add recipes for schema import/export + versioning

[dbsrgits/DBIx-Class.git] / lib / DBIx / Class / Manual / Cookbook.pod

=head1 NAME 

DBIx::Class::Manual::Cookbook - Miscellaneous recipes

=head1 RECIPES

=head2 Searching

=head3 Paged results

When you expect a large number of results, you can ask L<DBIx::Class> for a
paged resultset, which will fetch only a small number of records at a time:

  my $rs = $schema->resultset('Artist')->search(
    {},
    {
      page => 1,  # page to return (defaults to 1)
      rows => 10, # number of results per page
    },
  );

  return $rs->all(); # all records for page 1

The C<page> attribute does not have to be specified in your search:

  my $rs = $schema->resultset('Artist')->search(
    {},
    {
      rows => 10,
    }
  );

  return $rs->page(1); # DBIx::Class::ResultSet containing first 10 records

In either of the above cases, you can return a L<Data::Page> object for the
resultset (suitable for use in e.g. a template) using the C<pager> method:

  return $rs->pager();

=head3 Complex WHERE clauses

Sometimes you need to formulate a query using specific operators:

  my @albums = $schema->resultset('Album')->search({
    artist => { 'like', '%Lamb%' },
    title  => { 'like', '%Fear of Fours%' },
  });

This results in something like the following C<WHERE> clause:

  WHERE artist LIKE '%Lamb%' AND title LIKE '%Fear of Fours%'

Other queries might require slightly more complex logic:

  my @albums = $schema->resultset('Album')->search({
    -or => [
      -and => [
        artist => { 'like', '%Smashing Pumpkins%' },
        title  => 'Siamese Dream',
      ],
      artist => 'Starchildren',
    ],
  });

This results in the following C<WHERE> clause:

  WHERE ( artist LIKE '%Smashing Pumpkins%' AND title = 'Siamese Dream' )
    OR artist = 'Starchildren'

For more information on generating complex queries, see
L<SQL::Abstract/WHERE CLAUSES>.

=head3 Using specific columns

When you only want selected columns from a table, you can use C<cols> to
specify which ones you need:

  my $rs = $schema->resultset('Artist')->search(
    {},
    {
      cols => [qw/ name /]
    }
  );

  # Equivalent SQL:
  # SELECT artist.name FROM artist

=head3 Using database functions or stored procedures

The combination of C<select> and C<as> can be used to return the result of a
database function or stored procedure as a column value. You use C<select> to
specify the source for your column value (e.g. a column name, function, or
stored procedure name). You then use C<as> to set the column name you will use
to access the returned value:

  my $rs = $schema->resultset('Artist')->search(
    {},
    {
      select => [ 'name', { LENGTH => 'name' } ],
      as     => [qw/ name name_length /],
    }
  );

  # Equivalent SQL:
  # SELECT name name, LENGTH( name ) name_length
  # FROM artist

If your alias exists as a column in your base class (i.e. it was added with
C<add_columns>), you just access it as normal. Our C<Artist> class has a C<name>
column, so we just use the C<name> accessor:

  my $artist = $rs->first();
  my $name = $artist->name();

If on the other hand the alias does not correspond to an existing column, you
can get the value using the C<get_column> accessor:

  my $name_length = $artist->get_column('name_length');

If you don't like using C<get_column>, you can always create an accessor for
any of your aliases using either of these:

  # Define accessor manually:
  sub name_length { shift->get_column('name_length'); }
    
  # Or use DBIx::Class::AccessorGroup:
  __PACKAGE__->mk_group_accessors('column' => 'name_length');

=head3 SELECT DISTINCT with multiple columns

  my $rs = $schema->resultset('Foo')->search(
    {},
    {
      select => [
        { distinct => [ $source->columns ] }
      ],
      as => [ $source->columns ]
    }
  );

=head3 SELECT COUNT(DISTINCT colname)

  my $rs = $schema->resultset('Foo')->search(
    {},
    {
      select => [
        { count => { distinct => 'colname' } }
      ],
      as => [ 'count' ]
    }
  );

=head3 Grouping results

L<DBIx::Class> supports C<GROUP BY> as follows:

  my $rs = $schema->resultset('Artist')->search(
    {},
    {
      join     => [qw/ cds /],
      select   => [ 'name', { count => 'cds.cdid' } ],
      as       => [qw/ name cd_count /],
      group_by => [qw/ name /]
    }
  );

  # Equivalent SQL:
  # SELECT name, COUNT( cds.cdid ) FROM artist me
  # LEFT JOIN cd cds ON ( cds.artist = me.artistid )
  # GROUP BY name

=head2 Using joins and prefetch

You can use the C<join> attribute to allow searching on, or sorting your
results by, one or more columns in a related table. To return all CDs matching
a particular artist name:

  my $rs = $schema->resultset('CD')->search(
    {
      'artist.name' => 'Bob Marley'    
    },
    {
      join => [qw/artist/], # join the artist table
    }
  );

  # Equivalent SQL:
  # SELECT cd.* FROM cd
  # JOIN artist ON cd.artist = artist.id
  # WHERE artist.name = 'Bob Marley'

If required, you can now sort on any column in the related tables by including
it in your C<order_by> attribute:

  my $rs = $schema->resultset('CD')->search(
    {
      'artist.name' => 'Bob Marley'
    },
    {
      join     => [qw/ artist /],
      order_by => [qw/ artist.name /]
    }
  };

  # Equivalent SQL:
  # SELECT cd.* FROM cd
  # JOIN artist ON cd.artist = artist.id
  # WHERE artist.name = 'Bob Marley'
  # ORDER BY artist.name

Note that the C<join> attribute should only be used when you need to search or
sort using columns in a related table. Joining related tables when you only
need columns from the main table will make performance worse!

Now let's say you want to display a list of CDs, each with the name of the
artist. The following will work fine:

  while (my $cd = $rs->next) {
    print "CD: " . $cd->title . ", Artist: " . $cd->artist->name;
  }

There is a problem however. We have searched both the C<cd> and C<artist> tables
in our main query, but we have only returned data from the C<cd> table. To get
the artist name for any of the CD objects returned, L<DBIx::Class> will go back
to the database:

  SELECT artist.* FROM artist WHERE artist.id = ?

A statement like the one above will run for each and every CD returned by our
main query. Five CDs, five extra queries. A hundred CDs, one hundred extra
queries!

Thankfully, L<DBIx::Class> has a C<prefetch> attribute to solve this problem.
This allows you to fetch results from a related table as well as the main table
for your class:

  my $rs = $schema->resultset('CD')->search(
    {
      'artist.name' => 'Bob Marley'
    },
    {
      join     => [qw/ artist /],
      order_by => [qw/ artist.name /],
      prefetch => [qw/ artist /] # return artist data too!
    }
  );

  # Equivalent SQL (note SELECT from both "cd" and "artist"):
  # SELECT cd.*, artist.* FROM cd
  # JOIN artist ON cd.artist = artist.id
  # WHERE artist.name = 'Bob Marley'
  # ORDER BY artist.name

The code to print the CD list remains the same:

  while (my $cd = $rs->next) {
    print "CD: " . $cd->title . ", Artist: " . $cd->artist->name;
  }

L<DBIx::Class> has now prefetched all matching data from the C<artist> table,
so no additional SQL statements are executed. You now have a much more
efficient query.

Note that as of L<DBIx::Class> 0.04, C<prefetch> cannot be used with
C<has_many> relationships. You will get an error along the lines of "No
accessor for prefetched ..." if you try.

Also note that C<prefetch> should only be used when you know you will
definitely use data from a related table. Pre-fetching related tables when you
only need columns from the main table will make performance worse!

=head3 Multi-step joins

Sometimes you want to join more than one relationship deep. In this example,
we want to find all C<Artist> objects who have C<CD>s whose C<LinerNotes>
contain a specific string:

  # Relationships defined elsewhere:
  # Artist->has_many('cds' => 'CD', 'artist');
  # CD->has_one('liner_notes' => 'LinerNotes', 'cd');

  my $rs = $schema->resultset('Artist')->search(
    {
      'liner_notes.notes' => { 'like', '%some text%' },
    },
    {
      join => {
        'cds' => 'liner_notes'
      }
    }
  );

  # Equivalent SQL:
  # SELECT artist.* FROM artist
  # JOIN ( cd ON artist.id = cd.artist )
  # JOIN ( liner_notes ON cd.id = liner_notes.cd )
  # WHERE liner_notes.notes LIKE '%some text%'

Joins can be nested to an arbitrary level. So if we decide later that we
want to reduce the number of Artists returned based on who wrote the liner
notes:

  # Relationship defined elsewhere:
  # LinerNotes->belongs_to('author' => 'Person');

  my $rs = $schema->resultset('Artist')->search(
    {
      'liner_notes.notes' => { 'like', '%some text%' },
      'author.name' => 'A. Writer'
    },
    {
      join => {
        'cds' => {
          'liner_notes' => 'author'
        }
      }
    }
  );

  # Equivalent SQL:
  # SELECT artist.* FROM artist
  # JOIN ( cd ON artist.id = cd.artist )
  # JOIN ( liner_notes ON cd.id = liner_notes.cd )
  # JOIN ( author ON author.id = liner_notes.author )
  # WHERE liner_notes.notes LIKE '%some text%'
  # AND author.name = 'A. Writer'

=head2 Transactions

As of version 0.04001, there is improved transaction support in
L<DBIx::Class::Storage::DBI>.  Here is an example of the recommended
way to use it:

  my $genus = Genus->find(12);
  eval {
    MyDB->txn_begin;
    $genus->add_to_species({ name => 'troglodyte' });
    $genus->wings(2);
    $genus->update;
    cromulate($genus); # Can have a nested transation
    MyDB->txn_commit;
  };
  if ($@) {
    # Rollback might fail, too
    eval {
      MyDB->txn_rollback
    };
  }

Currently, a nested commit will do nothing and a nested rollback will
die.  The code at each level must be sure to call rollback in the case
of an error, to ensure that the rollback will propagate to the top
level and be issued.  Support for savepoints and for true nested
transactions (for databases that support them) will hopefully be added
in the future.

=head2 Many-to-many relationships

This is straightforward using L<DBIx::Class::Relationship::ManyToMany>:

  package My::DB;
  # ... set up connection ...

  package My::User;
  use base 'My::DB';
  __PACKAGE__->table('user');
  __PACKAGE__->add_columns(qw/id name/);
  __PACKAGE__->set_primary_key('id');
  __PACKAGE__->has_many('user_address' => 'My::UserAddress', 'user');
  __PACKAGE__->many_to_many('addresses' => 'user_address', 'address');

  package My::UserAddress;
  use base 'My::DB';
  __PACKAGE__->table('user_address');
  __PACKAGE__->add_columns(qw/user address/);
  __PACKAGE__->set_primary_key(qw/user address/);
  __PACKAGE__->belongs_to('user' => 'My::User');
  __PACKAGE__->belongs_to('address' => 'My::Address');

  package My::Address;
  use base 'My::DB';
  __PACKAGE__->table('address');
  __PACKAGE__->add_columns(qw/id street town area_code country/);
  __PACKAGE__->set_primary_key('id');
  __PACKAGE__->has_many('user_address' => 'My::UserAddress', 'address');
  __PACKAGE__->many_to_many('users' => 'user_address', 'user');

  $rs = $user->addresses(); # get all addresses for a user
  $rs = $address->users(); # get all users for an address

=head2 Setting default values for a row

It's as simple as overriding the C<new> method.  Note the use of
C<next::method>.

  sub new {
    my ( $class, $attrs ) = @_;

    $attrs->{foo} = 'bar' unless defined $attrs->{foo};

    $class->next::method($attrs);
  }

=head2 Stringification

Employ the standard stringification technique by using the C<overload>
module.  Replace C<foo> with the column/method of your choice.

  use overload '""' => 'foo', fallback => 1;

=head2 Disconnecting cleanly

If you find yourself quitting an app with Control-C a lot during
development, you might like to put the following signal handler in
your main database class to make sure it disconnects cleanly:

  $SIG{INT} = sub {
    __PACKAGE__->storage->dbh->disconnect;
  };

=head2 Schema import/export

This functionality requires you to have L<SQL::Translator> (also known as
"SQL Fairy") installed.

To create a DBIx::Class schema from an existing database:

 sqlt --from DBI
      --to DBIx::Class::File
      --prefix "MySchema" > MySchema.pm

To create a MySQL database from an existing L<DBIx::Class> schema, convert the
schema to MySQL's dialect of SQL:

  sqlt --from DBIx::Class --to MySQL --DBIx::Class "MySchema.pm" > Schema1.sql
  
And import using the mysql client:

  mysql -h "host" -D "database" -u "user" -p < Schema1.sql

=head2 Schema versioning

The following example shows simplistically how you might use DBIx::Class to
deploy versioned schemas to your customers. The basic process is as follows:

1) Create a DBIx::Class schema
2) Save the schema
3) Deploy to customers
4) Modify schema to change functionality
5) Deploy update to customers

=head3 Create a DBIx::Class schema

This can either be done manually, or generated from an existing database as
described under C<Schema import/export>.

=head3 Save the schema

Use C<sqlt> to transform your schema into an SQL script suitable for your
customer's database. E.g. for MySQL:

  sqlt --from DBIx::Class
       --to MySQL
       --DBIx::Class "MySchema.pm" > Schema1.mysql.sql

If you need to target databases from multiple vendors, just generate an SQL
script suitable for each. To support PostgreSQL too:

  sqlt --from DBIx::Class
       --to PostgreSQL
       --DBIx::Class "MySchema.pm" > Schema1.pgsql.sql

=head3 Deploy to customers

There are several ways you could deploy your schema. These are probably
beyond the scope of this recipe, but might include:

1) Require customer to apply manually using their RDBMS.
2) Package along with your app, making database dump/schema update/tests
all part of your install.

=head3 Modify the schema to change functionality

As your application evolves, it may be necessary to modify your schema to
change functionality. Once the changes are made to your schema in DBIx::Class,
export the modified schema as before, taking care not to overwrite the original:

  sqlt --from DBIx::Class
       --to MySQL
       --DBIx::Class "Anything.pm" > Schema2.mysql.sql

Next, use sqlt-diff to create an SQL script that will update the customer's
database schema:

  sqlt-diff --to MySQL Schema1=MySQL Schema2=MySQL > SchemaUpdate.mysql.sql

=head3 Deploy update to customers

The schema update can be deployed to customers using the same method as before.

=cut