package DB2::Table; use diagnostics; use strict; use warnings; use Carp; use DBI qw(:sql_types); our $VERSION = '0.23'; =head1 NAME DB2::Table - Framework wrapper around tables using DBD::DB2 =head1 SYNOPSIS package myTable; use DB2::Table; our @ISA = qw( DB2::Table ); ... use myDB; use myTable; my $db = myDB->new; my $tbl = $db->get_table('myTable'); my $row = $tbl->find($id); =head1 FUNCTIONS =over 4 =item C Do not call this - you should get your table through your database object. =cut sub new { my $class = shift; my $self = {}; bless $self, ref $class || $class || confess("Unknown table"); my $db = shift; confess("Need the db handle as parameter") unless $db and ref $db and $db->isa("DB2::db"); $self->{db} = $db; my %tableOrder; my @cl = $self->column_list; @tableOrder{ @cl } = (0..$#cl); $self->{tableOrder} = \%tableOrder; $self; } =item C the key sub to override! The data must be a reference to an array of hashes. Each element (hash) in the array must contain certain keys, others are optional. =over 2 =item Required: =over 2 =item C Column Name (must be upper case) =item C SQL type or one of: =over 4 =item C This will be represented by a NOT NULL CHAR that is limited to 'Y' or 'N'. In perl, this will be auto-converted to perlish true/false values. An undef will be treated as expected in perl: as false. =item C As above, but NULLs will be permitted. In this case, an 'N' in the database will become a false, but defined, value. Only a NULL in the database will translate to undef in perl. =back =back =item Optional: =over 2 =item C for CHAR, VARCHAR, etc. =item C optional stuff - C, C, etc. (Should use C rather than C 'PRIMARY KEY'>.) =item C default value =item C true for the primary key =item C stuff that is placed in the table create independantly =item C For this column, will create a FOREIGN KEY statement. The value here is used during creation of the table, and should begin with the foreign table name and include any "ON DELETE", "ON UPDATE", etc., portions. This may change in the future where C will be itself another hashref with all these fields. =item C For this column, will create as a generated identity. If this is undef or the word 'default', the option will be C<(START WITH 0, INCREMENT BY 1, NO CACHE)>, otherwise it will use whatever you provide here. =back =back This is somewhat based on a single column for a primary key, which is not necessarily the "right" thing to do in relational design, but sure as heck simplifies coding! NOTE: Other columns may be present, but would only be used by the subclass. =cut sub data_order { die "Gotta override data_order!"; } sub _internal_data_order { my $self = shift; unless ($self->{_data_order}) { $self->{_data_order} = $self->data_order(); } $self->{_data_order}; } sub _internal_data_reset { my $self = shift; delete $self->{_data_order}; delete $self->{column_list}; delete $self->{ALL_DATA}; delete $self->{PRIMARY}; delete $self->{GENERATEDIDENTITY}; } =item C When allowing the framework to create your row type object because there is no backing module, we need to know what to derive it from. If you have a generic row type that is derived from DB2::Row that you want all your rows to be derived from, you can override this. If all your empty Row types are derived from a single type that is not DB2::Row, you should create a single Table type and have all your tables derived from that. That is, to create a derivation tree for your row such as: DB2::Row -> My::Row -> My::UserR your derivation tree for your tables should look like: DB2::Table -> My::Table -> My::User And then C can override C to return C =cut sub get_base_row_type { q(DB2::Row); } =item C Gets the DB2::db object that contains this table =cut sub getDB { shift->{db}; } =item C You need to override this. Must return the DB2 Schema to use for this table. Generally, you may want to derive a single "schema" class from DB2::Table which only overrides this method, and then derive each table in that schema from that class. =cut sub schema_name { confess("You must override schema_name") } sub _connection { my $self = shift; $self->getDB->connection; } sub _find_create_row { my $self = shift; my $type = $self->{db}->get_row_type_for_table(ref $self); my @row = @_; my %params = ( _db_object => $self->getDB ); if ($row[-1] and ref $row[-1] eq 'HASH') { %params = ( %params, %{$row[-1]} ); pop @row; } my $data_order = $self->_internal_data_order(); foreach my $i (0..$#$data_order) { my $column = $data_order->[$i]{column}; if (defined $row[$i] and not exists $params{$column}) { ($params{$column} = $row[$i]) =~ s/\s*$//; } } return $type->new(\%params); } =item C Creates a new DB2::Row object for this table. Called instead of the constructor for the DB2::Row object. Sets up defaults, etc. B: this will not generate any identity column! We leave that up to the database, so we will retrieve that during the save before committing. =cut sub create_row { my $self = shift; $self->_find_create_row( map ( { $self->get_column($_, 'default'); } $self->column_list ), @_ ); } =item C Should be obvious - a full count of all the rows in this table =cut sub count { my $self = shift; $self->SELECT('COUNT(*)')->[0][0]; } =item C Similar to C, except that the first parameter will be the SQL WHERE condition while the rest of the parameters will be the bind values for that WHERE condition. =cut sub count_where { my $self = shift; $self->SELECT('COUNT(*)', @_)->[0][0]; } =item C Finds all rows with the primary column matching any of the parameters. For example, $tbl->find_id(1, 2, 10) will return an array of DB2::Row derived objects with all the data from 0-3 rows from this table, if the primary column for that row is either 1, 2, or 10. =cut sub find_id { my $self = shift; $self->find_where( $self->primaryColumn . ' IN (' . join (', ', map {'?'} @_) . ')', @_ ); } =item C Similar to C, the first parameter is the SQL WHERE condition while the rest of the parameters are the bind values for the WHERE condition. In array context, will return the array of DB2::Row derived objects returned, whether empty or not. In scalar context, will return undef if no rows are found, will return the single Row object if only one row is found, or an array ref if more than one row is found. =cut sub find_where { my $self = shift; $self->find_join($self->full_table_name, @_); } =item C Similar to C, the first parameter is the tables to join and how they are joined (any '!!!' found will be replaced with the current table's full name), the second parameter is the where condition, if any, and the rest are bind values. =cut sub find_join { my $self = shift; my @cols = $self->column_list(); my $prefix = ""; my $tables = shift; if (ref $tables and ref $tables eq 'ARRAY') { $tables = join ' ', @$tables; } if ($tables and ( $tables =~ /!!!\s+[Aa][Ss]\s+(\w+)/ or $tables =~ /$self->full_table_name()\s+[Aa][Ss]\s+(\w+)/ or $tables =~ /$self->table_name()\s+[Aa][Ss]\s+(\w+)/ ) ) { $prefix = "$1."; } my $ary_ref = $self->SELECT_join( { forreadonly => 1, #distinct => 1, prepare_attributes => $self->_prepare_attributes('SELECT'), }, join(', ', map {$prefix . $_} $self->column_list), $tables, @_); my @rc; foreach my $row (@$ary_ref) { push @rc, $self->_find_create_row(@$row); } # array, empty or not. if (wantarray) { return @rc; } # if there aren't any, send back undef. if (scalar @rc < 1) { return undef; } # no array wanted, and only one answer, send it back. if (scalar @rc == 1) { return $rc[0]; } # no array wanted, send back ref to array. return \@rc; } =item C<_prepare_attributes> Internally used to set any prepare attributes. Parameter says what type of prepare this is, although the list is not finalised yet. =cut sub _prepare_attributes { {} } =item C<_prepare> Internally used to cache statements. This may change to C if it is found to be useful. =cut sub _prepare { my $self = shift; my $stmt = shift; my $attr = shift; DB2::db::_debug("$stmt\n"); my $sth = $self->_connection->prepare_cached($stmt, $attr); croak "Can't prepare [$stmt]: " . $self->_connection->errstr() unless $sth; $sth; } sub _execute { my $self = shift; my $sth = shift; delete $self->{_dbi}; unless ($sth->execute(@_)) { $self->{_dbi}{err} = $sth->err; $self->{_dbi}{errstr} = $sth->errstr; $self->{_dbi}{state} = $sth->state; DB2::db::_debug("Failed to execute $sth->{Statement}: ", $sth->errstr()); undef; } } =item C =item C =item C Shortcuts to get the DBI err, errstr, and state's, respectively. =cut sub dbi_err { shift->{_dbi}{err} } sub dbi_errstr { shift->{_dbi}{errstr} } sub dbi_state { shift->{_dbi}{state} } sub _already_exists_in_db { my $self = shift; my $obj = shift; my $dbh = $self->_connection; my $count = 0; my $column = $self->primaryColumn; if (ref $obj) { if ($column) { $obj = $obj->column($column); } } if (defined $obj and not ref $obj) { #my $stmt = "SELECT COUNT(*) FROM " . $self->full_table_name . # " WHERE $column IN ?"; #$count = $dbh->selectrow_array($stmt, undef, $objval); $count = $self->SELECT('COUNT(*)', "$column IN ?", $obj)->[0][0]; } return $count; } sub _update_db { my $self = shift; my $obj = shift; my $prep_attr = shift; # it's an update. my $stmt = "UPDATE " . $self->full_table_name . " SET "; my $prim_key = $self->primaryColumn; # find all modified fields. my @sets; my @newVal; my @bind; { for my $key (keys %{$obj->{modified}}) { next if $key eq $prim_key; push @sets, "$key = ?"; push @bind, [$obj->{CONFIG}{$key}]; if ($self->get_column($key,'type') =~ /LOB$/) { push @{$bind[$#bind]}, 'SQL_BLOB'; } } } if (@sets) { $stmt .= join(", ", @sets); $stmt .= " WHERE " . $self->primaryColumn . " IN ?"; my $sth = $self->_prepare($stmt, $prep_attr); my $i = 0; for (; $i < @bind; ++$i) { if ($DB2::db::debug) { print "Binding ", $i + 1, " => "; if (scalar @{$bind[$i]} > 1 and $bind[$i][1] == SQL_BLOB) { print "[blob],", SQL_BLOB; } else { print join(",",@{$bind[$i]}); } print "\n"; } $sth->bind_param($i + 1, @{$bind[$i]}); } $sth->bind_param($i + 1, $obj->{CONFIG}{$prim_key}); print "stmt = $stmt\n" if $DB2::db::debug; $self->_execute($sth); #, @newVal); $sth->finish(); $self->commit(); } else { '0E0'; # default return value. } } sub _insert_into_db { my $self = shift; my $obj = shift; my $prep_attr = shift; my @cols = grep { not $self->get_column($_, 'NOCREATE') and $_ ne $self->generatedIdentityColumn() } $self->column_list; my $stmt = "INSERT INTO " . $self->full_table_name . " (" . join(', ', @cols) . ") VALUES(" . join(', ', map {'?'} @cols) . ")"; DB2::db::_debug("$stmt\n"); my $sth = $self->_prepare($stmt, $prep_attr); my @bind; { my $i = 0; for my $key (map { uc $_ } @cols) { ++$i; push @bind, [$obj->{CONFIG}{$key}]; if ($self->get_column($key,'type') =~ /LOB$/) { my $x = $obj->{CONFIG}{$key}; #$bind[$#bind] = [\$x, {TYPE => SQL_BLOB}]; $bind[$#bind] = [$x, {TYPE => SQL_BLOB}]; #$bind[$#bind] = [$x, SQL_BLOB]; } } } #print STDERR "stmt = $stmt -- ", join @newVal, "\n"; for (my $i = 0; $i < @bind; ++$i) { if ($DB2::db::debug) { print "Binding ", $i + 1, " => "; if (scalar @{$bind[$i]} > 1 and $bind[$i][1] == SQL_BLOB) { print "[blob],", SQL_BLOB; } else { print join(",", map { defined $_ ? $_ : "" } @{$bind[$i]}); } print "\n"; } $sth->bind_param($i + 1, @{$bind[$i]}); } my $rc = $self->_execute($sth); $sth->finish(); $rc; } =item C The table is what saves a row. If you've made changes to a row, this function will save it. Not really needed since the Row's destructor will save, but doesn't hurt. =cut sub save { my $self = shift; my $obj = shift; my $prep_attr = shift; unless (ref $obj and $obj->isa("DB2::Row")) { croak("Got a " . ref($obj) . " which isn't a 'DB2::Row'"); } if ($self->_already_exists_in_db($obj)) { if ($self->primaryColumn) { $self->_update_db($obj, $prep_attr); } } # else it's new else { $self->_insert_into_db($obj, $prep_attr); } } =item C Commits all current actions =cut sub commit { my $self = shift; $self->_connection->commit; } =item C Deletes the given row from the database. =cut sub delete { my $self = shift; my $obj = shift; my $prep_attr = shift; unless (ref $obj and $obj->isa("DB2::Row")) { croak("Got a " . ref($obj) . " which isn't a 'DB2::Row'"); } if ($self->_already_exists_in_db($obj)) { $self->_delete_db($obj, $prep_attr); } } =item delete_id Deletes a row based on its ID. To delete multiple IDs simultaneously, simply pass in an array ref. =cut sub delete_id { my $self = shift; my $id = shift; my $prep_attr = shift; if (ref $id ne 'ARRAY') { $id = [ $id ]; } if ($self->primaryColumn() and $self->_already_exists_in_db($id)) { my $stmt = 'DELETE FROM ' . $self->full_table_name() . ' WHERE ' . $self->primaryColumn() . ' IN (' . join(',', map { '?' } @$id) . ')'; my $sth = $self->_prepare($stmt, $prep_attr); $self->_execute($sth, @$id); $sth->finish(); } } =item delete_where Deletes rows based on the given WHERE clause. Further parameters are then bound to the DELETE statement. =cut sub delete_where { my $self = shift; my $opts = ref $_[0] eq 'HASH' ? shift : {}; my $where = shift; my $stmt = 'DELETE FROM ' . $self->full_table_name() . ' WHERE ' . $self->_replace_bangs($where); my %prep_attr = exists $opts->{prepare_attributes} ? %{$opts->{prepare_attributes}} : (); my $sth = $self->_prepare($stmt, \%prep_attr); my $rc = $self->_execute($sth, @_); $sth->finish(); $rc; } sub _delete_db { my $self = shift; my $obj = shift; my $prep_attr = shift; my $primcol = $self->primaryColumn; if ($primcol) { my $stmt = 'DELETE FROM ' . $self->full_table_name . ' WHERE ' . $primcol . ' IN ?'; my $sth = $self->_prepare($stmt, $prep_attr); $self->_execute($sth, $obj->column($primcol)); $sth->finish(); $self->commit(); } else { my $stmt = 'DELETE FROM ' . $self->full_table_name . ' WHERE ' . join (' AND ', map { "$_ IN ?" } $self->column_list()); my $sth = $self->_prepare($stmt, $prep_attr); $self->_execute($sth, map { $obj->column($_) } $self->column_list()); $sth->finish(); $self->commit(); } } =item C above =cut sub SELECT_distinct { my $self = shift; my $opts = {}; my $cols = shift; if (ref $cols and ref $cols eq 'HASH') { $opts = $cols; $cols = shift; } $opts->{distinct}++; return $self->SELECT($opts, $cols, @_); } =item C Wrapper around performing an SQL SELECT statement where you may be joining with other tables. The first argument is the columns you want, the second is the tables, and how they are to be joined, while the third is the WHERE condition. Further parameters are bind values. Any text matching '!!!' in the columns text will be replaced with this table's full table name. Any text matching '!(\S+?)!' will be replaced with $1's full table name. =cut sub _replace_bangs { my $self = shift; $_[0] =~ s/!!!/$self->full_table_name()/ge; $_[0] =~ s/!(\S+?)!/$self->getDB()->get_table("$1")->full_table_name()/ge; $_[0]; } sub SELECT_join { my $self = shift; my $opts = {}; my $cols = shift; if (ref $cols and ref $cols eq 'HASH') { $opts = $cols; $cols = shift; } $opts->{tables} = shift; return $self->SELECT($opts, $cols, @_); } =item C The name of this table, excluding schema. This will default to the part of the current package after the last double-colon. For example, if your table is in package "myDB2::foo", then the table name will be "foo". =cut sub table_name { my $self = shift; unless (exists $self->{table_name}) { my $type = ref $self; ( my $tbl = $type ) =~ s/.*::(\w+)/$1/; $self->{table_name} = uc $tbl; } $self->{table_name}; } =item C Shortcut to schema.table_name =cut sub full_table_name { my $self = shift; unless (exists $self->{full_table_name}) { $self->{full_table_name} = uc $self->schema_name . '.' . $self->table_name; } $self->{full_table_name}; } =item C Returns an array of all the column names, in order =cut sub column_list { my $self = shift; if (not exists $self->{column_list}) { $self->{column_list} = [map { $_->{column} } @{$self->_internal_data_order}]; } @{$self->{column_list}} } =item C Returns a hash ref which is all the data from C, but in no particular order (it's a hash, right?). =cut sub all_data { my $self = shift; unless ($self->{ALL_DATA}) { foreach my $h (@{$self->_internal_data_order()}) { $self->{ALL_DATA}{uc $h->{column}} = $h; } } $self->{ALL_DATA} } =item C Gets information about a column or its data. First parameter is the column. Second parameter is the key (NAME, type, etc.). If the key is not given, a hash ref is returned with all the data for this column. If the key is given, only that scalar is returned. =cut sub get_column { my $self = shift; my $column = uc shift; my $data = @_ ? lc shift : undef; my $all_data = $self->all_data; return undef unless exists $all_data->{$column}; if ($data) { exists $all_data->{$column}{$data} ? $all_data->{$column}{$data} : undef; } else { $all_data->{$column}; } } =item C Find the primary column. First time it is called, it will determine the primary column, and then it will cache this for later calls. If you want a table with no primary column, you must override this method to return undef. If no column has the primary attribute, then the last column is defaulted to be the primary column. =cut # Find the primary column (and cache it) sub primaryColumn { my $self = shift; # Check cache. if (not exists $self->{PRIMARY}) { # default to last one. $self->{PRIMARY} = $self->_internal_data_order()->[$#{$self->_internal_data_order()}]{column}; my $data_order = $self->_internal_data_order(); for (my $i = 0; $i < scalar @$data_order; ++$i) { if (exists $data_order->[$i]{primary} and $data_order->[$i]{primary}) { $self->{PRIMARY} = $data_order->[$i]{column}; last; } } } $self->{PRIMARY}; } =item C Determine the generated identity column, if any. This is determined by looking for the string 'GENERATED ALWAYS AS IDENTITY' in the opts of the column. Again, this is cached on first use. =cut sub generatedIdentityColumn { my $self = shift; if (not exists $self->{GENERATEDIDENTITY}) { $self->{GENERATEDIDENTITY} = ''; foreach my $col (@{$self->_internal_data_order()}) { if (exists $col->{generatedidentity} or ( exists $col->{opts} and $col->{opts} =~ /GENERATED ALWAYS AS IDENTITY/i) ) { $self->{GENERATEDIDENTITY} = $col->{column}; last; } } } $self->{GENERATEDIDENTITY}; } =item C Check if the table already exists. Normally only called by create_table. =cut sub table_exists { my $self = shift; my $dbh = $self->_connection; my @tables = $dbh->tables( { TABLE_SCHEM => uc $self->schema_name, TABLE_NAME => uc $self->table_name, } ); die "Unexpected - more than one table with same schema/name!" if scalar @tables > 1; scalar @tables; } # INTERNAL - get current table structure (column names) sub create_table_get_current { my $self = shift; my $dbh = $self->_connection; my @row; if ($self->table_exists()) { my $query = 'SELECT * FROM ' . $self->full_table_name . ' WHERE 1 = 0'; my $sth = $dbh->prepare($query); $self->_execute($sth); @row = @{$sth->{NAME}}; $sth->finish; } @row; } # INTERNAL - common code between CREATE and ALTER - column definitions sub _create_table_column_definition { my $self = shift; my $column = shift; my $tbl = $column->{column} . ' '; $tbl .= uc $column->{type} =~ /(?:NULL)?BOOL/ ? 'CHAR' : $column->{type}; $tbl .= ' (' . $column->{length} . ')' if exists $column->{length}; $tbl .= ' ' . $column->{opts} if $column->{opts}; $tbl .= ' NOT NULL' if ( $column->{primary} or uc $column->{type} ne 'NULLBOOL' and (not $column->{opts} or $column->{opts} !~ /NOT NULL/)) or ($column->{type} eq 'BOOL' and $column->{opts} !~ /NOT NULL/); if (exists $column->{sqldefault}) { $tbl .= ' WITH DEFAULT ' . $column->{sqldefault}; } $tbl .= ' CHECK (' . $column->{column} . q[ in ('Y','N'))] if uc $column->{type} eq 'BOOL'; $tbl .= ' CHECK (' . $column->{column} . q[ in ('Y','N') OR ] . $column->{column} . q[ IS NULL)] if uc $column->{type} eq 'NULLBOOL'; if (exists $column->{generatedidentity}) { $tbl .= ' GENERATED ALWAYS AS IDENTITY '; if (not defined $column->{generatedidentity} or $column->{generatedidentity} eq 'default') { $tbl .= '(START WITH 0, INCREMENT BY 1, NO CACHE)'; } else { $tbl .= $column->{generatedidentity}; } } $self->_replace_bangs($tbl); } # Create the table as given by data_order. =item C Creates the current table. Normally only called by L. =cut sub create_table { my $self = shift; my $dbh = $self->_connection; my %current_col_names = map { $_ => 1 } $self->create_table_get_current(); if (scalar keys %current_col_names == 0) { # new table my $tbl = 'CREATE TABLE ' . $self->full_table_name . ' ('; my @columns; my @constraints; my @foreign_keys; foreach my $f ( $self->column_list ) { my $column = $self->get_column($f); push @columns, $self->_create_table_column_definition($column); if (exists $column->{constraint}) { push @constraints, map { my $x = 'CONSTRAINT ' . $_; $self->_replace_bangs($x); } ref($column->{constraint}) eq 'ARRAY' ? @{$column->{constraint}} : $column->{constraint}; } if (exists $column->{foreignkey}) { push @foreign_keys, map { my $x = 'FOREIGN KEY (' . $column->{column} . ') REFERENCES ' . $_; $self->_replace_bangs($x); } ref($column->{foreignkey}) eq 'ARRAY' ? @{$column->{foreignkey}} : $column->{foreignkey}; } } if ($self->primaryColumn) { push @constraints, 'PRIMARY KEY (' . $self->primaryColumn . ')'; } $tbl .= join(', ', @columns, @constraints, @foreign_keys); $tbl .= ') DATA CAPTURE NONE'; print "$tbl\n"; unless ($dbh->do($tbl)) { print $DBI::err, '[', $DBI::state, '] : ', $DBI::errstr, "\n"; } $self->create_table_initialise('CREATE', $self->column_list()); } else { # existing table - anything need to be updated? my $alter = 'ALTER TABLE ' . $self->full_table_name; my @add = grep { not exists $current_col_names{uc $_} } ($self->column_list); if (scalar @add) { foreach my $add (@add) { my $column = $self->get_column($add); $alter .= ' ADD ' . $self->_create_table_column_definition($column); } print $alter, "\n"; $dbh->do($alter); $self->create_table_initialise('ALTER', @add); } } } =item C A hook that will allow you to initialise the table immediately after its creation. If the table is newly created, the only parameter will be 'CREATE'. If the table is being altered, the first parameter will be 'ALTER' while the rest of the parameters will be the list of columns added. The default action is mildly dangerous. It grants full select, insert, update, and delete authority to the user 'nobody'. This is the user that many daemons, including the default Apache http daemon, run under. If you override this, you can do whatever you want, including nothing. This default was put in primarily because many perl DBI scripts are expected to also be CGI scripts, so this may make certain things easier. This does not change the fact that when this grant is executed you will need some admin authority on the database. =cut sub create_table_initialise { my $self = shift; my $action = shift; if ($action eq 'CREATE') { # default: grant authority to nobody (useful for web apps) my $grant = 'GRANT SELECT,INSERT,UPDATE,DELETE ON TABLE ' . $self->full_table_name . ' TO USER NOBODY'; $self->_connection->do($grant); } } =back =cut 1;