The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.

NAME

DBIx::MoCo - Light & Fast Model Component

SYNOPSIS

# First, set up your db.
package Blog::DataBase;
use base qw(DBIx::MoCo::DataBase);

__PACKAGE__->dsn('dbi:mysql:dbname=blog');
__PACKAGE__->username('test');
__PACKAGE__->password('test');

1;

# Second, create a base class for all models.
package Blog::MoCo;
use base qw 'DBIx::MoCo'; # Inherit DBIx::MoCo
use Blog::DataBase;

__PACKAGE__->db_object('Blog::DataBase');

# If you want to use caching feature, you must explicitly set a
# cache object via cache_object() method.

use Cache::Memcached;
my $cache = Cache::Memcached->new;
$cache->set_servers([ ... ])
__PACKAGE__->cache_object($cache); # Enables caching by memcached

1;

# Third, create your models.
package Blog::User;
use base qw 'Blog::MoCo';

__PACKAGE__->table('user');
__PACKAGE__->has_many(
    entries => 'Blog::Entry',
    { key => 'user_id' }
);
__PACKAGE__->has_many(
    bookmarks => 'Blog::Bookmark',
    { key => 'user_id' }
);

1;

package Blog::Entry;
use base qw 'Blog::MoCo';

__PACKAGE__->table('entry');
__PACKAGE__->has_a(
    user => 'Blog::User',
    { key => 'user_id' }
);
__PACKAGE__->has_many(
    bookmarks => 'Blog::Bookmark',
    { key => 'entry_id' }
);

1;

package Blog::Bookmark;
use base qw 'Blog::MoCo';

__PACKAGE__->table('bookmark');
__PACKAGE__->has_a(
    user => 'Blog::User',
    { key => 'user_id' }
);
__PACKAGE__->has_a(
    entry => 'Blog::Entry',
    { key => 'entry_id' }
);

1;

# Now, You can use some methods same as Class::DBI.
# And, all objects are stored in cache automatically.
my $user = Blog::User->retrieve(user_id => 123);
print $user->name;
$user->name('jkontan'); # update db immediately
print $user->name; # jkontan

my $user2 = Blog::User->retrieve(user_id => 123);
# $user is same as $user2

# You can easily get has_many objects array.
my $entries = $user->entries;
my $entries2 = $user->entries;
# $entries is same reference as $entries2
my $entry = $entries->first; # isa Blog::Entry
print $entry->title; # you can use methods in Entry class.

Blog::Entry->create(
  user_id => 123,
  title => 'new entry!',
);
# $user->entries will be flushed automatically.
my $entries3 = $user->entries;
# $entries3 isnt $entries

print ($entries->last eq $entries2->last); # 1
print ($entries->last eq $entries3->last); # 1
# same instance

# You can delay update/create query to database using session.
DBIx::MoCo->start_session;
$user->name('jkondo'); # not saved now. changed in cache.
print $user->name; # 'jkondo'
$user->save; # update db
print Blog::User->retrieve(123)->name; # 'jkondo'

# Or, update queries will be thrown automatically after ending session.
$user->name('jkontan');
DBIx::MoCo->end_session;
print Blog::User->retrieve(123)->name; # 'jkontan'

DESCRIPTION

Light & Fast Model Component

CLASS DEFINITION METHODS

Here are common methods related with class definitions.

add_trigger

Adds triggers. Here are the types which called from DBIx::MoCo.

before_create
after_create
before_update
after_update
before_delete

You can add your trigger like this.

package Blog::User;
__PACKAGE__->add_trigger(before_create => sub
  my ($class, $args) = @_;
  $args->{name} .= '-san';
});

# in your scripts
my $u = Blog::User->create(name => 'ishizaki');
is ($u->name, 'ishizaki-san'); # ok.

before_create passes a hash reference of new object data as the second argument, and all other triggers pass the instance $self.

has_a

Defines has_a relationship between 2 models.

has_many

Defines has_many relationship between 2 models. You can define additional conditions as below.

Blog::User->has_many(
  root_messages => 'Blog::Message', {
    key => {name => 'to_name'},
    condition => 'reference_id is null',
    order => 'modified desc',
  },
);

condition is additional sql statement will be used in where statement. order is used for specifying order statement.

In above case, SQL statement will be

SELECT message_id FROM message
WHERE to_name = 'myname' AND reference_id is null
ORDER BY modified desc

And, all each results will be inflated as Blog::Message by retrieving all records again (with using cache).

retrieve_keys

Defines keys for retrieving by retrieve_all etc.

If there aren't any unique keys in your table, please specify these keys.

package Blog::Bookmark;

__PACKAGE__->retrieve_keys(['user_id', 'entry_id']);
# When user can add multiple bookmarks onto same entry.
primary_keys

Returns primary keys. Usually it returns them automatically by retrieving schema data from database.

But you can also redefine this parameter by overriding this method. It's useful when MoCo cannot get schema data from your dsn.

sub primary_keys {['user_id']}
unique_keys

Returns unique keys including primary keys. You can override this as same as primary_keys.

sub unique_keys {['user_id','name']}
schema

Returns DBIx::MoCo::Schema object reference related with your model class. You can set/get any parameters using Schema's param method. See DBIx::MoCo::Schema for details.

columns

Returns array reference of column names.

has_column(col_name)

Returns which the table has the column or not.

utf8_columns

Receives array reference and defines utf8 columns.

When you call utf8 column method, you'll get string with utf8 flag on. But you can get raw string when you call param('colname') method.

__PACKAGE__->utf8_columns([qw(title body)]);

my $e = Blog::Entry->retrieve(1);
print Encode::is_utf8($e->title); # true
print Encode::is_utf8($e->param('title')); # false
print Encode::is_utf8($e->uri); # false
list_class

By default, retrieve_all(), search(), etc. return results as a DBIx::MoCo::List object when in scalar context. If you want to add some features into the list class, you can make a subclass of DBIx::MoCo::List and tell your model class to use your own class instead by specifying the class via list_class() method.

# In Blog::Entry
__PACKAGE__->list_class('Blog::Entry::List');

# In Blog::Entry::List
package Blog::Entry::List;
use base qw/DBIx::MoCo::List/;

sub to_rss {
    processing rss from entries ...
}

1;

# The return value now has to_rss() method.
my $entries = Blog::Entry->search( ... ); # is a Blog::Entry::List
$entries->to_rss;

CACHING FEATURE

Setup

If you want to use caching feature provided by DBIx::MoCo, you must explicitly set the object via cache_object() method explained below, which sets an object to be used when caching data from database. The object can be, for example, a Cache::* modules such as Cache::Memory, Cache::Memecached, etc.

# In your Moco.pm
package Blog::MoCo;
use base qw 'DBIx::MoCo';

...

use Cache::Memcached;
my $cache = Cache::Memcached->new;
$cache->set_servers([ ... ])

__PACKAGE__->cache_object($cache); # Enables caching by memcached

Cache Algorithm

MoCo caches objects effectively.

There are 3 functions to control MoCo's cache. Their functions are called appropriately when some operations are called to a particular object.

Here are the 3 functions.

store_self_cache

Stores self instance for all own possible object ids.

flush_self_cache

Flushes all caches for all own possible object ids.

flush_belongs_to

Flushes all caches whose have has_many arrays including the object.

And, here are the triggers which call their functions.

_after_create

Calls store_self_cache and flush_belongs_to.

_before_update

Calls flush_self_cache.

_after_update

Calls store_self_cache.

_before_delete

Calls flush_self_cache and flush_belongs_to.

SESSION & CACHE METHODS

Here are common methods related with session.

start_session

Starts session.

end_session

Ends session.

is_in_session

Returns DBIx::MoCo is in session or not.

cache_object

Sets an object to be used when caching data from database. For example, the object can be a Cache::* modules such as Cache::Memory, Cache::Memecached, etc.

cache_status

Returns cache status of the current session as a hash reference. cache_status provides retrieve_count, retrieve_cache_count, retrieved_oids retrieve_all_count, has_many_count, has_many_cache_count,

flush

Delete attribute from given attr. name.

save

Saves changed columns in the current session.

icache_expiration

Specifies instance cache expiration time in seconds. MoCo store has_a, has_many instances in instance variable if this value is set.

__PACKAGE__->icache_expiration(30);

It's not necessary to setup icache if you are runnnig MoCo with DBIx::MoCo::Cache object because it is more powerful and as fast as icache.

You'd better to consider this option when you are running MoCo with centralized cache mechanism such as memcached.

cache_null_object

Specifies which MoCo will store null object when retrieve will fail.

object_id_prefix

This prefix is used for generating object ids and the ids are used as cache keys. Default value of this prefix is the name of class.

This option is effective when you use some classes which have parent -child relationships and they represent same table.

package Blog::Entry;

sub object_id_prefix { 'Blog::Entry' }

1;

package Blog::Entry::Video;
use base qw(Blog::Entry);

1;

MUID value is used for object_id when the class has muid field even if this prefix is specified.

DATA OPERATIONAL METHODS

Here are common methods related with operating data.

retrieve

Retrieves an object and returns that using cache (if possible).

my $u1 = Blog::User->retrieve(123); # retrieve by primary_key
my $u2 = Blog::User->retrieve(user_id => 123); # same as above
my $u3 = Blog::User->retrieve(name => 'jkondo'); # retrieve by name
restore_from_db

Restores self attributes from db.

retrieve_by_db

Retrieves an object from db.

retrieve_multi

Returns results of given array of conditions.

my $users = Blog::User->retrieve_multi(
  {user_id => 123},
  {user_id => 234},
);
retrieve_all

Returns results of given conditions as DBIx::MoCo::List instance.

my $users = Blog::User->retrieve_all(birthday => '2001-07-15');
retrieve_or_create

Retrieves a object or creates new record with given data and returns that.

my $user = Blog::User->retrieve_or_create(name => 'jkondo');
create

Creates new object and returns that.

my $user = Blog::User->create(
  name => 'jkondo',
  birthday => '2001-07-15',
);
delete

Deletes a object. You can call delete as both of class and instance method.

$user->delete;
Blog::User->delete($user);
delete_all

Deletes all records with given conditions. You should specify the conditions as a hash reference.

Blog::User->delete_all(where => {birthday => '2001-07-15'});

Returns results of given conditions as DBIx::MoCo::List instance. You can specify search conditions in 3 diferrent ways. "Hash reference style", "Array reference style" and "Scalar style".

Hash reference style is same as SQL::Abstract style and like this.

Blog::User->search(where => {name => 'jkondo'});

Array style is the most flexible. You can use placeholder.

Blog::User->search(
  where => ['name = ?', 'jkondo'],
);
Blog::User->search(
  where => ['name in (?,?)', 'jkondo', 'cinnamon'],
);
Blog::Entry->search(
  where => ['name = :name and date like :date'],
           name => 'jkondo', date => '2007-04%'],
);

Scalar style is the simplest one, and most flexible in other word.

Blog::Entry->search(
  where => "name = 'jkondo' and DATE_ADD(date, INTERVAL 1 DAY) > NOW()',
);

You can also specify field, order, offset, limit, group, with too. Full spec search statement will be like the following.

Blog::Entry->search(
  field => 'entry_id',
  where => ['name = ?', 'jkondo'],
  order => 'created desc',
  offset => 0,
  limit => 1,
  group => 'title',
  with  => [qw(user)], # for prefetching users related to each entry
);

Search results will not be cached because MoCo expects that the conditions for search will be complicated and should not be cached. You should use retrieve or retrieve_all method instead of search if you'll use simple conditions.

See Prefetching section below for details of with option in search() method.

count

Returns the count of results matched with given conditions. You can specify the conditions in same way as search's where spec.

Blog::User->count({name => 'jkondo'}); # Hash reference style
Blog::User->count(['name => ?', 'jkondo']); # Array reference style
Blog::User->count("name => 'jkondo'"); # Scalar style
find

Similar to search, but returns only the first item as a reference (not as an array).

retrieve_by_column(_and_column2)

Auto generated method which returns an object by using key defined is method and given value.

my $user = Blog::User->retrieve_by_name('jkondo');
retrieve_by_column(_and_column2)_or_create

Similar to retrieve_or_create.

my $user = Blog::User->retrieve_by_name_or_create('jkondo');
retrieve_by_column_or_column2

Returns an object matched with given column names.

my $user = Blog::User->retrieve_by_user_id_or_name('jkondo');
param

Set or get attribute from given attr. name.

set

Set attribute which is not related with DB schema or set temporary.

column_as_something

Inflate column value by using DBIx::MoCo::Column::* plugins. If you set up your plugin like this,

package DBIx::MoCo::Column::URI;

sub URI {
  my $self = shift;
  return URI->new($$self);
}

sub URI_as_string {
  my $class = shift;
  my $uri = shift or return;
  return $uri->as_string;
}

1;

Then, you can use column_as_URI method as following,

my $e = MyEntry->retrieve(..);
print $e->uri; # 'http://test.com/test'
print $e->uri_as_URI->host; # 'test.com';

my $uri = URI->new('http://www.test.com/test');
$e->uri_as_URI($uri); # set uri by using URI instance

The name of infrate method which will be imported must be same as the package name.

If you don't define "as string" method (such as URI_as_string), scalar evaluated value of given argument will be used for new value instead.

has_a, has_many auto generated methods

If you define has_a, has_many relationships,

package Blog::Entry;
use base qw 'Blog::MoCo';

__PACKAGE__->table('entry');
__PACKAGE__->has_a(
    user => 'Blog::User',
    { key => 'user_id' }
);
__PACKAGE__->has_many(
    bookmarks => 'Blog::Bookmark',
    { key => 'entry_id' }
);

You can use those keys as methods.

my $e = Blog::Entry->retrieve(..);
print $e->user; # isa Blog::User
print $e->bookmarks; # isa ARRAY of Blog::Bookmark
quote

Quotes given string using DBI's quote method.

HINTS FOR PERFORMANCE

Prefetching

By default, DBIx::MoCo can issue too many queries in such case as below:

my $user = Blog::User->retrieve(name => $name);
for my $entry ( $user->entries ) {
    ## Entry has a user
    $entry->user->name;
}

The code above executes more than twice as many queries as the count $user-entries->size> method returns, which can cause problems on performance when the count is large. DBIx::MoCo provides prefetching feature to solve the problem.

You can specify the target to be prefetched by with option in model class definitions as below:

package Blog::User;
use base qw 'Blog::MoCo';

__PACKAGE__->table('user');
__PACKAGE__->has_many(
    entries => 'Blog::Entry',
    {
        key  => 'user_id',
        with => [qw(user)],  # Added
    }
);
1;

package Blog::Entry;
use base qw 'Blog::MoCo';

__PACKAGE__->table('entry');
__PACKAGE__->has_a(
    user => 'Blog::User',
    { key => 'user_id' }
);
1;

As a result, $user-entry> prefetches users of all entries, and you'll see the performance is drastically improved.

my $user = Blog::User->retrieve(name => $name);
for my $entry ( $user->entries ) {              # Does prefetching
    ## Entry has a user
    $entry->user->name;
}

In case that you temporally don't want the method to prefetch, you can inhibit prefetching as below:

$user->entries({ without => 'user' });

with option described above can appear not only in has_many() method, but also in search() method.

my $entries = Blog::Entry->search(
    ...
    with => [qw(user)],
);

for my $entry ( @$entries ) {
    $entry->user->name;       # $entry->user is already prefetched
}

FORM VALIDATION

You can validate user parameters using moco's schema. For example you can define your validation profile using param like this,

package Blog::User;

__PACKAGE__->schema->param([
  name => ['NOT_BLANK', 'ASCII', ['DBIC_UNIQUE', 'Blog::User', 'name']],
  mail => ['NOT_BLANK', 'EMAIL_LOOSE'],
]);

And then,

# In your scripts
sub validate {
  my $self = shift;
  my $q = $self->query;
  my $prof = Blog::User->schema->param('validation');
  my $result = FormValidator::Simple->check($q => $prof);
  # handle errors ...
}

SEE ALSO

DBIx::MoCo::DataBase, SQL::Abstract, Class::DBI, Cache,

AUTHOR

Junya Kondo, <jkondo@hatena.com>, Naoya Ito, <naoya@hatena.ne.jp>, Kentaro Kuribayashi, <kentarok@gmail.com>

COPYRIGHT AND LICENSE

Copyright (C) Hatena Inc. All Rights Reserved.

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