>.
732 return 0 unless $self->_seems_connected;
735 local $self->_dbh->{RaiseError} = 1;
740 sub _seems_connected {
743 my $dbh = $self->_dbh
746 if(defined $self->_conn_tid && $self->_conn_tid != threads->tid) {
753 return 0 if !$self->_dbh;
756 return $dbh->FETCH('Active');
762 my $dbh = $self->_dbh or return 0;
767 # handle pid changes correctly
768 # NOTE: assumes $self->_dbh is a valid $dbh
772 return if defined $self->_conn_pid && $self->_conn_pid == $$;
774 $self->_dbh->{InactiveDestroy} = 1;
781 sub ensure_connected {
784 unless ($self->connected) {
785 $self->_populate_dbh;
791 Returns a C<$dbh> - a data base handle of class L<DBI>. The returned handle
792 is guaranteed to be healthy by implicitly calling L</connected>, and if
793 necessary performing a reconnection before returning. Keep in mind that this
794 is very B<expensive> on some database engines. Consider using L<dbh_do>
802 if (not $self->_dbh) {
803 $self->_populate_dbh;
805 $self->ensure_connected;
810 # this is the internal "get dbh or connect (don't check)" method
813 $self->_verify_pid if $self->_dbh;
814 $self->_populate_dbh unless $self->_dbh;
818 sub _sql_maker_args {
823 array_datatypes => 1,
824 limit_dialect => $self->_get_dbh,
825 %{$self->_sql_maker_opts}
831 unless ($self->_sql_maker) {
832 my $sql_maker_class = $self->sql_maker_class;
833 $self->ensure_class_loaded ($sql_maker_class);
834 $self->_sql_maker($sql_maker_class->new( $self->_sql_maker_args ));
836 return $self->_sql_maker;
839 # nothing to do by default
846 my @info = @{$self->_dbi_connect_info || []};
847 $self->_dbh(undef); # in case ->connected failed we might get sent here
848 $self->_dbh($self->_connect(@info));
850 $self->_conn_pid($$);
851 $self->_conn_tid(threads->tid) if $INC{'threads.pm'};
853 $self->_determine_driver;
855 # Always set the transaction depth on connect, since
856 # there is no transaction in progress by definition
857 $self->{transaction_depth} = $self->_dbh_autocommit ? 0 : 1;
859 $self->_run_connection_actions unless $self->{_in_determine_driver};
862 sub _run_connection_actions {
866 push @actions, ( $self->on_connect_call || () );
867 push @actions, $self->_parse_connect_do ('on_connect_do');
869 $self->_do_connection_actions(connect_call_ => $_) for @actions;
872 sub _determine_driver {
875 if ((not $self->_driver_determined) && (not $self->{_in_determine_driver})) {
876 my $started_unconnected = 0;
877 local $self->{_in_determine_driver} = 1;
879 if (ref($self) eq __PACKAGE__) {
881 if ($self->_dbh) { # we are connected
882 $driver = $self->_dbh->{Driver}{Name};
884 # if connect_info is a CODEREF, we have no choice but to connect
885 if (ref $self->_dbi_connect_info->[0] &&
886 Scalar::Util::reftype($self->_dbi_connect_info->[0]) eq 'CODE') {
887 $self->_populate_dbh;
888 $driver = $self->_dbh->{Driver}{Name};
891 # try to use dsn to not require being connected, the driver may still
892 # force a connection in _rebless to determine version
893 ($driver) = $self->_dbi_connect_info->[0] =~ /dbi:([^:]+):/i;
894 $started_unconnected = 1;
898 my $storage_class = "DBIx::Class::Storage::DBI::${driver}";
899 if ($self->load_optional_class($storage_class)) {
900 mro::set_mro($storage_class, 'c3');
901 bless $self, $storage_class;
906 $self->_driver_determined(1);
908 $self->_init; # run driver-specific initializations
910 $self->_run_connection_actions
911 if $started_unconnected && defined $self->_dbh;
915 sub _do_connection_actions {
917 my $method_prefix = shift;
920 if (not ref($call)) {
921 my $method = $method_prefix . $call;
923 } elsif (ref($call) eq 'CODE') {
925 } elsif (ref($call) eq 'ARRAY') {
926 if (ref($call->[0]) ne 'ARRAY') {
927 $self->_do_connection_actions($method_prefix, $_) for @$call;
929 $self->_do_connection_actions($method_prefix, @$_) for @$call;
932 $self->throw_exception (sprintf ("Don't know how to process conection actions of type '%s'", ref($call)) );
938 sub connect_call_do_sql {
940 $self->_do_query(@_);
943 sub disconnect_call_do_sql {
945 $self->_do_query(@_);
948 # override in db-specific backend when necessary
949 sub connect_call_datetime_setup { 1 }
952 my ($self, $action) = @_;
954 if (ref $action eq 'CODE') {
955 $action = $action->($self);
956 $self->_do_query($_) foreach @$action;
959 # Most debuggers expect ($sql, @bind), so we need to exclude
960 # the attribute hash which is the second argument to $dbh->do
961 # furthermore the bind values are usually to be presented
962 # as named arrayref pairs, so wrap those here too
963 my @do_args = (ref $action eq 'ARRAY') ? (@$action) : ($action);
964 my $sql = shift @do_args;
965 my $attrs = shift @do_args;
966 my @bind = map { [ undef, $_ ] } @do_args;
968 $self->_query_start($sql, @bind);
969 $self->_get_dbh->do($sql, $attrs, @do_args);
970 $self->_query_end($sql, @bind);
977 my ($self, @info) = @_;
979 $self->throw_exception("You failed to provide any connection info")
982 my ($old_connect_via, $dbh);
984 if ($INC{'Apache/DBI.pm'} && $ENV{MOD_PERL}) {
985 $old_connect_via = $DBI::connect_via;
986 $DBI::connect_via = 'connect';
990 if(ref $info[0] eq 'CODE') {
994 $dbh = DBI->connect(@info);
997 if($dbh && !$self->unsafe) {
998 my $weak_self = $self;
999 Scalar::Util::weaken($weak_self);
1000 $dbh->{HandleError} = sub {
1002 $weak_self->throw_exception("DBI Exception: $_[0]");
1005 croak ("DBI Exception: $_[0]");
1008 $dbh->{ShowErrorStatement} = 1;
1009 $dbh->{RaiseError} = 1;
1010 $dbh->{PrintError} = 0;
1014 $DBI::connect_via = $old_connect_via if $old_connect_via;
1016 $self->throw_exception("DBI Connection failed: " . ($@||$DBI::errstr))
1019 $self->_dbh_autocommit($dbh->{AutoCommit});
1025 my ($self, $name) = @_;
1027 $name = $self->_svp_generate_name
1028 unless defined $name;
1030 $self->throw_exception ("You can't use savepoints outside a transaction")
1031 if $self->{transaction_depth} == 0;
1033 $self->throw_exception ("Your Storage implementation doesn't support savepoints")
1034 unless $self->can('_svp_begin');
1036 push @{ $self->{savepoints} }, $name;
1038 $self->debugobj->svp_begin($name) if $self->debug;
1040 return $self->_svp_begin($name);
1044 my ($self, $name) = @_;
1046 $self->throw_exception ("You can't use savepoints outside a transaction")
1047 if $self->{transaction_depth} == 0;
1049 $self->throw_exception ("Your Storage implementation doesn't support savepoints")
1050 unless $self->can('_svp_release');
1052 if (defined $name) {
1053 $self->throw_exception ("Savepoint '$name' does not exist")
1054 unless grep { $_ eq $name } @{ $self->{savepoints} };
1056 # Dig through the stack until we find the one we are releasing. This keeps
1057 # the stack up to date.
1060 do { $svp = pop @{ $self->{savepoints} } } while $svp ne $name;
1062 $name = pop @{ $self->{savepoints} };
1065 $self->debugobj->svp_release($name) if $self->debug;
1067 return $self->_svp_release($name);
1071 my ($self, $name) = @_;
1073 $self->throw_exception ("You can't use savepoints outside a transaction")
1074 if $self->{transaction_depth} == 0;
1076 $self->throw_exception ("Your Storage implementation doesn't support savepoints")
1077 unless $self->can('_svp_rollback');
1079 if (defined $name) {
1080 # If they passed us a name, verify that it exists in the stack
1081 unless(grep({ $_ eq $name } @{ $self->{savepoints} })) {
1082 $self->throw_exception("Savepoint '$name' does not exist!");
1085 # Dig through the stack until we find the one we are releasing. This keeps
1086 # the stack up to date.
1087 while(my $s = pop(@{ $self->{savepoints} })) {
1088 last if($s eq $name);
1090 # Add the savepoint back to the stack, as a rollback doesn't remove the
1091 # named savepoint, only everything after it.
1092 push(@{ $self->{savepoints} }, $name);
1094 # We'll assume they want to rollback to the last savepoint
1095 $name = $self->{savepoints}->[-1];
1098 $self->debugobj->svp_rollback($name) if $self->debug;
1100 return $self->_svp_rollback($name);
1103 sub _svp_generate_name {
1106 return 'savepoint_'.scalar(@{ $self->{'savepoints'} });
1111 if($self->{transaction_depth} == 0) {
1112 $self->debugobj->txn_begin()
1114 $self->_dbh_begin_work;
1116 elsif ($self->auto_savepoint) {
1119 $self->{transaction_depth}++;
1122 sub _dbh_begin_work {
1125 # if the user is utilizing txn_do - good for him, otherwise we need to
1126 # ensure that the $dbh is healthy on BEGIN.
1127 # We do this via ->dbh_do instead of ->dbh, so that the ->dbh "ping"
1128 # will be replaced by a failure of begin_work itself (which will be
1129 # then retried on reconnect)
1130 if ($self->{_in_dbh_do}) {
1131 $self->_dbh->begin_work;
1133 $self->dbh_do(sub { $_[1]->begin_work });
1139 if ($self->{transaction_depth} == 1) {
1140 my $dbh = $self->_dbh;
1141 $self->debugobj->txn_commit()
1144 $self->{transaction_depth} = 0
1145 if $self->_dbh_autocommit;
1147 elsif($self->{transaction_depth} > 1) {
1148 $self->{transaction_depth}--;
1150 if $self->auto_savepoint;
1156 $self->_dbh->commit;
1161 my $dbh = $self->_dbh;
1163 if ($self->{transaction_depth} == 1) {
1164 $self->debugobj->txn_rollback()
1166 $self->{transaction_depth} = 0
1167 if $self->_dbh_autocommit;
1168 $self->_dbh_rollback;
1170 elsif($self->{transaction_depth} > 1) {
1171 $self->{transaction_depth}--;
1172 if ($self->auto_savepoint) {
1173 $self->svp_rollback;
1178 die DBIx::Class::Storage::NESTED_ROLLBACK_EXCEPTION->new;
1183 my $exception_class = "DBIx::Class::Storage::NESTED_ROLLBACK_EXCEPTION";
1184 $error =~ /$exception_class/ and $self->throw_exception($error);
1185 # ensure that a failed rollback resets the transaction depth
1186 $self->{transaction_depth} = $self->_dbh_autocommit ? 0 : 1;
1187 $self->throw_exception($error);
1193 $self->_dbh->rollback;
1196 # This used to be the top-half of _execute. It was split out to make it
1197 # easier to override in NoBindVars without duping the rest. It takes up
1198 # all of _execute's args, and emits $sql, @bind.
1199 sub _prep_for_execute {
1200 my ($self, $op, $extra_bind, $ident, $args) = @_;
1202 if( Scalar::Util::blessed($ident) && $ident->isa("DBIx::Class::ResultSource") ) {
1203 $ident = $ident->from();
1206 my ($sql, @bind) = $self->sql_maker->$op($ident, @$args);
1209 map { ref $_ eq 'ARRAY' ? $_ : [ '!!dummy', $_ ] } @$extra_bind)
1211 return ($sql, \@bind);
1215 sub _fix_bind_params {
1216 my ($self, @bind) = @_;
1218 ### Turn @bind from something like this:
1219 ### ( [ "artist", 1 ], [ "cdid", 1, 3 ] )
1221 ### ( "'1'", "'1'", "'3'" )
1224 if ( defined( $_ && $_->[1] ) ) {
1225 map { qq{'$_'}; } @{$_}[ 1 .. $#$_ ];
1232 my ( $self, $sql, @bind ) = @_;
1234 if ( $self->debug ) {
1235 @bind = $self->_fix_bind_params(@bind);
1237 $self->debugobj->query_start( $sql, @bind );
1242 my ( $self, $sql, @bind ) = @_;
1244 if ( $self->debug ) {
1245 @bind = $self->_fix_bind_params(@bind);
1246 $self->debugobj->query_end( $sql, @bind );
1251 my ($self, $dbh, $op, $extra_bind, $ident, $bind_attributes, @args) = @_;
1253 my ($sql, $bind) = $self->_prep_for_execute($op, $extra_bind, $ident, \@args);
1255 $self->_query_start( $sql, @$bind );
1257 my $sth = $self->sth($sql,$op);
1259 my $placeholder_index = 1;
1261 foreach my $bound (@$bind) {
1262 my $attributes = {};
1263 my($column_name, @data) = @$bound;
1265 if ($bind_attributes) {
1266 $attributes = $bind_attributes->{$column_name}
1267 if defined $bind_attributes->{$column_name};
1270 foreach my $data (@data) {
1271 my $ref = ref $data;
1272 $data = $ref && $ref ne 'ARRAY' ? ''.$data : $data; # stringify args (except arrayrefs)
1274 $sth->bind_param($placeholder_index, $data, $attributes);
1275 $placeholder_index++;
1279 # Can this fail without throwing an exception anyways???
1280 my $rv = $sth->execute();
1281 $self->throw_exception($sth->errstr) if !$rv;
1283 $self->_query_end( $sql, @$bind );
1285 return (wantarray ? ($rv, $sth, @$bind) : $rv);
1290 $self->dbh_do('_dbh_execute', @_); # retry over disconnects
1294 my ($self, $source, $to_insert) = @_;
1296 # redispatch to insert method of storage we reblessed into, if necessary
1297 if (not $self->_driver_determined) {
1298 $self->_determine_driver;
1299 goto $self->can('insert');
1302 my $ident = $source->from;
1303 my $bind_attributes = $self->source_bind_attributes($source);
1305 my $updated_cols = {};
1307 foreach my $col ( $source->columns ) {
1308 if ( !defined $to_insert->{$col} ) {
1309 my $col_info = $source->column_info($col);
1311 if ( $col_info->{auto_nextval} ) {
1312 $updated_cols->{$col} = $to_insert->{$col} = $self->_sequence_fetch(
1314 $col_info->{sequence} ||
1315 $self->_dbh_get_autoinc_seq($self->_get_dbh, $source)
1321 $self->_execute('insert' => [], $source, $bind_attributes, $to_insert);
1323 return $updated_cols;
1326 ## Still not quite perfect, and EXPERIMENTAL
1327 ## Currently it is assumed that all values passed will be "normal", i.e. not
1328 ## scalar refs, or at least, all the same type as the first set, the statement is
1329 ## only prepped once.
1331 my ($self, $source, $cols, $data, $sth_attr) = @_;
1333 # redispatch to insert_bulk method of storage we reblessed into, if necessary
1334 if (not $self->_driver_determined) {
1335 $self->_determine_driver;
1336 goto $self->can('insert_bulk');
1340 @colvalues{@$cols} = (0..$#$cols);
1342 # bind literal sql if it's the same in all slices
1343 for my $i (0..$#$cols) {
1344 my $first_val = $data->[0][$i];
1345 next unless (Scalar::Util::reftype($first_val)||'') eq 'SCALAR';
1347 $colvalues{ $cols->[$i] } = $first_val
1349 (Scalar::Util::reftype($_)||'') eq 'SCALAR' &&
1351 } map $data->[$_][$i], (1..$#$data)) == (@$data - 1);
1354 my ($sql, $bind) = $self->_prep_for_execute (
1355 'insert', undef, $source, [\%colvalues]
1359 my $empty_bind = 1 if (not @bind) &&
1360 (grep { (Scalar::Util::reftype($_)||'') eq 'SCALAR' } values %colvalues)
1363 if ((not @bind) && (not $empty_bind)) {
1364 croak 'Cannot insert_bulk without support for placeholders';
1367 $self->_query_start( $sql, @bind );
1368 my $sth = $self->sth($sql, 'insert', $sth_attr);
1371 # bind_param_array doesn't work if there are no binds
1373 local $self->_get_dbh->{RaiseError} = 1;
1374 local $self->_get_dbh->{PrintError} = 0;
1375 foreach (0..$#$data) {
1377 $sth->fetchall_arrayref;
1382 $self->throw_exception($exception) if $exception;
1386 # @bind = map { ref $_ ? ''.$_ : $_ } @bind; # stringify args
1388 ## This must be an arrayref, else nothing works!
1389 my $tuple_status = [];
1391 ## Get the bind_attributes, if any exist
1392 my $bind_attributes = $self->source_bind_attributes($source);
1394 ## Bind the values and execute
1395 my $placeholder_index = 1;
1397 foreach my $bound (@bind) {
1399 my $attributes = {};
1400 my ($column_name, $data_index) = @$bound;
1402 if( $bind_attributes ) {
1403 $attributes = $bind_attributes->{$column_name}
1404 if defined $bind_attributes->{$column_name};
1407 my @data = map { $_->[$data_index] } @$data;
1409 $sth->bind_param_array( $placeholder_index, [@data], $attributes );
1410 $placeholder_index++;
1413 my $rv = eval { $sth->execute_array({ArrayTupleStatus => $tuple_status}) };
1415 if (my $err = $@ || $sth->errstr) {
1417 ++$i while $i <= $#$tuple_status && !ref $tuple_status->[$i];
1419 $self->throw_exception("Unexpected populate error: $err")
1420 if ($i > $#$tuple_status);
1422 $self->throw_exception(sprintf "%s for populate slice:\n%s",
1423 ($tuple_status->[$i][1] || $err),
1424 $self->_pretty_print ({
1425 map { $cols->[$_] => $data->[$i][$_] } (0 .. $#$cols)
1430 $self->_query_end( $sql, @bind );
1431 return (wantarray ? ($rv, $sth, @bind) : $rv);
1435 my ($self, $source, @args) = @_;
1437 # redispatch to update method of storage we reblessed into, if necessary
1438 if (not $self->_driver_determined) {
1439 $self->_determine_driver;
1440 goto $self->can('update');
1443 my $bind_attributes = $self->source_bind_attributes($source);
1445 return $self->_execute('update' => [], $source, $bind_attributes, @args);
1450 my $self = shift @_;
1451 my $source = shift @_;
1452 $self->_determine_driver;
1453 my $bind_attrs = $self->source_bind_attributes($source);
1455 return $self->_execute('delete' => [], $source, $bind_attrs, @_);
1458 # We were sent here because the $rs contains a complex search
1459 # which will require a subquery to select the correct rows
1460 # (i.e. joined or limited resultsets)
1462 # Genarating a single PK column subquery is trivial and supported
1463 # by all RDBMS. However if we have a multicolumn PK, things get ugly.
1464 # Look at _multipk_update_delete()
1465 sub _subq_update_delete {
1467 my ($rs, $op, $values) = @_;
1469 my $rsrc = $rs->result_source;
1471 # we already check this, but double check naively just in case. Should be removed soon
1472 my $sel = $rs->_resolved_attrs->{select};
1473 $sel = [ $sel ] unless ref $sel eq 'ARRAY';
1474 my @pcols = $rsrc->primary_columns;
1475 if (@$sel != @pcols) {
1476 $self->throw_exception (
1477 'Subquery update/delete can not be called on resultsets selecting a'
1478 .' number of columns different than the number of primary keys'
1485 $op eq 'update' ? $values : (),
1486 { $pcols[0] => { -in => $rs->as_query } },
1491 return $self->_multipk_update_delete (@_);
1495 # ANSI SQL does not provide a reliable way to perform a multicol-PK
1496 # resultset update/delete involving subqueries. So by default resort
1497 # to simple (and inefficient) delete_all style per-row opearations,
1498 # while allowing specific storages to override this with a faster
1501 sub _multipk_update_delete {
1502 return shift->_per_row_update_delete (@_);
1505 # This is the default loop used to delete/update rows for multi PK
1506 # resultsets, and used by mysql exclusively (because it can't do anything
1509 # We do not use $row->$op style queries, because resultset update/delete
1510 # is not expected to cascade (this is what delete_all/update_all is for).
1512 # There should be no race conditions as the entire operation is rolled
1515 sub _per_row_update_delete {
1517 my ($rs, $op, $values) = @_;
1519 my $rsrc = $rs->result_source;
1520 my @pcols = $rsrc->primary_columns;
1522 my $guard = $self->txn_scope_guard;
1524 # emulate the return value of $sth->execute for non-selects
1525 my $row_cnt = '0E0';
1527 my $subrs_cur = $rs->cursor;
1528 while (my @pks = $subrs_cur->next) {
1531 for my $i (0.. $#pcols) {
1532 $cond->{$pcols[$i]} = $pks[$i];
1537 $op eq 'update' ? $values : (),
1552 # localization is neccessary as
1553 # 1) there is no infrastructure to pass this around before SQLA2
1554 # 2) _select_args sets it and _prep_for_execute consumes it
1555 my $sql_maker = $self->sql_maker;
1556 local $sql_maker->{_dbic_rs_attrs};
1558 return $self->_execute($self->_select_args(@_));
1561 sub _select_args_to_query {
1564 # localization is neccessary as
1565 # 1) there is no infrastructure to pass this around before SQLA2
1566 # 2) _select_args sets it and _prep_for_execute consumes it
1567 my $sql_maker = $self->sql_maker;
1568 local $sql_maker->{_dbic_rs_attrs};
1570 # my ($op, $bind, $ident, $bind_attrs, $select, $cond, $order, $rows, $offset)
1571 # = $self->_select_args($ident, $select, $cond, $attrs);
1572 my ($op, $bind, $ident, $bind_attrs, @args) =
1573 $self->_select_args(@_);
1575 # my ($sql, $prepared_bind) = $self->_prep_for_execute($op, $bind, $ident, [ $select, $cond, $order, $rows, $offset ]);
1576 my ($sql, $prepared_bind) = $self->_prep_for_execute($op, $bind, $ident, \@args);
1577 $prepared_bind ||= [];
1580 ? ($sql, $prepared_bind, $bind_attrs)
1581 : \[ "($sql)", @$prepared_bind ]
1586 my ($self, $ident, $select, $where, $attrs) = @_;
1588 my ($alias2source, $rs_alias) = $self->_resolve_ident_sources ($ident);
1590 my $sql_maker = $self->sql_maker;
1591 $sql_maker->{_dbic_rs_attrs} = {
1597 ? ( _source_handle => $alias2source->{$rs_alias}->handle )
1602 # calculate bind_attrs before possible $ident mangling
1603 my $bind_attrs = {};
1604 for my $alias (keys %$alias2source) {
1605 my $bindtypes = $self->source_bind_attributes ($alias2source->{$alias}) || {};
1606 for my $col (keys %$bindtypes) {
1608 my $fqcn = join ('.', $alias, $col);
1609 $bind_attrs->{$fqcn} = $bindtypes->{$col} if $bindtypes->{$col};
1611 # Unqialified column names are nice, but at the same time can be
1612 # rather ambiguous. What we do here is basically go along with
1613 # the loop, adding an unqualified column slot to $bind_attrs,
1614 # alongside the fully qualified name. As soon as we encounter
1615 # another column by that name (which would imply another table)
1616 # we unset the unqualified slot and never add any info to it
1617 # to avoid erroneous type binding. If this happens the users
1618 # only choice will be to fully qualify his column name
1620 if (exists $bind_attrs->{$col}) {
1621 $bind_attrs->{$col} = {};
1624 $bind_attrs->{$col} = $bind_attrs->{$fqcn};
1631 $attrs->{software_limit}
1633 $sql_maker->_default_limit_syntax eq "GenericSubQ"
1635 $attrs->{software_limit} = 1;
1638 $self->throw_exception("rows attribute must be positive if present")
1639 if (defined($attrs->{rows}) && !($attrs->{rows} > 0));
1641 # MySQL actually recommends this approach. I cringe.
1642 $attrs->{rows} = 2**48 if not defined $attrs->{rows} and defined $attrs->{offset};
1647 # see if we need to tear the prefetch apart (either limited has_many or grouped prefetch)
1648 # otherwise delegate the limiting to the storage, unless software limit was requested
1650 ( $attrs->{rows} && keys %{$attrs->{collapse}} )
1652 ( $attrs->{group_by} && @{$attrs->{group_by}} &&
1653 $attrs->{_prefetch_select} && @{$attrs->{_prefetch_select}} )
1655 ($ident, $select, $where, $attrs)
1656 = $self->_adjust_select_args_for_complex_prefetch ($ident, $select, $where, $attrs);
1658 elsif (! $attrs->{software_limit} ) {
1659 push @limit, $attrs->{rows}, $attrs->{offset};
1663 # This would be the point to deflate anything found in $where
1664 # (and leave $attrs->{bind} intact). Problem is - inflators historically
1665 # expect a row object. And all we have is a resultsource (it is trivial
1666 # to extract deflator coderefs via $alias2source above).
1668 # I don't see a way forward other than changing the way deflators are
1669 # invoked, and that's just bad...
1673 { $attrs->{$_} ? ( $_ => $attrs->{$_} ) : () }
1674 (qw/order_by group_by having/ )
1677 return ('select', $attrs->{bind}, $ident, $bind_attrs, $select, $where, $order, @limit);
1681 # This is the code producing joined subqueries like:
1682 # SELECT me.*, other.* FROM ( SELECT me.* FROM ... ) JOIN other ON ...
1684 sub _adjust_select_args_for_complex_prefetch {
1685 my ($self, $from, $select, $where, $attrs) = @_;
1687 $self->throw_exception ('Nothing to prefetch... how did we get here?!')
1688 if not @{$attrs->{_prefetch_select}};
1690 $self->throw_exception ('Complex prefetches are not supported on resultsets with a custom from attribute')
1691 if (ref $from ne 'ARRAY' || ref $from->[0] ne 'HASH' || ref $from->[1] ne 'ARRAY');
1694 # generate inner/outer attribute lists, remove stuff that doesn't apply
1695 my $outer_attrs = { %$attrs };
1696 delete $outer_attrs->{$_} for qw/where bind rows offset group_by having/;
1698 my $inner_attrs = { %$attrs };
1699 delete $inner_attrs->{$_} for qw/for collapse _prefetch_select _collapse_order_by select as/;
1702 # bring over all non-collapse-induced order_by into the inner query (if any)
1703 # the outer one will have to keep them all
1704 delete $inner_attrs->{order_by};
1705 if (my $ord_cnt = @{$outer_attrs->{order_by}} - @{$outer_attrs->{_collapse_order_by}} ) {
1706 $inner_attrs->{order_by} = [
1707 @{$outer_attrs->{order_by}}[ 0 .. $ord_cnt - 1]
1712 # generate the inner/outer select lists
1713 # for inside we consider only stuff *not* brought in by the prefetch
1714 # on the outside we substitute any function for its alias
1715 my $outer_select = [ @$select ];
1716 my $inner_select = [];
1717 for my $i (0 .. ( @$outer_select - @{$outer_attrs->{_prefetch_select}} - 1) ) {
1718 my $sel = $outer_select->[$i];
1720 if (ref $sel eq 'HASH' ) {
1721 $sel->{-as} ||= $attrs->{as}[$i];
1722 $outer_select->[$i] = join ('.', $attrs->{alias}, ($sel->{-as} || "inner_column_$i") );
1725 push @$inner_select, $sel;
1728 # normalize a copy of $from, so it will be easier to work with further
1729 # down (i.e. promote the initial hashref to an AoH)
1731 $from->[0] = [ $from->[0] ];
1732 my %original_join_info = map { $_->[0]{-alias} => $_->[0] } (@$from);
1735 # decide which parts of the join will remain in either part of
1736 # the outer/inner query
1738 # First we compose a list of which aliases are used in restrictions
1739 # (i.e. conditions/order/grouping/etc). Since we do not have
1740 # introspectable SQLA, we fall back to ugly scanning of raw SQL for
1741 # WHERE, and for pieces of ORDER BY in order to determine which aliases
1742 # need to appear in the resulting sql.
1743 # It may not be very efficient, but it's a reasonable stop-gap
1744 # Also unqualified column names will not be considered, but more often
1745 # than not this is actually ok
1747 # In the same loop we enumerate part of the selection aliases, as
1748 # it requires the same sqla hack for the time being
1749 my ($restrict_aliases, $select_aliases, $prefetch_aliases);
1751 # produce stuff unquoted, so it can be scanned
1752 my $sql_maker = $self->sql_maker;
1753 local $sql_maker->{quote_char};
1754 my $sep = $self->_sql_maker_opts->{name_sep} || '.';
1757 my $non_prefetch_select_sql = $sql_maker->_recurse_fields ($inner_select);
1758 my $prefetch_select_sql = $sql_maker->_recurse_fields ($outer_attrs->{_prefetch_select});
1759 my $where_sql = $sql_maker->where ($where);
1760 my $group_by_sql = $sql_maker->_order_by({
1761 map { $_ => $inner_attrs->{$_} } qw/group_by having/
1763 my @non_prefetch_order_by_chunks = (map
1764 { ref $_ ? $_->[0] : $_ }
1765 $sql_maker->_order_by_chunks ($inner_attrs->{order_by})
1769 for my $alias (keys %original_join_info) {
1770 my $seen_re = qr/\b $alias $sep/x;
1772 for my $piece ($where_sql, $group_by_sql, @non_prefetch_order_by_chunks ) {
1773 if ($piece =~ $seen_re) {
1774 $restrict_aliases->{$alias} = 1;
1778 if ($non_prefetch_select_sql =~ $seen_re) {
1779 $select_aliases->{$alias} = 1;
1782 if ($prefetch_select_sql =~ $seen_re) {
1783 $prefetch_aliases->{$alias} = 1;
1789 # Add any non-left joins to the restriction list (such joins are indeed restrictions)
1790 for my $j (values %original_join_info) {
1791 my $alias = $j->{-alias} or next;
1792 $restrict_aliases->{$alias} = 1 if (
1793 (not $j->{-join_type})
1795 ($j->{-join_type} !~ /^left (?: \s+ outer)? $/xi)
1799 # mark all join parents as mentioned
1800 # (e.g. join => { cds => 'tracks' } - tracks will need to bring cds too )
1801 for my $collection ($restrict_aliases, $select_aliases) {
1802 for my $alias (keys %$collection) {
1803 $collection->{$_} = 1
1804 for (@{ $original_join_info{$alias}{-join_path} || [] });
1808 # construct the inner $from for the subquery
1809 my %inner_joins = (map { %{$_ || {}} } ($restrict_aliases, $select_aliases) );
1811 for my $j (@$from) {
1812 push @inner_from, $j if $inner_joins{$j->[0]{-alias}};
1815 # if a multi-type join was needed in the subquery ("multi" is indicated by
1816 # presence in {collapse}) - add a group_by to simulate the collapse in the subq
1817 unless ($inner_attrs->{group_by}) {
1818 for my $alias (keys %inner_joins) {
1820 # the dot comes from some weirdness in collapse
1821 # remove after the rewrite
1822 if ($attrs->{collapse}{".$alias"}) {
1823 $inner_attrs->{group_by} ||= $inner_select;
1829 # demote the inner_from head
1830 $inner_from[0] = $inner_from[0][0];
1832 # generate the subquery
1833 my $subq = $self->_select_args_to_query (
1840 my $subq_joinspec = {
1841 -alias => $attrs->{alias},
1842 -source_handle => $inner_from[0]{-source_handle},
1843 $attrs->{alias} => $subq,
1846 # Generate the outer from - this is relatively easy (really just replace
1847 # the join slot with the subquery), with a major caveat - we can not
1848 # join anything that is non-selecting (not part of the prefetch), but at
1849 # the same time is a multi-type relationship, as it will explode the result.
1851 # There are two possibilities here
1852 # - either the join is non-restricting, in which case we simply throw it away
1853 # - it is part of the restrictions, in which case we need to collapse the outer
1854 # result by tackling yet another group_by to the outside of the query
1856 # so first generate the outer_from, up to the substitution point
1858 while (my $j = shift @$from) {
1859 if ($j->[0]{-alias} eq $attrs->{alias}) { # time to swap
1864 last; # we'll take care of what's left in $from below
1867 push @outer_from, $j;
1871 # see what's left - throw away if not selecting/restricting
1872 # also throw in a group_by if restricting to guard against
1873 # cross-join explosions
1875 while (my $j = shift @$from) {
1876 my $alias = $j->[0]{-alias};
1878 if ($select_aliases->{$alias} || $prefetch_aliases->{$alias}) {
1879 push @outer_from, $j;
1881 elsif ($restrict_aliases->{$alias}) {
1882 push @outer_from, $j;
1884 # FIXME - this should be obviated by SQLA2, as I'll be able to
1885 # have restrict_inner and restrict_outer... or something to that
1886 # effect... I think...
1888 # FIXME2 - I can't find a clean way to determine if a particular join
1889 # is a multi - instead I am just treating everything as a potential
1890 # explosive join (ribasushi)
1892 # if (my $handle = $j->[0]{-source_handle}) {
1893 # my $rsrc = $handle->resolve;
1894 # ... need to bail out of the following if this is not a multi,
1895 # as it will be much easier on the db ...
1897 $outer_attrs->{group_by} ||= $outer_select;
1902 # demote the outer_from head
1903 $outer_from[0] = $outer_from[0][0];
1905 # This is totally horrific - the $where ends up in both the inner and outer query
1906 # Unfortunately not much can be done until SQLA2 introspection arrives, and even
1907 # then if where conditions apply to the *right* side of the prefetch, you may have
1908 # to both filter the inner select (e.g. to apply a limit) and then have to re-filter
1909 # the outer select to exclude joins you didin't want in the first place
1911 # OTOH it can be seen as a plus: <ash> (notes that this query would make a DBA cry ;)
1912 return (\@outer_from, $outer_select, $where, $outer_attrs);
1915 sub _resolve_ident_sources {
1916 my ($self, $ident) = @_;
1918 my $alias2source = {};
1921 # the reason this is so contrived is that $ident may be a {from}
1922 # structure, specifying multiple tables to join
1923 if ( Scalar::Util::blessed($ident) && $ident->isa("DBIx::Class::ResultSource") ) {
1924 # this is compat mode for insert/update/delete which do not deal with aliases
1925 $alias2source->{me} = $ident;
1928 elsif (ref $ident eq 'ARRAY') {
1932 if (ref $_ eq 'HASH') {
1934 $rs_alias = $tabinfo->{-alias};
1936 if (ref $_ eq 'ARRAY' and ref $_->[0] eq 'HASH') {
1940 $alias2source->{$tabinfo->{-alias}} = $tabinfo->{-source_handle}->resolve
1941 if ($tabinfo->{-source_handle});
1945 return ($alias2source, $rs_alias);
1948 # Takes $ident, \@column_names
1950 # returns { $column_name => \%column_info, ... }
1951 # also note: this adds -result_source => $rsrc to the column info
1954 # my $col_sources = $self->_resolve_column_info($ident, @column_names);
1955 sub _resolve_column_info {
1956 my ($self, $ident, $colnames) = @_;
1957 my ($alias2src, $root_alias) = $self->_resolve_ident_sources($ident);
1959 my $sep = $self->_sql_maker_opts->{name_sep} || '.';
1962 my (%return, %seen_cols);
1964 # compile a global list of column names, to be able to properly
1965 # disambiguate unqualified column names (if at all possible)
1966 for my $alias (keys %$alias2src) {
1967 my $rsrc = $alias2src->{$alias};
1968 for my $colname ($rsrc->columns) {
1969 push @{$seen_cols{$colname}}, $alias;
1974 foreach my $col (@$colnames) {
1975 my ($alias, $colname) = $col =~ m/^ (?: ([^$sep]+) $sep)? (.+) $/x;
1978 # see if the column was seen exactly once (so we know which rsrc it came from)
1979 if ($seen_cols{$colname} and @{$seen_cols{$colname}} == 1) {
1980 $alias = $seen_cols{$colname}[0];
1987 my $rsrc = $alias2src->{$alias};
1988 $return{$col} = $rsrc && {
1989 %{$rsrc->column_info($colname)},
1990 -result_source => $rsrc,
1991 -source_alias => $alias,
1998 # Returns a counting SELECT for a simple count
1999 # query. Abstracted so that a storage could override
2000 # this to { count => 'firstcol' } or whatever makes
2001 # sense as a performance optimization
2003 #my ($self, $source, $rs_attrs) = @_;
2004 return { count => '*' };
2007 # Returns a SELECT which will end up in the subselect
2008 # There may or may not be a group_by, as the subquery
2009 # might have been called to accomodate a limit
2011 # Most databases would be happy with whatever ends up
2012 # here, but some choke in various ways.
2014 sub _subq_count_select {
2015 my ($self, $source, $rs_attrs) = @_;
2016 return $rs_attrs->{group_by} if $rs_attrs->{group_by};
2018 my @pcols = map { join '.', $rs_attrs->{alias}, $_ } ($source->primary_columns);
2019 return @pcols ? \@pcols : [ 1 ];
2023 # Returns an ordered list of column names before they are used
2024 # in a SELECT statement. By default simply returns the list
2027 # This may be overridden in a specific storage when there are
2028 # requirements such as moving BLOB columns to the end of the
2030 sub _order_select_columns {
2031 #my ($self, $source, $columns) = @_;
2035 sub source_bind_attributes {
2036 my ($self, $source) = @_;
2038 my $bind_attributes;
2039 foreach my $column ($source->columns) {
2041 my $data_type = $source->column_info($column)->{data_type} || '';
2042 $bind_attributes->{$column} = $self->bind_attribute_by_data_type($data_type)
2046 return $bind_attributes;
2053 =item Arguments: $ident, $select, $condition, $attrs
2057 Handle a SQL select statement.
2063 my ($ident, $select, $condition, $attrs) = @_;
2064 return $self->cursor_class->new($self, \@_, $attrs);
2069 my ($rv, $sth, @bind) = $self->_select(@_);
2070 my @row = $sth->fetchrow_array;
2071 my @nextrow = $sth->fetchrow_array if @row;
2072 if(@row && @nextrow) {
2073 carp "Query returned more than one row. SQL that returns multiple rows is DEPRECATED for ->find and ->single";
2075 # Need to call finish() to work round broken DBDs
2084 =item Arguments: $sql
2088 Returns a L<DBI> sth (statement handle) for the supplied SQL.
2093 my ($self, $dbh, $sql, $op, $sth_attr) = @_;
2094 # $op is ignored right now
2098 # 3 is the if_active parameter which avoids active sth re-use
2099 my $sth = $self->disable_sth_caching
2100 ? $dbh->prepare($sql, $sth_attr)
2101 : $dbh->prepare_cached($sql, $sth_attr, 3);
2103 # XXX You would think RaiseError would make this impossible,
2104 # but apparently that's not true :(
2105 $self->throw_exception($dbh->errstr) if !$sth;
2111 my ($self, $sql, $op, $sth_attr) = @_;
2112 $self->dbh_do('_dbh_sth', $sql, $op, $sth_attr); # retry over disconnects
2115 sub _dbh_columns_info_for {
2116 my ($self, $dbh, $table) = @_;
2118 if ($dbh->can('column_info')) {
2121 my ($schema,$tab) = $table =~ /^(.+?)\.(.+)$/ ? ($1,$2) : (undef,$table);
2122 my $sth = $dbh->column_info( undef,$schema, $tab, '%' );
2124 while ( my $info = $sth->fetchrow_hashref() ){
2126 $column_info{data_type} = $info->{TYPE_NAME};
2127 $column_info{size} = $info->{COLUMN_SIZE};
2128 $column_info{is_nullable} = $info->{NULLABLE} ? 1 : 0;
2129 $column_info{default_value} = $info->{COLUMN_DEF};
2130 my $col_name = $info->{COLUMN_NAME};
2131 $col_name =~ s/^\"(.*)\"$/$1/;
2133 $result{$col_name} = \%column_info;
2136 return \%result if !$@ && scalar keys %result;
2140 my $sth = $dbh->prepare($self->sql_maker->select($table, undef, \'1 = 0'));
2142 my @columns = @{$sth->{NAME_lc}};
2143 for my $i ( 0 .. $#columns ){
2145 $column_info{data_type} = $sth->{TYPE}->[$i];
2146 $column_info{size} = $sth->{PRECISION}->[$i];
2147 $column_info{is_nullable} = $sth->{NULLABLE}->[$i] ? 1 : 0;
2149 if ($column_info{data_type} =~ m/^(.*?)\((.*?)\)$/) {
2150 $column_info{data_type} = $1;
2151 $column_info{size} = $2;
2154 $result{$columns[$i]} = \%column_info;
2158 foreach my $col (keys %result) {
2159 my $colinfo = $result{$col};
2160 my $type_num = $colinfo->{data_type};
2162 if(defined $type_num && $dbh->can('type_info')) {
2163 my $type_info = $dbh->type_info($type_num);
2164 $type_name = $type_info->{TYPE_NAME} if $type_info;
2165 $colinfo->{data_type} = $type_name if $type_name;
2172 sub columns_info_for {
2173 my ($self, $table) = @_;
2174 $self->_dbh_columns_info_for ($self->_get_dbh, $table);
2177 =head2 last_insert_id
2179 Return the row id of the last insert.
2183 sub _dbh_last_insert_id {
2184 # All Storage's need to register their own _dbh_last_insert_id
2185 # the old SQLite-based method was highly inappropriate
2188 my $class = ref $self;
2189 $self->throw_exception (<<EOE);
2191 No _dbh_last_insert_id() method found in $class.
2192 Since the method of obtaining the autoincrement id of the last insert
2193 operation varies greatly between different databases, this method must be
2194 individually implemented for every storage class.
2198 sub last_insert_id {
2200 $self->_dbh_last_insert_id ($self->_dbh, @_);
2203 =head2 _native_data_type
2207 =item Arguments: $type_name
2211 This API is B<EXPERIMENTAL>, will almost definitely change in the future, and
2212 currently only used by L<::AutoCast|DBIx::Class::Storage::DBI::AutoCast> and
2213 L<::Sybase|DBIx::Class::Storage::DBI::Sybase>.
2215 The default implementation returns C<undef>, implement in your Storage driver if
2216 you need this functionality.
2218 Should map types from other databases to the native RDBMS type, for example
2219 C<VARCHAR2> to C<VARCHAR>.
2221 Types with modifiers should map to the underlying data type. For example,
2222 C<INTEGER AUTO_INCREMENT> should become C<INTEGER>.
2224 Composite types should map to the container type, for example
2225 C<ENUM(foo,bar,baz)> becomes C<ENUM>.
2229 sub _native_data_type {
2230 #my ($self, $data_type) = @_;
2234 # Check if placeholders are supported at all
2235 sub _placeholders_supported {
2237 my $dbh = $self->_get_dbh;
2239 # some drivers provide a $dbh attribute (e.g. Sybase and $dbh->{syb_dynamic_supported})
2240 # but it is inaccurate more often than not
2242 local $dbh->{PrintError} = 0;
2243 local $dbh->{RaiseError} = 1;
2244 $dbh->do('select ?', {}, 1);
2249 # Check if placeholders bound to non-string types throw exceptions
2251 sub _typeless_placeholders_supported {
2253 my $dbh = $self->_get_dbh;
2256 local $dbh->{PrintError} = 0;
2257 local $dbh->{RaiseError} = 1;
2258 # this specifically tests a bind that is NOT a string
2259 $dbh->do('select 1 where 1 = ?', {}, 1);
2266 Returns the database driver name.
2273 if (not $self->_driver_determined) {
2274 $self->_determine_driver;
2275 goto $self->can ('sqlt_type');
2278 $self->_get_dbh->{Driver}->{Name};
2281 =head2 bind_attribute_by_data_type
2283 Given a datatype from column info, returns a database specific bind
2284 attribute for C<< $dbh->bind_param($val,$attribute) >> or nothing if we will
2285 let the database planner just handle it.
2287 Generally only needed for special case column types, like bytea in postgres.
2291 sub bind_attribute_by_data_type {
2295 =head2 is_datatype_numeric
2297 Given a datatype from column_info, returns a boolean value indicating if
2298 the current RDBMS considers it a numeric value. This controls how
2299 L<DBIx::Class::Row/set_column> decides whether to mark the column as
2300 dirty - when the datatype is deemed numeric a C<< != >> comparison will
2301 be performed instead of the usual C<eq>.
2305 sub is_datatype_numeric {
2306 my ($self, $dt) = @_;
2308 return 0 unless $dt;
2310 return $dt =~ /^ (?:
2311 numeric | int(?:eger)? | (?:tiny|small|medium|big)int | dec(?:imal)? | real | float | double (?: \s+ precision)? | (?:big)?serial
2316 =head2 create_ddl_dir (EXPERIMENTAL)
2320 =item Arguments: $schema \@databases, $version, $directory, $preversion, \%sqlt_args
2324 Creates a SQL file based on the Schema, for each of the specified
2325 database engines in C<\@databases> in the given directory.
2326 (note: specify L<SQL::Translator> names, not L<DBI> driver names).
2328 Given a previous version number, this will also create a file containing
2329 the ALTER TABLE statements to transform the previous schema into the
2330 current one. Note that these statements may contain C<DROP TABLE> or
2331 C<DROP COLUMN> statements that can potentially destroy data.
2333 The file names are created using the C<ddl_filename> method below, please
2334 override this method in your schema if you would like a different file
2335 name format. For the ALTER file, the same format is used, replacing
2336 $version in the name with "$preversion-$version".
2338 See L<SQL::Translator/METHODS> for a list of values for C<\%sqlt_args>.
2339 The most common value for this would be C<< { add_drop_table => 1 } >>
2340 to have the SQL produced include a C<DROP TABLE> statement for each table
2341 created. For quoting purposes supply C<quote_table_names> and
2342 C<quote_field_names>.
2344 If no arguments are passed, then the following default values are assumed:
2348 =item databases - ['MySQL', 'SQLite', 'PostgreSQL']
2350 =item version - $schema->schema_version
2352 =item directory - './'
2354 =item preversion - <none>
2358 By default, C<\%sqlt_args> will have
2360 { add_drop_table => 1, ignore_constraint_names => 1, ignore_index_names => 1 }
2362 merged with the hash passed in. To disable any of those features, pass in a
2363 hashref like the following
2365 { ignore_constraint_names => 0, # ... other options }
2368 Note that this feature is currently EXPERIMENTAL and may not work correctly
2369 across all databases, or fully handle complex relationships.
2371 WARNING: Please check all SQL files created, before applying them.
2375 sub create_ddl_dir {
2376 my ($self, $schema, $databases, $version, $dir, $preversion, $sqltargs) = @_;
2378 if(!$dir || !-d $dir) {
2379 carp "No directory given, using ./\n";
2382 $databases ||= ['MySQL', 'SQLite', 'PostgreSQL'];
2383 $databases = [ $databases ] if(ref($databases) ne 'ARRAY');
2385 my $schema_version = $schema->schema_version || '1.x';
2386 $version ||= $schema_version;
2389 add_drop_table => 1,
2390 ignore_constraint_names => 1,
2391 ignore_index_names => 1,
2395 $self->throw_exception("Can't create a ddl file without SQL::Translator: " . $self->_sqlt_version_error)
2396 if !$self->_sqlt_version_ok;
2398 my $sqlt = SQL::Translator->new( $sqltargs );
2400 $sqlt->parser('SQL::Translator::Parser::DBIx::Class');
2401 my $sqlt_schema = $sqlt->translate({ data => $schema })
2402 or $self->throw_exception ($sqlt->error);
2404 foreach my $db (@$databases) {
2406 $sqlt->{schema} = $sqlt_schema;
2407 $sqlt->producer($db);
2410 my $filename = $schema->ddl_filename($db, $version, $dir);
2411 if (-e $filename && ($version eq $schema_version )) {
2412 # if we are dumping the current version, overwrite the DDL
2413 carp "Overwriting existing DDL file - $filename";
2417 my $output = $sqlt->translate;
2419 carp("Failed to translate to $db, skipping. (" . $sqlt->error . ")");
2422 if(!open($file, ">$filename")) {
2423 $self->throw_exception("Can't open $filename for writing ($!)");
2426 print $file $output;
2429 next unless ($preversion);
2431 require SQL::Translator::Diff;
2433 my $prefilename = $schema->ddl_filename($db, $preversion, $dir);
2434 if(!-e $prefilename) {
2435 carp("No previous schema file found ($prefilename)");
2439 my $difffile = $schema->ddl_filename($db, $version, $dir, $preversion);
2441 carp("Overwriting existing diff file - $difffile");
2447 my $t = SQL::Translator->new($sqltargs);
2452 or $self->throw_exception ($t->error);
2454 my $out = $t->translate( $prefilename )
2455 or $self->throw_exception ($t->error);
2457 $source_schema = $t->schema;
2459 $source_schema->name( $prefilename )
2460 unless ( $source_schema->name );
2463 # The "new" style of producers have sane normalization and can support
2464 # diffing a SQL file against a DBIC->SQLT schema. Old style ones don't
2465 # And we have to diff parsed SQL against parsed SQL.
2466 my $dest_schema = $sqlt_schema;
2468 unless ( "SQL::Translator::Producer::$db"->can('preprocess_schema') ) {
2469 my $t = SQL::Translator->new($sqltargs);
2474 or $self->throw_exception ($t->error);
2476 my $out = $t->translate( $filename )
2477 or $self->throw_exception ($t->error);
2479 $dest_schema = $t->schema;
2481 $dest_schema->name( $filename )
2482 unless $dest_schema->name;
2485 my $diff = SQL::Translator::Diff::schema_diff($source_schema, $db,
2489 if(!open $file, ">$difffile") {
2490 $self->throw_exception("Can't write to $difffile ($!)");
2498 =head2 deployment_statements
2502 =item Arguments: $schema, $type, $version, $directory, $sqlt_args
2506 Returns the statements used by L</deploy> and L<DBIx::Class::Schema/deploy>.
2508 The L<SQL::Translator> (not L<DBI>) database driver name can be explicitly
2509 provided in C<$type>, otherwise the result of L</sqlt_type> is used as default.
2511 C<$directory> is used to return statements from files in a previously created
2512 L</create_ddl_dir> directory and is optional. The filenames are constructed
2513 from L<DBIx::Class::Schema/ddl_filename>, the schema name and the C<$version>.
2515 If no C<$directory> is specified then the statements are constructed on the
2516 fly using L<SQL::Translator> and C<$version> is ignored.
2518 See L<SQL::Translator/METHODS> for a list of values for C<$sqlt_args>.
2522 sub deployment_statements {
2523 my ($self, $schema, $type, $version, $dir, $sqltargs) = @_;
2524 $type ||= $self->sqlt_type;
2525 $version ||= $schema->schema_version || '1.x';
2527 my $filename = $schema->ddl_filename($type, $version, $dir);
2531 open($file, "<$filename")
2532 or $self->throw_exception("Can't open $filename ($!)");
2535 return join('', @rows);
2538 $self->throw_exception("Can't deploy without either SQL::Translator or a ddl_dir: " . $self->_sqlt_version_error )
2539 if !$self->_sqlt_version_ok;
2541 # sources needs to be a parser arg, but for simplicty allow at top level
2543 $sqltargs->{parser_args}{sources} = delete $sqltargs->{sources}
2544 if exists $sqltargs->{sources};
2546 my $tr = SQL::Translator->new(
2547 producer => "SQL::Translator::Producer::${type}",
2549 parser => 'SQL::Translator::Parser::DBIx::Class',
2552 return $tr->translate;
2556 my ($self, $schema, $type, $sqltargs, $dir) = @_;
2559 return if($line =~ /^--/);
2561 # next if($line =~ /^DROP/m);
2562 return if($line =~ /^BEGIN TRANSACTION/m);
2563 return if($line =~ /^COMMIT/m);
2564 return if $line =~ /^\s+$/; # skip whitespace only
2565 $self->_query_start($line);
2567 # do a dbh_do cycle here, as we need some error checking in
2568 # place (even though we will ignore errors)
2569 $self->dbh_do (sub { $_[1]->do($line) });
2572 carp qq{$@ (running "${line}")};
2574 $self->_query_end($line);
2576 my @statements = $self->deployment_statements($schema, $type, undef, $dir, { %{ $sqltargs || {} }, no_comments => 1 } );
2577 if (@statements > 1) {
2578 foreach my $statement (@statements) {
2579 $deploy->( $statement );
2582 elsif (@statements == 1) {
2583 foreach my $line ( split(";\n", $statements[0])) {
2589 =head2 datetime_parser
2591 Returns the datetime parser class
2595 sub datetime_parser {
2597 return $self->{datetime_parser} ||= do {
2598 $self->build_datetime_parser(@_);
2602 =head2 datetime_parser_type
2604 Defines (returns) the datetime parser class - currently hardwired to
2605 L<DateTime::Format::MySQL>
2609 sub datetime_parser_type { "DateTime::Format::MySQL"; }
2611 =head2 build_datetime_parser
2613 See L</datetime_parser>
2617 sub build_datetime_parser {
2618 if (not $_[0]->_driver_determined) {
2619 $_[0]->_determine_driver;
2620 goto $_[0]->can('build_datetime_parser');
2624 my $type = $self->datetime_parser_type(@_);
2625 $self->ensure_class_loaded ($type);
2630 =head2 is_replicating
2632 A boolean that reports if a particular L<DBIx::Class::Storage::DBI> is set to
2633 replicate from a master database. Default is undef, which is the result
2634 returned by databases that don't support replication.
2638 sub is_replicating {
2643 =head2 lag_behind_master
2645 Returns a number that represents a certain amount of lag behind a master db
2646 when a given storage is replicating. The number is database dependent, but
2647 starts at zero and increases with the amount of lag. Default in undef
2651 sub lag_behind_master {
2658 $self->_verify_pid if $self->_dbh;
2660 # some databases need this to stop spewing warnings
2661 if (my $dbh = $self->_dbh) {
2663 eval { $dbh->disconnect };
2673 =head2 DBIx::Class and AutoCommit
2675 DBIx::Class can do some wonderful magic with handling exceptions,
2676 disconnections, and transactions when you use C<< AutoCommit => 1 >>
2677 (the default) combined with C<txn_do> for transaction support.
2679 If you set C<< AutoCommit => 0 >> in your connect info, then you are always
2680 in an assumed transaction between commits, and you're telling us you'd
2681 like to manage that manually. A lot of the magic protections offered by
2682 this module will go away. We can't protect you from exceptions due to database
2683 disconnects because we don't know anything about how to restart your
2684 transactions. You're on your own for handling all sorts of exceptional
2685 cases if you choose the C<< AutoCommit => 0 >> path, just as you would
2691 Matt S. Trout <mst@shadowcatsystems.co.uk>
2693 Andy Grundman <andy@hybridized.org>
2697 You may distribute this code under the same terms as Perl itself.