SPOPS::SQLInterface - Generic routines for DBI database interaction


SPOPS documentation Contained in the SPOPS distribution.

Index


Code Index:

NAME

Top

SPOPS::SQLInterface - Generic routines for DBI database interaction

SYNOPSIS

Top

 # Make this class a parent of my class

 package My::DBIStuff;
 use SPOPS::SQLInterface;
 @My::DBIStuff::ISA = qw( SPOPS::SQLInterface );

 # You should also be able to use it directly, but you
 # need to pass in a database object with every request

 use SPOPS::SQLInterface;
 my $dbc = 'SPOPS::SQLInterface';
 my $db = DBI->connect( ... ) || die $DBI::errstr;
 my $rows = $dbc->db_select({ select => [ qw/ uid first_name last_name / ],
                              from   => [ 'users' ],
                              where  => 'first_name = ? or last_name = ?',
                              value  => [ 'fozzie', "th' bear" ],
                              db     => $db });
 foreach my $row ( @{ $results } ) {
   print "User ID $row->[0] is $row->[1] $row->[2]\n";
 }

DESCRIPTION

Top

You are meant to inherit from this class, although you can use it as a standalone SQL abstraction tool as well, as long as you pass the database handle into every routine you call.

DATABASE METHODS

Top

Relatively simple methods to do the select, update, delete and insert statements, with the right values and table names being passed in.

All parameters are passed in via named values, such as:

 $t->db_select({ select => [ 'this', 'that' ],
                 from   => [ 'mytable' ] });

VERY IMPORTANT

The subclass that uses these methods must either pass in a DBI database handle via a named parameter (db) or make it available through a method of the class called 'global_datasource_handle'.

METHODS

Top

There are very few methods in this class, but each one can do quite a bit.

sql_quote( $value, $type, $db )

Quotes a value for insertion/update or selection when bound parameters are inappropriate or unavailable. Returns 'NULL' if $value is undef, otherwise calls quote( $value, $type ) on the DBI handle $db.

Drivers that do not implement the two-argument form of quote() should override this method.

Returns: quoted $value appropriate to use in a SQL statement

db_select( \%params )

Executes a SELECT. Return value depends on what you ask for. Many of the parameters are optional unless you pass in SQL to execute.

Parameters:

sql ($) (optional)

Full statement to execute, although you may put '?' in the where clause and pass values for substitution. (No quoting hassles...)

select (\@) (optional unless 'sql' defined)

Fields to select

select_modifier ($) (optional)

Clause to insert between 'SELECT' and fields (e.g., DISTINCT)

from (\@ or $) (optional unless 'sql' defined)

List of tables to select from. (You can pass a single tablename as a scalar if you wish.)

order ($) (optional)

Clause to order results by; if not given, the order depends entirely on the database.

group ($) (optional)

Clause to group results by (in a 'GROUP BY' clause). This is normally only done with 'COUNT(*)' and such features. See your favorite SQL reference for more info.

where ($) (optional unless 'sql' defined)

Clause to limit results. Note that you can use '?' for field values but they will get quoted as if they were a SQL_VARCHAR type of value.

return ($) (optional)

What the method should return. Potential values are:

value (\@) (optional unless you use '?' placeholders)

List of values to bind, all as SQL_VARCHAR; they must match order of '?' in the where clause either passed in or within the SQL statement passed in.

Examples:

Perl statement:

 $t->db_select( { select => [ qw/ first_name last_name /],
                  from   => [ 'users' ],
                  where  => 'last_name LIKE ?',
                  value  => 'moo%' } );

SQL statement:

 SELECT first_name, last_name
   FROM users
  WHERE last_name LIKE 'moo%'

Returns:

 [ [ 'stephen', 'moore' ],
   [ 'charles', 'mooron' ],
   [ 'stacy', 'moonshine' ] ]

Perl statement:

 $t->db_select( { select => [ qw/ u.username l.login_date / ],
                  from   => [ 'users u', 'logins l' ],
                  where  => "l.login_date > '2000-04-18' and u.uid = l.uid"
                  return => 'hash' } );

SQL statement:

 SELECT u.username, l.login_date
   FROM users u, logins l
  WHERE l.login_date > '2000-04-18' and u.uid = l.uid

Returns:

 [ { username => 'smoore',
     login_date => '2000-05-01' },
   { username => 'cmooron',
     login_date => '2000-04-19' },
   { username => 'smoonshine',
     login_date => '2000-05-02' } ]

Perl statement:

 $t->db_select( { select => [ qw/ login_name first_name last_name /],
                  from   => [ 'users' ],
                  where  => 'last_name LIKE ?',
                  value  => 'moo%',
                  return => 'single-list' } );

SQL statement:

 SELECT login_name, first_name, last_name
   FROM users
  WHERE last_name LIKE 'moo%'

Returns:

 [ 'smoore',
   'cmooron',
   'smoonshine' ]




db_insert( \%params )

Create and execute an INSERT statement given the parameters passed in. Return value is true is insert was successful -- the exact value is whatever is returned from the execute() statement handle call from your database. (See DBI and your driver docs.)

Parameters:

sql ($) (optional)

Full SQL statement to run; you can still pass in values to quote/bind if you use '?' in the statement.

table ($) (optional unless 'sql' defined)

Name of table to insert into

field (\@) (optional unless 'sql' defined)

List of fieldnames to insert

value (\@) (optional unless you use '?' placeholders)

List of values, matching up with order of field list.

no_quote (\%) (optional)

Fields that we should not quote

return_sth ($) (optional)

If true, return the statement handle rather than a status.

Examples:

Perl statement:

 $t->db_insert( { table => 'users',
                  field => [ qw/ username first_name last_name password / ],
                  value => [ 'cmw817', "Chris O'Winters" ] } );

SQL statement:

 INSERT INTO users
 ( username, first_name, last_name, password )
 VALUES
 ( 'cmw817', 'Chris', 'O''Winters', NULL )

Perl statement:

 my $sql = qq/
   INSERT INTO users ( username ) VALUES ( ? )
 /;

 foreach my $username ( qw/ chuck stinky jackson / ) {
   $t->db_insert({ sql   => $sql,
                   value => [ $username ] } );
 }

SQL statements:

 INSERT INTO users ( username ) VALUES ( 'chuck' )
 INSERT INTO users ( username ) VALUES ( 'stinky' )
 INSERT INTO users ( username ) VALUES ( 'jackson' )

db_update( \%params )

Create and execute an UPDATE statement given the parameters passed in. Return value is true is update was successful -- the exact value is whatever is returned from the execute() statement handle call from your database, which many times is the number of rows affected by the update. (See DBI and your driver docs -- in particular, note that the return value from an UPDATE can vary depending on the database being used as well as the number of records actually updated versus those that matched the criteria but were not updated because they already matched the value(s). In particular, see the discussion in DBD::mysql under 'mysql_client_found_rows'.)

Parameters:

sql ($) (optional)

Full SQL statement to run; note that you can use '?' for values and pass in the raw values via the 'value' parameter, and they will be quoted as necessary.

field (\@) (optional unless 'sql' defined)

List of fieldnames we are updating

value (\@) (optional unless you use '?' placeholders)

List of values corresponding to the fields we are updating and to parameters in the WHERE clause.

table ($) (optional unless 'sql' defined)

Name of table we are updating

where ($) (optional unless 'sql' defined)

Clause that specifies the rows we are updating

no_quote (\%) (optional)

Specify fields not to quote

Examples:

Perl statement:

 $t->db_update( { field => [ qw/ first_name last_name / ],
                  value => [ 'Chris', "O'Donohue" ],
                  table => 'users',
                  where => 'user_id = 98172' } );

SQL statement (assuming "'" gets quoted as "''"):

 UPDATE users
    SET first_name = 'Chris',
        last_name = 'O''Donohue',
  WHERE user_id = 98172

db_delete( \%params )

Removes the record indicated by \%params from the database. Return value is true is delete was successful -- the exact value is whatever is returned from the execute() statement handle call from your database. (See DBI)

Parameters:

sql ($) (optional)

Full SQL statement to execute directly, although you can use '?' for values and pass the actual values in via the 'value' parameter.

table ($) (optional unless 'sql' defined)

Name of table from which we are removing records.

where ($) (optional unless 'sql' defined)

Specify the records we are removing. Be careful: if you pass in the table but not the criteria, you will clear out your table! (Just like real SQL...)

value (\@) (optional unless you use '?' placeholders)

List of values to bind to '?' that may be found either in the where clause passed in or in the where clause found in the SQL statement.

Examples:

Perl statement:

 $t->db_delete( { table => 'users', where => 'user_id = 98172' } );

SQL statement:

 DELETE FROM users
  WHERE user_id = 98172

Perl statement:

 $t->db_delete( { table => 'users', where => 'last_name LIKE ?',
                  value => [ 'moo%' ] } );

SQL statement:

 DELETE FROM users
  WHERE last_name LIKE 'moo%'

Perl statement:

 $t->db_delete( { table => 'users' } );

SQL statement:

 DELETE FROM users

Oops, just cleared out the 'users' table. Be careful!

db_discover_types( $table, \%params )

Retrieve field type information for $table. Normally we simply issue a dummy query to a particular table to get its schema -- field names and field types. We cache the information (in a SPOPS::DBI::TypeInfo object) and then query it for the different field types as we need them.

If a DBD driver does not support the {TYPE} attribute of the DBI statement handle, you have to specify some simple types in your class configuration or provide them either in the call to db_discover_types() or in the object configuration using the key 'dbi_type_info' as listed in the parameters below. (See SPOPS::DBI::TypeInfo for an enumeration of the the values to use.)

Return a SPOPS::DBI::TypeInfo object for $table.

Parameters:

table ($)

The name of a particular table. Note that this routine is not smart enough to distinguish between: users and dbo.users even though they might refer to the same table in the database. It is not harmful if you use the same name twice in this manner, the module just has to do a little extra work.

Other parameters:

db (object) (optional)

DBI database handle. (Optional only if you have a global_datasource_handle() class method defined.

dbi_type_info (\%) (optional)

If your DBD driver cannot retrieve type information from the database, you need to give this module a hint as to what type of datatypes you will be working with. See Fake Types in SPOPS::DBI::TypeInfo for the types you can use.

Example:

  my $type_info = $class->db_discover_types( $class->table_name );
  foreach my $field ( $type_info->get_fields ) {
      print "$field is DBI type ", $type_info->get_type( $field ), "\n";
  }

ERROR HANDLING

Top

All errors encountered by this module throw a SPOPS::Exception object (in case of a lack of required information) or, in most cases, a SPOPS::Exception::DBI object.

TO DO

Top

DBI binding conventions

One of the things the DBI allows you to do is prepare a statement once and then execute it many times -- particularly useful for INSERTs and UPDATEs. It would be nice to be able to do that.

Datasource Names

Be able to pass a name to 'global_datasource_handle' (and to pass in that name to the relevant 'db_*' calls).

BUGS

Top

None known.

SEE ALSO

Top

DBI

COPYRIGHT

Top

AUTHORS

Top

Chris Winters <chris@cwinters.com>

See the SPOPS module for the full author list.


SPOPS documentation Contained in the SPOPS distribution.

package SPOPS::SQLInterface;

# $Id: SQLInterface.pm,v 3.12 2004/06/02 00:48:21 lachoy Exp $

use strict;
use Data::Dumper          qw( Dumper );
use DBI;
use Log::Log4perl         qw( get_logger );
use SPOPS::DBI::TypeInfo;
use SPOPS::Exception      qw( spops_error );
use SPOPS::Exception::DBI qw( spops_dbi_error );

my $log = get_logger();

$SPOPS::SQLInterface::VERSION = sprintf("%d.%02d", q$Revision: 3.12 $ =~ /(\d+)\.(\d+)/);

my %TYPE_INFO = ();

sub throw_no_database_handle_error {
    my ( $item ) = @_;
    my $class = ref $item || $item;
    spops_error "No database handle available; pass using the 'db' ",
                "parameter or define ${class}->global_datasource_handle ",
                "to return a valid DBI handle.";
}

# Drivers that don't support the two-argument form of DBI->quote
# should override...

sub sql_quote {
    my ( $class, $value, $type, $db ) = @_;
    return 'NULL' unless ( defined $value );
    $db ||= $class->global_datasource_handle;
    unless ( ref $db ) {
        spops_error "No database handle could be found!";
    }

    # This issues a warning when '$type' is unknown; need to figure
    # out how to set '$type' to a known value

    return $db->quote( $value, $type );
}

# Note: not sure how to integrate the fieldtype discovery
# stuff in here. What if you do:
#
# select u.username from users u, logins l
#  where l.login_date > ?  and
#        l.user_id = u.user_id
#
# and pass '2000/1/13' is passed in? This seems to be a little
# too much in-depth sql processing than this library is
# appropriate for; so you can still pass in values for binding,
# but they'll all be bound with SQL_VARCHAR
#
# select   => \@ of fields to select
# select_modifier => $ to insert between 'SELECT' and fields (e.g., DISTINCT)
# from     => \@ of tables to select from ($ allowed if only one)
# order    => $ clause to order by
# group    => $ clause to group by
# where    => $ clause to limit results
# return   => sth | single | list | hash | single-list
# value    => \@ of values to bind, all as SQL_VARCHAR; they must match order of '?' in where
# sql      => $ statement to execute

sub db_select {
    my ( $class, $p ) = @_;
    my $db    = $p->{db} || $class->global_datasource_handle;
    $class->throw_no_database_handle_error unless ( $db );

  # Don't do anything if the SQL isn't passed in and you don't have
  # either a list of fields to select or a table to select them from

    unless ( $p->{sql} or ( $p->{select} and $p->{from} ) ) {
        spops_error 'Cannot run without select/from statements!';
    }

    $log->is_debug &&
        $log->debug( "Entering db_select with ", Dumper( $p ) );
    $p->{return} ||= 'list';
    $p->{value}  ||= [];
    my $sql = $p->{sql};

    # If we don't have any SQL, build it (straightforward).

    unless ( $sql ) {
        $log->is_info &&
            $log->info( "No SQL passed in to execute directly; building." );

        $p->{from} ||= $p->{table}; # allow an alias
        if ( $p->{from} and ref $p->{from} ne 'ARRAY' ) {
            $p->{from} = [ $p->{from} ];
        }

        $p->{select_modifier} ||= '';
        my $select = join ', ', @{ $p->{select} };
        my $from   = join ', ', @{ $p->{from} };
        my $order  = ( $p->{order} ) ? "ORDER BY $p->{order}" : '';
        my $where  = ( $p->{where} ) ? "WHERE $p->{where}" : '';
        my $group  = ( $p->{group} ) ? "GROUP BY $p->{group}" : '';
        $sql = qq/
                      SELECT $p->{select_modifier} $select
                          FROM $from
                        $where
                        $group
                        $order
                /;
    }
    $log->is_info &&
        $log->info( "SQL for select: $sql" );

    # First prepare and check for errors...

    my $sth = eval { $db->prepare( $sql ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'prepare' };
    }

    # Execute with any bound parameters; note that for Sybase you do
    # not need to pass any types at all.

    $log->is_info &&
        $log->info( "Values bound: ", join( '//', @{ $p->{value} } ) );
    eval { $sth->execute( @{ $p->{value} } ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql         => $sql,
                              bound_value => $p->{value},
                              action      => 'execute' };
    }

    # If they asked for the handle back, give it to them

    if ( $p->{return} eq 'sth' ) {
        $log->is_info &&
            $log->info( "Returning statement handle (after prepare/execute)" );
        return $sth;
    }

    # If they asked for a single row, return it in arrayref format [
    # field1, field2, ...]

    if ( $p->{return} eq 'single' ) {
        $log->is_info &&
            $log->info( "Returning single row." );
        my $row =  eval { $sth->fetchrow_arrayref; };
        if ( $@ ) {
            spops_dbi_error $@, { sql         => $sql,
                                  bound_value => $p->{value},
                                  action      => 'fetchrow_arrayref' };
        }
        return $row;
    }

    # If they asked for a list of results, return an arrayref of arrayrefs

    elsif ( $p->{return} eq 'list' ) {
        $log->is_info &&
            $log->info( "Returning list of lists." );
        my $rows = eval { $sth->fetchall_arrayref; };
        if ( $@ ) {
            spops_dbi_error $@, { sql         => $sql,
                                  bound_value => $p->{value},
                                  action      => 'fetchall_arrayref' };
        }
        return $rows;
    }

    # return the first element of each record in an arrayref

    elsif ( $p->{return} eq 'single-list' ) {
        $log->is_info &&
            $log->info( "Returning list of single items." );
        my $rows = eval { $sth->fetchall_arrayref };
        if ( $@ ) {
            spops_dbi_error $@, { sql         => $sql,
                                  bound_value => $p->{value},
                                  action      => 'fetchall_arrayref' };
        }
        return [ map { $_->[0] } @{ $rows } ];
    }

    # If they asked for a hash, return a list of hashrefs

    elsif ( $p->{return} eq 'hash' ) {
        $log->is_info &&
            $log->info( "Returning list of hashrefs." );
        my @rows = ();

        # Note -- we may need to change this to zip through $row every
        # time and push a new reference onto @rows

        eval {
            while ( my $row = $sth->fetchrow_hashref ) {
                push @rows, \%{ $row };
            }
        };
        if ( $@ ) {
            spops_dbi_error $@, { sql         => $sql,
                                  bound_value => $p->{value},
                                  action      => 'fetchall_arrayref' };
        }
        return \@rows;
    }
    return [];
}

# field    => \@ of fieldnames
# value    => \@ of values
# table    => $ of table to insert into
# no_quote => \% of fields not to quote
# sql      => $ of sql to run
# return_sth => $ if true, return the statement handle rather than status

sub db_insert {
    my ( $class, $p ) = @_;
    my $db    = $p->{db} || $class->global_datasource_handle;
    $class->throw_no_database_handle_error unless ( $db );

    $log->is_debug &&
        $log->debug( "Enter insert procedure\n", Dumper( $p ) );

    # If we weren't given direct sql or a list of values or table, bail

    unless ( $p->{sql} or ( $p->{value} and $p->{table} ) ) {
        spops_error 'Cannot continue with no SQL, values or table name';
    }

    my $sql = $p->{sql};

    # If we weren't given SQL, build it.

    unless ( $sql ) {

        # Find the types for all fields in this table (we don't have
        # to use them all...); let any errors trickle up

        my $type_info = $class->db_discover_types(
                                   $p->{table},
                                   { dbi_type_info => $p->{dbi_type_info},
                                     db            => $db });

        my ( $fields, $values );

        # Be sure these are at least empty hashrefs, otherwise we
        # will get an error

        $p->{no_quote} ||= {};
        $p->{field}    ||= [];
        $p->{value}    ||= [];
        $log->is_debug &&
            $log->debug( "Fields/values: ", Dumper( $p->{field}, $p->{value} ) );

        # Cycle through the fields and values, creating lists
        # suitable for join()ing into the SQL statement.

        my @value_list = ();
        my $count = 0;
        foreach my $field ( @{ $p->{field} } ) {
            next unless ( $field );

            # Quote the value unless the user asked us not to
            my $value = ( $p->{no_quote}{ $field } )
                          ? $p->{value}->[ $count ]
                          : $class->sql_quote( $p->{value}->[ $count ],
                                               $type_info->get_type( $field ),
                                               $db );
            $log->is_info &&
                $log->info( "Trying to add quoted value [$value] ",
                                      "for field [$field]" );
            push @value_list, $value;
            $count++;
        }

        my $field_listing = join ', ', @{ $p->{field} };
        my $value_listing = join ', ', @value_list;
        $sql = qq/
                      INSERT INTO $p->{table}
                      ( $field_listing )
                      VALUES
                      ( $value_listing )
                /;
    }

    # Note that we use the prepare()/execute() method of
    # getting this data in rather than the simpler do(),
    # since the user might want the statement handle afterward;
    # if this becomes a performance hang (doubtful), we can only
    # do p/e if the user's asked for the statement handle

    $log->is_info &&
        $log->info( "Preparing\n$sql" );
    my $sth = eval { $db->prepare( $sql ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'prepare' };
    }

    my $rv = eval { $sth->execute };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'execute' };
    }
    $log->is_info &&
        $log->info( "Prepare/execute went ok." );

    # Everything is ok; return either a true value
    # or the statement handle, if they've asked for it.

    return $sth   if ( $p->{return_sth} );
    return $rv;
}


# field    => \@ of fieldnames
# value    => \@ of values (includes both update values and WHERE params)
# table    => $ of table to insert into
# where    => $ clause for which we're updating
# no_quote => \% of fields not to quote
# sql      => $ of sql to run

sub db_update {
    my ( $class, $p ) = @_;
    my $db    = $p->{db} || $class->global_datasource_handle;
    $class->throw_no_database_handle_error unless ( $db );

  # If we weren't given direct sql or a list of values or table, bail

    unless ( $p->{sql} or ( $p->{value} and $p->{table} ) ) {
        spops_error 'Cannot continue with no SQL, values or table name';
    }
    my $sql = $p->{sql};

    my @values = ();
    $p->{value} ||= [];

    # If SQL provided still pass the values as bound parameters

    if ( $sql ) {
        @values = @{ $p->{value} };
    }

    # Build the SQL

    unless ( $sql ) {

        # Find the types for all fields in this table (we don't have
        # to use them all...); let the error trickle up

        my $type_info = $class->db_discover_types(
                                       $p->{table},
                                       { dbi_type_info => $p->{dbi_type_info},
                                         db            => $db } );
        my ( @update );

        # make a copy of the values given so we can modify the list as
        # we go -- it may include both update parameters and WHERE
        # parameters

        @values = @{ $p->{value} };

        # Go through each field and setup an update assign subset
        # for each; most of them get a bound parameter and push the
        # value onto the stack, but values that cannot be bound push
        # the direct information onto the stack.

        $p->{no_quote} ||= {};
        foreach my $field ( @{ $p->{field} } ) {
            my $rawval = shift @values;
            $log->is_info &&
                $log->info( "Trying to add value [", defined $rawval ? $rawval : '', "] ",
                                      "with field [$field] and type ",
                                      "[", $type_info->get_type( $field ), "]" );

            # Quote the value unless the user asked us not to

            my $value = ( $p->{no_quote}{ $field } )
                          ? $rawval
                          : $class->sql_quote( $rawval,
                                               $type_info->get_type( $field ),
                                               $db );
            push @update, "$field = $value";
        }
        my $update = join ', ', @update;
        my $where  = ( $p->{where} ) ? "WHERE $p->{where}" : '';
        $sql = qq/
                      UPDATE $p->{table}
                            SET $update
                        $where
                /;
    }
    $log->is_info &&
        $log->info( "Prepare/execute\n$sql" );
    my $sth = eval { $db->prepare( $sql ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'prepare' };
    }

    # assign the remainder of the values in case the 'WHERE' clause
    # included any

    my $rv = eval { $sth->execute( @values ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'execute' };
    }
    return $rv;
}


# table  => $ of table we're deleting from
# where  => $ limiting our deletes
# value  => \@ of values to bind
# sql    => $ of statement to execute directly

sub db_delete {
    my ( $class, $p ) = @_;
    my $db    = $p->{db} || $class->global_datasource_handle;
    $class->throw_no_database_handle_error unless ( $db );

    # Gotta have a table to delete from

    unless ( $p->{table} or $p->{sql} ) {
        spops_error 'Cannot delete records without SQL or a table name';
    }

    # If we weren't given SQL, build it.

    my $sql = $p->{sql};
    unless ( $sql ) {

        # Hopefully you'll have a WHERE clause... but we'll let
        # you shoot yourself in the foot if you forget :)

        my $where = ( $p->{where} ) ? "WHERE $p->{where}" : '';
        $sql = qq/ DELETE FROM $p->{table} $where /;
    }

    $log->is_info &&
        $log->info( "SQL for DELETE:\n$sql" );
    $p->{value} ||= [];
    my $sth = eval { $db->prepare( $sql ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql => $sql, action => 'prepare' };
    }

    my $rv = eval { $sth->execute( @{ $p->{value} } ) };
    if ( $@ ) {
        spops_dbi_error $@, { sql         => $sql,
                              bound_value => $p->{value},
                              action      => 'execute' };
    }
    return $rv;
}



sub db_discover_types {
    my ( $class, $table, $p ) = @_;

    # Create the index used to find the table info later

    my $db       = $p->{db} || $class->global_datasource_handle;
    $class->throw_no_database_handle_error unless ( $db );

    my $db_name = eval { $db->{Name} } || eval { $db->{name} };
    my $type_idx = join( '-', lc $db_name , lc $table );
    $log->is_debug &&
        $log->debug( "Type index used to discover data types: [$type_idx]" );

    # If we've already discovered the types, get the cached copy

    return $TYPE_INFO{ $type_idx } if ( $TYPE_INFO{ $type_idx } );

    my $type_info = SPOPS::DBI::TypeInfo->new({ database => $db_name,
                                                table    => $table });

    my $conf = eval { $class->CONFIG };
    my $fake_types = $p->{dbi_type_info} || $conf->{dbi_type_info};

    # Certain databases (or more specifically, DBD drivers) do not
    # process $sth->{TYPE} requests properly, so we need the user to
    # specify the types by hand

    if ( ref $fake_types eq 'HASH' ) {
        while ( my ( $field, $fake_type ) = each %{ $fake_types } ) {
            $type_info->add_type( $field, $fake_type );
        }
    }

    # Otherwise, fetch the types from the database

    else {
        my $sql = $class->sql_fetch_types( $table );
        $type_info->fetch_types( $db, $sql );
    }

    return $TYPE_INFO{ $type_idx } = $type_info
}


# Default data type discovery statement

sub sql_fetch_types { return "SELECT * FROM $_[1] where 1 = 0" }

1;

__END__