implemented "naming" accessor, got rid of Loader/Compat/v0_040.pm

[dbsrgits/DBIx-Class-Schema-Loader.git] / lib / DBIx / Class / Schema / Loader.pm

package DBIx::Class::Schema::Loader;

use strict;
use warnings;
use base qw/DBIx::Class::Schema Class::Data::Accessor/;
use Carp::Clan qw/^DBIx::Class/;
use Class::C3;
use Scalar::Util qw/ weaken /;

# Always remember to do all digits for the version even if they're 0
# i.e. first release of 0.XX *must* be 0.XX000. This avoids fBSD ports
# brain damage and presumably various other packaging systems too
our $VERSION = '0.04999_12';

__PACKAGE__->mk_classaccessor('_loader_args' => {});
__PACKAGE__->mk_classaccessors(qw/
    dump_to_dir _loader_invoked _loader loader_class naming
/);

=head1 NAME

DBIx::Class::Schema::Loader - Dynamic definition of a DBIx::Class::Schema

=head1 SYNOPSIS

  ### use this module to generate a set of class files

  # in a script
  use DBIx::Class::Schema::Loader qw/ make_schema_at /;
  make_schema_at(
      'My::Schema',
      { debug => 1,
        dump_directory => './lib',
      },
      [ 'dbi:Pg:dbname="foo"', 'myuser', 'mypassword' ],
  );

  # from the command line or a shell script with dbicdump (distributed
  # with this module).  Do `perldoc dbicdump` for usage.
  dbicdump -o dump_directory=./lib \
           -o debug=1 \
           My::Schema \
           'dbi:Pg:dbname=foo' \
           myuser \
           mypassword

  ### or generate and load classes at runtime
  # note: this technique is not recommended
  # for use in production code

  package My::Schema;
  use base qw/DBIx::Class::Schema::Loader/;

  __PACKAGE__->loader_options(
      constraint              => '^foo.*',
      # debug                 => 1,
  );

  #### in application code elsewhere:

  use My::Schema;

  my $schema1 = My::Schema->connect( $dsn, $user, $password, $attrs);
  # -or-
  my $schema1 = "My::Schema"; $schema1->connection(as above);

=head1 DESCRIPTION 

DBIx::Class::Schema::Loader automates the definition of a
L<DBIx::Class::Schema> by scanning database table definitions and
setting up the columns, primary keys, and relationships.

DBIx::Class::Schema::Loader currently supports only the DBI storage type.
It has explicit support for L<DBD::Pg>, L<DBD::mysql>, L<DBD::DB2>,
L<DBD::SQLite>, and L<DBD::Oracle>.  Other DBI drivers may function to
a greater or lesser degree with this loader, depending on how much of the
DBI spec they implement, and how standard their implementation is.

Patches to make other DBDs work correctly welcome.

See L<DBIx::Class::Schema::Loader::DBI::Writing> for notes on writing
your own vendor-specific subclass for an unsupported DBD driver.

This module requires L<DBIx::Class> 0.07006 or later, and obsoletes
the older L<DBIx::Class::Loader>.

This module is designed more to get you up and running quickly against
an existing database, or to be effective for simple situations, rather
than to be what you use in the long term for a complex database/project.

That being said, transitioning your code from a Schema generated by this
module to one that doesn't use this module should be straightforward and
painless, so don't shy away from it just for fears of the transition down
the road.

=head1 METHODS

=head2 loader_class

=over 4

=item Argument: $loader_class

=back

Set the loader class to be instantiated when L</connection> is called.
If the classname starts with "::", "DBIx::Class::Schema::Loader" is
prepended. Defaults to L<DBIx::Class::Schema/storage_type> (which must
start with "::" when using L<DBIx::Class::Schema::Loader>).

This is mostly useful for subclassing existing loaders or in conjunction
with L</dump_to_dir>.

=head2 loader_options

=over 4

=item Argument: \%loader_options

=back

Example in Synopsis above demonstrates a few common arguments.  For
detailed information on all of the arguments, most of which are
only useful in fairly complex scenarios, see the
L<DBIx::Class::Schema::Loader::Base> documentation.

If you intend to use C<loader_options>, you must call
C<loader_options> before any connection is made, or embed the
C<loader_options> in the connection information itself as shown
below.  Setting C<loader_options> after the connection has
already been made is useless.

=cut

sub loader_options {
    my $self = shift;
    
    my %args = (ref $_[0] eq 'HASH') ? %{$_[0]} : @_;
    $self->_loader_args(\%args);

    $self;
}

sub _invoke_loader {
    my $self = shift;
    my $class = ref $self || $self;

    my $args = $self->_loader_args;

    # set up the schema/schema_class arguments
    $args->{schema} = $self;
    $args->{schema_class} = $class;
    weaken($args->{schema}) if ref $self;
    $args->{dump_directory} ||= $self->dump_to_dir;
    $args->{naming} = $self->naming;

    # XXX this only works for relative storage_type, like ::DBI ...
    my $impl = $self->loader_class
      || "DBIx::Class::Schema::Loader" . $self->storage_type;
    $impl = "DBIx::Class::Schema::Loader${impl}" if $impl =~ /^::/;
    eval { $self->ensure_class_loaded($impl) };
    croak qq/Could not load storage_type loader "$impl": "$@"/ if $@;

    $self->_loader($impl->new(%$args));
    $self->_loader->load;
    $self->_loader_invoked(1);

    $self;
}

=head2 connection

=over 4

=item Arguments: @args

=item Return Value: $new_schema

=back

See L<DBIx::Class::Schema/connection> for basic usage.

If the final argument is a hashref, and it contains the keys C<loader_options>
or C<loader_class>, those keys will be deleted, and their values value will be
used for the loader options or class, respectively, just as if set via the
L</loader_options> or L</loader_class> methods above.

The actual auto-loading operation (the heart of this module) will be invoked
as soon as the connection information is defined.

=cut

sub connection {
    my $self = shift;

    if($_[-1] && ref $_[-1] eq 'HASH') {
        for my $option (qw/ loader_class loader_options result_base_class schema_base_class/) {
            if(my $value = delete $_[-1]->{$option}) {
                $self->$option($value);
            }
        }
        pop @_ if !keys %{$_[-1]};
    }

    $self = $self->next::method(@_);

    my $class = ref $self || $self;
    if(!$class->_loader_invoked) {
        $self->_invoke_loader
    }

    return $self;
}

=head2 clone

See L<DBIx::Class::Schema/clone>.

=cut

sub clone {
    my $self = shift;

    my $clone = $self->next::method(@_);

    if($clone->_loader_args) {
        $clone->_loader_args->{schema} = $clone;
        weaken($clone->_loader_args->{schema});
    }

    $clone;
}

=head2 dump_to_dir

=over 4

=item Argument: $directory

=back

Calling this as a class method on either L<DBIx::Class::Schema::Loader>
or any derived schema class will cause all schemas to dump
manual versions of themselves to the named directory when they are
loaded.  In order to be effective, this must be set before defining a
connection on this schema class or any derived object (as the loading
happens as soon as both a connection and loader_options are set, and
only once per class).

See L<DBIx::Class::Schema::Loader::Base/dump_directory> for more
details on the dumping mechanism.

This can also be set at module import time via the import option
C<dump_to_dir:/foo/bar> to L<DBIx::Class::Schema::Loader>, where
C</foo/bar> is the target directory.

Examples:

    # My::Schema isa DBIx::Class::Schema::Loader, and has connection info
    #   hardcoded in the class itself:
    perl -MDBIx::Class::Schema::Loader=dump_to_dir:/foo/bar -MMy::Schema -e1

    # Same, but no hard-coded connection, so we must provide one:
    perl -MDBIx::Class::Schema::Loader=dump_to_dir:/foo/bar -MMy::Schema -e 'My::Schema->connection("dbi:Pg:dbname=foo", ...)'

    # Or as a class method, as long as you get it done *before* defining a
    #  connection on this schema class or any derived object:
    use My::Schema;
    My::Schema->dump_to_dir('/foo/bar');
    My::Schema->connection(........);

    # Or as a class method on the DBIx::Class::Schema::Loader itself, which affects all
    #   derived schemas
    use My::Schema;
    use My::OtherSchema;
    DBIx::Class::Schema::Loader->dump_to_dir('/foo/bar');
    My::Schema->connection(.......);
    My::OtherSchema->connection(.......);

    # Another alternative to the above:
    use DBIx::Class::Schema::Loader qw| dump_to_dir:/foo/bar |;
    use My::Schema;
    use My::OtherSchema;
    My::Schema->connection(.......);
    My::OtherSchema->connection(.......);

=cut

sub import {
    my $self = shift;

    return if !@_;

    my $cpkg = (caller)[0];

    foreach my $opt (@_) {
        if($opt =~ m{^dump_to_dir:(.*)$}) {
            $self->dump_to_dir($1)
        }
        elsif($opt eq 'make_schema_at') {
            no strict 'refs';
            *{"${cpkg}::make_schema_at"} = \&make_schema_at;
        }
        elsif($opt eq 'naming') {
            no strict 'refs';
            *{"${cpkg}::naming"} = sub { $self->naming(@_) };
        }
    }
}

=head2 make_schema_at

=over 4

=item Arguments: $schema_class_name, \%loader_options, \@connect_info

=item Return Value: $schema_class_name

=back

This function creates a DBIx::Class schema from an existing RDBMS
schema.  With the C<dump_directory> option, generates a set of
DBIx::Class classes from an existing database schema read from the
given dsn.  Without a C<dump_directory>, creates schema classes in
memory at runtime without generating on-disk class files.

For a complete list of supported loader_options, see
L<DBIx::Class::Schema::Loader::Base>

This function can be imported in the usual way, as illustrated in
these Examples:

    # Simple example, creates as a new class 'New::Schema::Name' in
    #  memory in the running perl interpreter.
    use DBIx::Class::Schema::Loader qw/ make_schema_at /;
    make_schema_at(
        'New::Schema::Name',
        { debug => 1 },
        [ 'dbi:Pg:dbname="foo"','postgres' ],
    );

    # Inside a script, specifying a dump directory in which to write
    # class files
    use DBIx::Class::Schema::Loader qw/ make_schema_at /;
    make_schema_at(
        'New::Schema::Name',
        { debug => 1, dump_directory => './lib' },
        [ 'dbi:Pg:dbname="foo"','postgres' ],
    );

=cut

sub make_schema_at {
    my ($target, $opts, $connect_info) = @_;

    {
        no strict 'refs';
        @{$target . '::ISA'} = qw/DBIx::Class::Schema::Loader/;
    }

    $target->loader_options($opts);
    $target->connection(@$connect_info);
}

=head2 rescan

=over 4

=item Return Value: @new_monikers

=back

Re-scans the database for newly added tables since the initial
load, and adds them to the schema at runtime, including relationships,
etc.  Does not process drops or changes.

Returns a list of the new monikers added.

=cut

sub rescan { my $self = shift; $self->_loader->rescan($self) }

=head2 naming

=over 4

=item Arguments: \%opts | $ver

=back

Controls the naming options for backward compatibility, see
L<DBIx::Class::Schema::Loader::Base/naming> for details.

To upgrade a dynamic schema, use:

    __PACKAGE__->naming('current');

Can be imported into your dump script and called as a function as well:

    naming('v4');

=head1 KNOWN ISSUES

=head2 Multiple Database Schemas

Currently the loader is limited to working within a single schema
(using the underlying RDBMS's definition of "schema").  If you have a
multi-schema database with inter-schema relationships (which is easy
to do in PostgreSQL or DB2 for instance), you currently can only
automatically load the tables of one schema, and relationships to
tables in other schemas will be silently ignored.

At some point in the future, an intelligent way around this might be
devised, probably by allowing the C<db_schema> option to be an
arrayref of schemas to load.

In "normal" L<DBIx::Class::Schema> usage, manually-defined
source classes and relationships have no problems crossing vendor schemas.

=head1 ACKNOWLEDGEMENTS

Matt S Trout, all of the #dbix-class folks, and everyone who's ever sent
in a bug report or suggestion.

Based on L<DBIx::Class::Loader> by Sebastian Riedel

Based upon the work of IKEBE Tomohiro

=head1 AUTHOR

blblack: Brandon Black <blblack@gmail.com>

=head1 CONTRIBUTORS

ilmarii: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>

arcanez: Justin Hunter <justin.d.hunter@gmail.com>

ash: Ash Berlin <ash@cpan.org>

Caelum: Rafael Kitover <rkitover@cpan.org>

TSUNODA Kazuya <drk@drk7.jp>

Robert Bohne <rbo@openserv.org>

ribasushi: Peter Rabbitson <rabbit+dbic@rabbit.us>

gugu: Andrey Kostenko <a.kostenko@rambler-co.ru>

... and lots of other folks. If we forgot you, please write the current
maintainer or RT.

=head1 LICENSE

This library is free software; you can redistribute it and/or modify it under
the same terms as Perl itself.

=head1 SEE ALSO

L<DBIx::Class>, L<DBIx::Class::Manual::ExampleSchema>

=cut

1;