In our example domain, a blog entry "has many" tags, but because a tag also "has many" blog entries, there has to be an association table mapping between the two. This kind of many-to-many relationship requires a third model to be present; it models the association mapping through which the native and foreign models relate to each other.
As such, setup for a many-to-many relationship is slightly more complicated than a regular has-many/belongs-to. First you have to relate to the association model, then you have to relate to the foreign model through the association. Do so for each side of the many-to-many relationship.
<?php
class Acme_Model_Blogs extends Acme_Sql_Model
{
protected function _setup()
{
// the "through" related association
$this->_hasMany('taggings');
// has many tags, through this relationship name
$this->_hasManyThrough('tags', 'taggings');
}
}
class Acme_Model_Tags extends Acme_Sql_Model
{
protected function _setup()
{
// the "through" related association
$this->_hasMany('taggings');
// has many tags, through this relationship name
$this->_hasManyThrough('blogs', 'taggings');
}
}
Finally, in the association model, state that it belongs-to both sides of the many-to-many relationship.
<?php
class Acme_Model_Taggings extends Acme_Sql_Model
{
protected function _setup()
{
$this->_belongsTo('blog');
$this->_belongsTo('tag');
}
}
The above code creates a pseudo-property on each blog record called
$tags
that gets lazy-loaded by a fetch from
the tags
entry in the model catalog. That is,
each time you ask for a $blog_record->tags
, the model
system will fetch the related tag collection from the database
and place it in $tags
for you. The SQL for
the lazy-load looks similar to this:
SELECT tags.*
FROM tags AS tags
LEFT JOIN taggings AS taggings ON taggings.tag_id = tags.id
WHERE taggings.blog_id = {$blog_record->id}
Similarly, the above code creates a pseudo-property on each tag record called
$blogs
that gets lazy-loaded by a fetch from
the blogs
entry in the model catalog. That is,
each time you ask for a $tag_record->blogs
, the model
system will fetch the related blog collection from the database
and place it in $blogs
for you. The SQL for
the lazy-load looks similar to this:
SELECT blogs.*
FROM blogs AS blogs
LEFT JOIN taggings AS taggings ON taggings.blog_id = blogs.id
WHERE taggings.tag_id = {$tag_record->id}
Lazy vs. Eager Loading | |
---|---|
Beware the dreaded N+1 problem! If you fetch a collection of
10 blog records and loop through them, displaying the
|
In a has-many-through relationship, the foreign key exists in the
"through" association model/table, and identifies which mappings
belong to the native model. In our example domain, where a blog
has many tags through taggings, this means the blog foreign key
blog_id
is on the taggings table, and maps to
the primary key on the blogs table. Similarly, where a tag has
many blogs through taggings, the tag foreign key
tag_id
is on the taggings table, and maps to the
primary key on the tags table.
In a has-many-through relationship, there is a second foreign key
to keep track of, the "through" key. The "through" key exists in
the "through" model/table, and identifies which mappings belong to
the foreign model. In our example domain, where a blog has many
tags through taggings, the through key tag_id
is
on the taggings table, and maps to the primary key on the tags
table. Similarly, where a tag has many blogs through taggings, the
through key blog_id
is on the taggings table,
and maps to the primary key on the blogs table.
Although the Solar model system makes some assumptions about how
tables and foreign keys are named, these assumptions are not
hard-coded. You can define every aspect of the foreign
relationship using an array of key-value pairs passed to the
_hasManyThrough()
method.
<?php
$this->_hasManyThrough('foreigns', 'mappings', array(
'option' => 'value',
// ...
));
-
string
foreign_name
-
Normally, the model system expects the foreign model to be in the catalog uner the plural form of the related name. E.g.,
{native} _hasManyThrough('foreigns', 'mappings')
means the foreign model on the other side of the mapping relationship uses the catalog name 'foreigns'.Use the
foreign_name
option to specify a different catalog name for the foreign model on the other side of the related mappings. This allows you to name the relationship "foo" (and thereby the record property for the foreign) but use the catalog name "bar" to do the work. (Theforeign_class
option takes precedence overforeign_name
.) -
string
foreign_key
-
Normally, the model system expects the foreign key on the foreign table to be the same as the foreign model's
primary_col
value.Use the
foreign_key
option to specify a different column in the foreign table. (This option will be ignored if a specificnative_col
orforeign_col
option is set.) -
string
through_key
-
Normally, the model system expects the through key on the through table to be the same as the one defined by the foreign model's
foreign_col
value.Use the
through_key
option to specify a different column in the through table. (This option will be ignored if a specificnative_col
orforeign_col
option is set.) -
array
cols
-
Fetch these columns for the foreign records.
-
array
conditions
-
Additional conditions on the foreign table when fetching related records. These will be used as WHERE conditions, or as JOIN ON conditions, as appropriate for the kind of fetch being performed (e.g. lazy vs. eager fetch).
-
array
through_conditions
-
Additional conditions on the through table when fetching related records. These will be used as WHERE conditions, or as JOIN ON conditions, as appropriate for the kind of fetch being performed (e.g. lazy vs. eager fetch).
-
array
order
-
Additional ORDER clauses when fetching related records.
Note | |
---|---|
The following are some less common, but more advanced and more finely grained, options for defining relationships: |
-
string
foreign_class
-
The class name of the foreign model on the other side of the association mapping. Default is the first matching class for the relationship name, as loaded from the parent class stack.
-
string
foreign_alias
-
Aliases the foreign table using this name. Default is the relationship name.
-
string
foreign_col
-
The name of the column to join with in the foreign table, matching against the
through_foreign_col
column in the through table. -
string
through_foreign_col
-
The name of the column to join with in the through table, matching against the
foreign_col
column in the foreign table. -
string
through native_col
-
The name of column to join with in the through table, matching against the
native_col
column in the native table. -
string
native_col
-
The name of column to join with in the native table, matching against the
through_native_col
column in the through table. -
string
native_by
-
The strategy to be used for connecting to native records when eager-fetching: 'wherein', meaning a "WHERE IN (...)" a list of native IDs, or 'select', meaning a join against a sub-SELECT.
-
int
wherein_max
-
When picking a native-by strategy, use 'wherein' for up to this many records in the native result; after this point, use a 'select' strategy.
-
string
merge
-
Indicates the strategy to use for merging joined foreign rows; 'server' means the database will do it via a single SELECT combined into the native fetch (only possible with to-one relationships), whereas 'client' means PHP will do it, using one additional SELECT for the relationship (always for to-many relationships, optionally for to-one relationships) and then merging the rows in a PHP loop.