^^&&^^

Edit File: Developers.pod

=head1 NAME

DBD::File::Developers - Developers documentation for DBD::File

=head1 SYNOPSIS

package DBD::myDriver;

use base qw( DBD::File );

sub driver
    {
	...
	my $drh = $proto->SUPER::driver ($attr);
	...
	return $drh->{class};
	}

sub CLONE { ... }

package DBD::myDriver::dr;

@ISA = qw( DBD::File::dr );

sub data_sources { ... }
    ...

package DBD::myDriver::db;

@ISA = qw( DBD::File::db );

sub init_valid_attributes { ... }
    sub init_default_attributes { ... }
    sub set_versions { ... }
    sub validate_STORE_attr { my ($dbh, $attrib, $value) = @_; ... }
    sub validate_FETCH_attr { my ($dbh, $attrib) = @_; ... }
    sub get_myd_versions { ... }

package DBD::myDriver::st;

@ISA = qw( DBD::File::st );

sub FETCH { ... }
    sub STORE { ... }

package DBD::myDriver::Statement;

@ISA = qw( DBD::File::Statement );

package DBD::myDriver::Table;

@ISA = qw( DBD::File::Table );

my %reset_on_modify = (
        myd_abc => "myd_foo",
        myd_mno => "myd_bar",
        );
    __PACKAGE__->register_reset_on_modify (\%reset_on_modify);
    my %compat_map = (
        abc => 'foo_abc',
        xyz => 'foo_xyz',
        );
    __PACKAGE__->register_compat_map (\%compat_map);

sub bootstrap_table_meta { ... }
    sub init_table_meta { ... }
    sub table_meta_attr_changed { ... }
    sub open_data { ... }

sub fetch_row { ... }
    sub push_row { ... }
    sub push_names { ... }

# optimize the SQL engine by add one or more of
    sub update_current_row { ... }
    # or
    sub update_specific_row { ... }
    # or
    sub update_one_row { ... }
    # or
    sub insert_new_row { ... }
    # or
    sub delete_current_row { ... }
    # or
    sub delete_one_row { ... }

=head1 DESCRIPTION

This document describes how DBD developers can write DBD::File based DBI
drivers. It supplements L<DBI::DBD> and L<DBI::DBD::SqlEngine::Developers>,
which you should read first.

=head1 CLASSES

Each DBI driver must provide a package global C<driver> method and three
DBI related classes:

=over 4

=item DBD::File::dr

Driver package, contains the methods DBI calls indirectly via DBI
interface:

DBI->connect ('DBI:DBM:', undef, undef, {})

# invokes
  package DBD::DBM::dr;
  @DBD::DBM::dr::ISA = qw( DBD::File::dr );

sub connect ($$;$$$)
  {
      ...
      }

Similar for C<< data_sources >> and C<< disconnect_all >>.

Pure Perl DBI drivers derived from DBD::File do not usually need to
override any of the methods provided through the DBD::XXX::dr package
however if you need additional initialization in the connect method
you may need to.

=item DBD::File::db

Contains the methods which are called through DBI database handles
(C<< $dbh >>). e.g.,

$sth = $dbh->prepare ("select * from foo");
  # returns the f_encoding setting for table foo
  $dbh->csv_get_meta ("foo", "f_encoding");

DBD::File provides the typical methods required here. Developers who
write DBI drivers based on DBD::File need to override the methods C<<
set_versions >> and C<< init_valid_attributes >>.

=item DBD::File::st

Contains the methods to deal with prepared statement handles. e.g.,

$sth->execute () or die $sth->errstr;

=back

=head2 DBD::File

This is the main package containing the routines to initialize
DBD::File based DBI drivers. Primarily the C<< DBD::File::driver >>
method is invoked, either directly from DBI when the driver is
initialized or from the derived class.

package DBD::DBM;

use base qw( DBD::File );

sub driver
  {
      my ($class, $attr) = @_;
      ...
      my $drh = $class->SUPER::driver ($attr);
      ...
      return $drh;
      }

It is not necessary to implement your own driver method as long as
additional initialization (e.g. installing more private driver
methods) is not required.  You do not need to call C<< setup_driver >>
as DBD::File takes care of it.

=head2 DBD::File::dr

The driver package contains the methods DBI calls indirectly via the DBI
interface (see L<DBI/DBI Class Methods>).

DBD::File based DBI drivers usually do not need to implement anything here,
it is enough to do the basic initialization:

package DBD:XXX::dr;

@DBD::XXX::dr::ISA = qw (DBD::File::dr);
  $DBD::XXX::dr::imp_data_size     = 0;
  $DBD::XXX::dr::data_sources_attr = undef;
  $DBD::XXX::ATTRIBUTION = "DBD::XXX $DBD::XXX::VERSION by Hans Mustermann";

=head2 DBD::File::db

This package defines the database methods, which are called via the DBI
database handle C<< $dbh >>.

Methods provided by DBD::File:

=over 4

=item ping

Simply returns the content of the C<< Active >> attribute. Override
when your driver needs more complicated actions here.

=item prepare

Prepares a new SQL statement to execute. Returns a statement handle,
C<< $sth >> - instance of the DBD:XXX::st. It is neither required nor
recommended to override this method.

=item FETCH

Fetches an attribute of a DBI database object. Private handle attributes
must have a prefix (this is mandatory). If a requested attribute is
detected as a private attribute without a valid prefix, the driver prefix
(written as C<$drv_prefix>) is added.

The driver prefix is extracted from the attribute name and verified against
C<< $dbh->{$drv_prefix . "valid_attrs"} >> (when it exists). If the
requested attribute value is not listed as a valid attribute, this method
croaks. If the attribute is valid and readonly (listed in C<< $dbh->{
$drv_prefix . "readonly_attrs" } >> when it exists), a real copy of the
attribute value is returned. So it's not possible to modify
C<f_valid_attrs> from outside of DBD::File::db or a derived class.

=item STORE

Stores a database private attribute. Private handle attributes must have a
prefix (this is mandatory). If a requested attribute is detected as a private
attribute without a valid prefix, the driver prefix (written as
C<$drv_prefix>) is added. If the database handle has an attribute
C<${drv_prefix}_valid_attrs> - for attribute names which are not listed in
that hash, this method croaks. If the database handle has an attribute
C<${drv_prefix}_readonly_attrs>, only attributes which are not listed there
can be stored (once they are initialized). Trying to overwrite such an
immutable attribute forces this method to croak.

An example of a valid attributes list can be found in
C<< DBD::File::db::init_valid_attributes >>.

=item set_versions

This method sets the attribute C<f_version> with the version of DBD::File.

This method is called at the begin of the C<connect ()> phase.

When overriding this method, do not forget to invoke the superior one.

=item init_valid_attributes

This method is called after the database handle is instantiated as the
first attribute initialization.

C<< DBD::File::db::init_valid_attributes >> initializes the attributes
C<f_valid_attrs> and C<f_readonly_attrs>.

When overriding this method, do not forget to invoke the superior one,
preferably before doing anything else. Compatibility table attribute
access must be initialized here to allow DBD::File to instantiate the
map tie:

# for DBD::CSV
    $dbh->{csv_meta} = "csv_tables";
    # for DBD::DBM
    $dbh->{dbm_meta} = "dbm_tables";
    # for DBD::AnyData
    $dbh->{ad_meta}  = "ad_tables";

=item init_default_attributes

This method is called after the database handle is instantiated to
initialize the default attributes.

C<< DBD::File::db::init_default_attributes >> initializes the attributes
C<f_dir>, C<f_meta>, C<f_meta_map>, C<f_version>.

When the derived implementor class provides the attribute to validate
attributes (e.g. C<< $dbh->{dbm_valid_attrs} = {...}; >>) or the attribute
containing the immutable attributes (e.g.
C<< $dbh->{dbm_readonly_attrs} = {...}; >>), the attributes
C<drv_valid_attrs>, C<drv_readonly_attrs>, C<drv_version> and C<drv_meta>
are added (when available) to the list of valid and immutable attributes
(where C<drv_> is interpreted as the driver prefix).

If C<drv_meta> is set, an attribute with the name in C<drv_meta> is
initialized providing restricted read/write access to the meta data of the
tables using C<DBD::File::TieTables> in the first (table) level and
C<DBD::File::TieMeta> for the meta attribute level. C<DBD::File::TieTables>
uses C<DBD::DRV::Table::get_table_meta> to initialize the second level
tied hash on FETCH/STORE. The C<DBD::File::TieMeta> class uses
C<DBD::DRV::Table::get_table_meta_attr> to FETCH attribute values and
C<DBD::DRV::Table::set_table_meta_attr> to STORE attribute values. This
allows it to map meta attributes for compatibility reasons.

=item get_single_table_meta

=item get_file_meta

Retrieve an attribute from a table's meta information. The method
signature is C<< get_file_meta ($dbh, $table, $attr) >>. This method
is called by the injected db handle method C<< ${drv_prefix}get_meta >>.

While get_file_meta allows C<$table> or C<$attr> to be a list of tables or
attributes to retrieve, get_single_table_meta allows only one table name
and only one attribute name. A table name of C<'.'> (single dot) is
interpreted as the default table and this will retrieve the appropriate
attribute globally from the dbh. This has the same restrictions as
C<< $dbh->{$attrib} >>.

get_file_meta allows C<'+'> and C<'*'> as wildcards for table names and
C<$table> being a regular expression matching against the table names
(evaluated without the default table). The table name C<'*'> is
I<all currently known tables, including the default one>. The table
name C<'+'> is I<all table names which conform to
ANSI file name restrictions> (/^[_A-Za-z0-9]+$/).

The table meta information is retrieved using the get_table_meta and
get_table_meta_attr methods of the table class of the implementation.

=item set_single_table_meta

=item set_file_meta

Sets an attribute in a table's meta information. The method signature is
C<< set_file_meta ($dbh, $table, $attr, $value) >>. This method is called
by the injected db handle method C<< ${drv_prefix}set_meta >>.

While set_file_meta allows C<$table> to be a list of tables and C<$attr>
to be a hash of several attributes to set, set_single_table_meta allows
only one table name and only one attribute name/value pair.

The wildcard characters for the table name are the same as for
get_file_meta.

The table meta information is updated using the get_table_meta and
set_table_meta_attr methods of the table class of the implementation.

=item clear_file_meta

Clears all meta information cached about a table. The method signature is
C<< clear_file_meta ($dbh, $table) >>. This method is called
by the injected db handle method C<< ${drv_prefix}clear_meta >>.

=back

=head2 DBD::File::st

Contains the methods to deal with prepared statement handles:

=over 4

=item FETCH

Fetches statement handle attributes. Supported attributes (for full overview
see L<DBI/Statement Handle Attributes>) are C<NAME>, C<TYPE>, C<PRECISION>
and C<NULLABLE> in case that SQL::Statement is used as SQL execution engine
and a statement is successful prepared.  When SQL::Statement has additional
information about a table, those information are returned. Otherwise, the
same defaults as in L<DBI::DBD::SqlEngine> are used.

This method usually requires extending in a derived implementation.
See L<DBD::CSV> or L<DBD::DBM> for some example.

=back

=head2 DBD::File::TableSource::FileSystem

Provides data sources and table information on database driver and database
handle level.

package DBD::File::TableSource::FileSystem;

sub data_sources ($;$)
  {
      my ($class, $drh, $attrs) = @_;
      ...
      }

sub avail_tables
  {
      my ($class, $drh) = @_;
      ...
      }

The C<data_sources> method is called when the user invokes any of the
following:

@ary = DBI->data_sources ($driver);
  @ary = DBI->data_sources ($driver, \%attr);
  
  @ary = $dbh->data_sources ();
  @ary = $dbh->data_sources (\%attr);

The C<avail_tables> method is called when the user invokes any of the
following:

@names = $dbh->tables ($catalog, $schema, $table, $type);
  
  $sth   = $dbh->table_info ($catalog, $schema, $table, $type);
  $sth   = $dbh->table_info ($catalog, $schema, $table, $type, \%attr);

$dbh->func ("list_tables");

Every time where an C<\%attr> argument can be specified, this C<\%attr>
object's C<sql_table_source> attribute is preferred over the C<$dbh>
attribute or the driver default.

=head2 DBD::File::DataSource::Stream

package DBD::File::DataSource::Stream;

@DBD::File::DataSource::Stream::ISA = 'DBI::DBD::SqlEngine::DataSource';

sub complete_table_name
  {
      my ($self, $meta, $file, $respect_case) = @_;
      ...
      }

Clears all meta attributes identifying a file: C<f_fqfn>, C<f_fqbn> and
C<f_fqln>. The table name is set according to C<$respect_case> and
C<< $meta->{sql_identifier_case} >> (SQL_IC_LOWER, SQL_IC_UPPER).

package DBD::File::DataSource::Stream;

sub apply_encoding
  {
      my ($self, $meta, $fn) = @_;
      ...
      }

Applies the encoding from I<meta information> (C<< $meta->{f_encoding} >>)
to the file handled opened in C<open_data>.

package DBD::File::DataSource::Stream;

sub open_data
  {
      my ($self, $meta, $attrs, $flags) = @_;
      ...
      }

Opens (C<dup (2)>) the file handle provided in C<< $meta->{f_file} >>.

package DBD::File::DataSource::Stream;

sub can_flock { ... }

Returns whether C<flock (2)> is available or not (avoids retesting in
subclasses).

=head2 DBD::File::DataSource::File

package DBD::File::DataSource::File;

sub complete_table_name ($$;$)
  {
      my ($self, $meta, $table, $respect_case) = @_;
      ...
      }

The method C<complete_table_name> tries to map a filename to the associated
table name.  It is called with a partially filled meta structure for the
resulting table containing at least the following attributes:
C<< f_ext >>, C<< f_dir >>, C<< f_lockfile >> and C<< sql_identifier_case >>.

If a file/table map can be found then this method sets the C<< f_fqfn
>>, C<< f_fqbn >>, C<< f_fqln >> and C<< table_name >> attributes in
the meta structure. If a map cannot be found the table name will be
undef.

package DBD::File::DataSource::File;

sub open_data ($)
  {
      my ($self, $meta, $attrs, $flags) = @_;
      ...
      }

Depending on the attributes set in the table's meta data, the
following steps are performed. Unless C<< f_dontopen >> is set to a
true value, C<< f_fqfn >> must contain the full qualified file name
for the table to work on (file2table ensures this). The encoding in
C<< f_encoding >> is applied if set and the file is opened. If
C<<f_fqln >> (full qualified lock name) is set, this file is opened,
too. Depending on the value in C<< f_lock >>, the appropriate lock is
set on the opened data file or lock file.

=head2 DBD::File::Statement

Derives from DBI::SQL::Nano::Statement to provide following method:

=over 4

=item open_table

Implements the open_table method required by L<SQL::Statement> and
L<DBI::SQL::Nano>. All the work for opening the file(s) belonging to the
table is handled and parametrized in DBD::File::Table. Unless you intend
to add anything to the following implementation, an empty DBD::XXX::Statement
package satisfies DBD::File.

sub open_table ($$$$$)
  {
      my ($self, $data, $table, $createMode, $lockMode) = @_;

my $class = ref $self;
      $class =~ s/::Statement/::Table/;

my $flags = {
  	  createMode => $createMode,
	  lockMode   => $lockMode,
	  };
      $self->{command} eq "DROP" and $flags->{dropMode} = 1;

return $class->new ($data, { table => $table }, $flags);
      } # open_table

=back

=head2 DBD::File::Table

Derives from DBI::SQL::Nano::Table and provides physical file access for
the table data which are stored in the files.

=over 4

=item bootstrap_table_meta

Initializes a table meta structure. Can be safely overridden in a
derived class, as long as the C<< SUPER >> method is called at the end
of the overridden method.

It copies the following attributes from the database into the table meta data
C<< f_dir >>, C<< f_ext >>, C<< f_encoding >>, C<< f_lock >>, C<< f_schema >>
and C<< f_lockfile >> and makes them sticky to the table.

This method should be called before you attempt to map between file
name and table name to ensure the correct directory, extension etc. are
used.

=item init_table_meta

Initializes more attributes of the table meta data - usually more
expensive ones (e.g. those which require class instantiations) - when
the file name and the table name could mapped.

=item get_table_meta

Returns the table meta data. If there are none for the required
table, a new one is initialized. When it fails, nothing is
returned. On success, the name of the table and the meta data
structure is returned.

=item get_table_meta_attr

Returns a single attribute from the table meta data. If the attribute
name appears in C<%compat_map>, the attribute name is updated from
there.

=item set_table_meta_attr

Sets a single attribute in the table meta data. If the attribute
name appears in C<%compat_map>, the attribute name is updated from
there.

=item table_meta_attr_changed

Called when an attribute of the meta data is modified.

If the modified attribute requires to reset a calculated attribute, the
calculated attribute is reset (deleted from meta data structure) and
the I<initialized> flag is removed, too. The decision is made based on
C<%register_reset_on_modify>.

=item register_reset_on_modify

Allows C<set_table_meta_attr> to reset meta attributes when special
attributes are modified. For DBD::File, modifying one of C<f_file>, C<f_dir>,
C<f_ext> or C<f_lockfile> will reset C<f_fqfn>. DBD::DBM extends the
list for C<dbm_type> and C<dbm_mldbm> to reset the value of C<dbm_tietype>.

If your DBD has calculated values in the meta data area, then call
C<register_reset_on_modify>:

my %reset_on_modify = (xxx_foo => "xxx_bar");
  __PACKAGE__->register_reset_on_modify (\%reset_on_modify);

=item register_compat_map

Allows C<get_table_meta_attr> and C<set_table_meta_attr> to update the
attribute name to the current favored one:

# from DBD::DBM
  my %compat_map = (dbm_ext => "f_ext");
  __PACKAGE__->register_compat_map (\%compat_map);

=item open_file

Called to open the table's data file.

After this is done, a derived class might add more steps in an overridden
C<< open_file >> method.

=item new

Instantiates the table. This is done in 3 steps:

1. get the table meta data
 2. open the data file
 3. bless the table data structure using inherited constructor new

It is not recommended to override the constructor of the table class.
Find a reasonable place to add you extensions in one of the above four
methods.

=item drop

Implements the abstract table method for the C<< DROP >>
command. Discards table meta data after all files belonging to the
table are closed and unlinked.

Overriding this method might be reasonable in very rare cases.

=item seek

Implements the abstract table method used when accessing the table from the
engine. C<< seek >> is called every time the engine uses dumb algorithms
for iterating over the table content.

=item truncate

Implements the abstract table method used when dumb table algorithms
for C<< UPDATE >> or C<< DELETE >> need to truncate the table storage
after the last written row.

=back

You should consult the documentation of C<< SQL::Eval::Table >> (see
L<SQL::Eval>) to get more information about the abstract methods of the
table's base class you have to override and a description of the table
meta information expected by the SQL engines.

=head1 AUTHOR

The module DBD::File is currently maintained by

H.Merijn Brand < h.m.brand at xs4all.nl > and
Jens Rehsack  < rehsack at googlemail.com >

The original author is Jochen Wiedmann.

=head1 COPYRIGHT AND LICENSE

You may freely distribute and/or modify this module under the terms of
either the GNU General Public License (GPL) or the Artistic License, as
specified in the Perl README file.

=cut