= Association Basics This guide is based on http://guides.rubyonrails.org/association_basics.html == Why Associations? Associations exist to simplify code that deals with related rows in separate database tables. Without associations, if you had classes such as: class Artist < Sequel::Model end class Album < Sequel::Model end And you wanted to get all of the albums for a given artist (assuming each album was associated with only one artist): Album.filter(:artist_id=>@artist.id).all Or maybe you want to add an album for a given artist: Album.create(:artist_id=>@artist.id, :name=>'RF') With Associations, you can make the above code simpler, by setting up associations between the two models: class Artist < Sequel::Model one_to_many :albums end class Album < Sequel::Model many_to_one :artist end Then, the code to retrieve albums related to the artist is simpler: @artist.albums As is the code to add a related album to an artist: @artist.add_album(:name=>'RF') It also makes it easier to creating queries that use joins based on the association: Artist.association_join(:albums) # SELECT * FROM artists # INNER JOIN albums ON (albums.artist_id = artists.id) == The Types of Associations Sequel has five different association types built in: * many_to_one * one_to_many * one_to_one * many_to_many * one_through_one It ships with additional association types via plugins. === many_to_one The many_to_one association is used when the table for the current class contains a foreign key that references the primary key in the table for the associated class. It is named because there can be many rows in the current table for each row in the associated table. # Database schema: # albums artists # :id /--> :id # :artist_id --/ :name # :name class Album # Uses singular form of associated model name many_to_one :artist end === one_to_many and one_to_one The one_to_many association is used when the table for the associated class contains a foreign key that references the primary key in the table for the current class. It is named because for each row in the current table there can be many rows in the associated table: The one_to_one association can be thought of as a subset of the one_to_many association, but where there can only be either 0 or 1 records in the associated table. This is useful if there is a unique constraint on the foreign key field in the associated table. It's also useful if you want to impose an order on the association and just want the first record returned. # Database schema: # artists albums # :id <----\ :id # :name \----- :artist_id # :name class Artist # Uses plural form of associated model name one_to_many :albums # Uses singular form of associated model name one_to_one :album end === many_to_many and one_through_one The many_to_many association allows each row in the current table to be associated to many rows in the associated table, and each row in the associated table to many rows in the current table, by using a join table to associate the two tables. If you assume each artist can have multiple albums and each album can have multiple artists: The one_through_one association can be thought of as a subset of the many_to_many association, but where there can only be 0 or 1 records in the associated table. This is useful if there is a unique constraint on the foreign key in the join table that refrences the current table. It's also useful if you want to impose an order on the association and just want the first record returned. The one_through_one association is so named because it sets up a one-to-one association through a single join table. # Database schema: # albums # :id <----\ # :name \ albums_artists # \---- :album_id # artists /---- :artist_id # :id <-----/ # :name class Artist # Uses plural form of associated model name many_to_many :albums # Uses singular form of associated model name one_through_one :album end === Differences Between many_to_one and one_to_one If you want to setup a 1-1 relationship between two models, where the foreign key in one table references the associated table directly, you have to use many_to_one in one model, and one_to_one in the other model. How do you know which to use in which model? The simplest way to remember is that the model whose table has the foreign key uses many_to_one, and the other model uses one_to_one: # Database schema: # artists albums # :id <----\ :id # :name \----- :artist_id # :name class Artist one_to_one :album end class Album many_to_one :artist end == Most Common Options === :key The :key option must be used if the default column symbol that Sequel would use is not the correct column. For example: class Album # Assumes :key is :artist_id, based on association name of :artist many_to_one :artist end class Artist # Assumes :key is :artist_id, based on class name of Artist one_to_many :albums end However, if your schema looks like: # Database schema: # artists albums # :id <----\ :id # :name \----- :artistid # Note missing underscore # :name Then the default :key option will not be correct. To fix this, you need to specify an explicit :key option: class Album many_to_one :artist, :key=>:artistid end class Artist one_to_many :albums, :key=>:artistid end For many_to_many associations, the :left_key and :right_key options can be used to specify the column names in the join table, and the :join_table option can be used to specify the name of the join table: # Database schema: # albums # :id <----\ # :name \ albumsartists # \---- :albumid # artists /---- :artistid # :id <-----/ # :name class Artist # Note that :left_key refers to the foreign key pointing to the # current table, and :right_key the foreign key pointing to the # associated table. many_to_many :albums, :left_key=>:artistid, :right_key=>:albumid, :join_table=>:albumsartists end class Album many_to_many :artists, :left_key=>:albumid, :right_key=>:artistid, :join_table=>:albumsartists end === :class If the class of the association can not be guessed directly by looking at the association name, you need to specify it via the :class option. For example, if you have two separate foreign keys in the albums table that both point to the artists table, maybe to indicate one artist is the vocalist and one is the composer, you'd have to use the :class option: # Database schema: # artists albums # :id <----\ :id # :name \----- :vocalist_id # \---- :composer_id # :name class Album many_to_one :vocalist, :class=>:Artist many_to_one :composer, :class=>:Artist end class Artist one_to_many :vocalist_albums, :class=>:Album, :key=>:vocalist_id one_to_many :composer_albums, :class=>:Album, :key=>:composer_id end == Self-referential Associations Self-referential associations are easy to handle in Sequel. The simplest example is a tree structure: # Database schema: # nodes # :id <--\ # :parent_id ---/ # :name class Node many_to_one :parent, :class=>self one_to_many :children, :key=>:parent_id, :class=>self end For many_to_many self_referential associations, it's fairly similar. Here's an example of a directed graph: # Database schema: # nodes edges # :id <----------- :successor_id # :name \----- :predecessor_id class Node many_to_many :direct_predecessors, :left_key=>:successor_id, :right_key=>:predecessor_id, :join_table=>:edges, :class=>self many_to_many :direct_successors, :right_key=>:successor_id, :left_key=>:predecessor_id, :join_table=>:edges, :class=>self end == Methods Added When you create an association, it's going to add instance methods to the class related to the association. All associations are going to have an instance method added with the same name as the association: @artist.albums @album.artists many_to_one and one_to_one associations will also have a setter method added to change the associated object: @album.artist = Artist.create(:name=>'YJM') many_to_many and one_to_many associations will have three methods added: - add_* to associate an object to the current object - remove_* to disassociate an object from the current object - remove_all_* to dissociate all currently associated objects Examples: @artist.add_album(@album) @artist.remove_album(@album) @artist.remove_all_albums one_through_one associations do not have any modification methods added. Note that the remove_all_* method does not call remove hooks defined on the association, it just issues a single query to the database. If you want to remove all associated objects and call remove hooks, iterate over the array of associated objects and call remove_* for each. == Caching Associations are cached after being retrieved: @artist.album # Not cached - Database Query @artist.album # Cached - No Database Query @album.artists # Not cached - Database Query @album.artists # Cached - No Database Query You can choose to ignore the cached versions and do a database query to retrieve results by passing a true argument to the association method: @album.artists # Not cached - Database Query @album.artists # Cached - No Database Query @album.artists(true) # Ignore cache - Database Query If you reload/refresh the object, it will automatically clear the associations cache for the object: @album.artists # Not cached - Database Query @album.artists # Cached - No Database Query @album.reload @album.artists # Not Cached - Database Query If you want direct access to the associations cache, use the associations instance method: @album.associations # {} @album.associations[:artists] # nil @album.artists # [, ...] @album.associations[:artists] # [, ...] == Dataset Method In addition to the above methods, associations also add a instance method ending in +_dataset+ that returns a dataset representing the objects in the associated table: @album.artist_id # 10 @album.artist_dataset # SELECT * FROM artists WHERE (id = 10) LIMIT 1 @artist.id # 20 @artist.albums_dataset # SELECT * FROM albums WHERE (artist_id = 20) The association dataset is just like any other Sequel dataset, in that it can be further filtered, ordered, etc.: @artist.albums_dataset. where(Sequel.like(:name, 'A%')). order(:copies_sold). limit(10) # SELECT * FROM albums # WHERE ((artist_id = 20) AND (name LIKE 'A%' ESCAPE '\')) # ORDER BY copies_sold LIMIT 10 Records retrieved using the +_dataset+ method are not cached in the associations cache. @album.artists_dataset.all # [, ...] @album.associations[:artists] # nil == Dynamic Association Modification Similar to the +_dataset+ method, you can provide a block to the association method to customize the dataset that will be used to retrieve the records. So you can apply a filter in either of these two ways: @artist.albums_dataset.where(Sequel.like(:name, 'A%')) @artist.albums{|ds| ds.where(Sequel.like(:name, 'A%'))} While they both apply the same filter, using the +_dataset+ method does not apply any of the association callbacks or handle association reciprocals (see below for details about callbacks and reciprocals). Using a block instead handles all those things, and also caches its results in the associations cache (ignoring any previously cached value). == Filtering By Associations In addition to using the association method to get associated objects, you can also use associated objects in filters. For example, to get all albums for a given artist, you would usually do: @artist.albums # or @artist.albums_dataset for a dataset You can also do the following: Album.where(:artist=>@artist).all # or leave off the .all for a dataset For filtering by a single association, this isn't very useful. However, unlike using the association method, using a filter allows you to filter by multiple associations: Album.where(:artist=>@artist, :publisher=>@publisher) This will return all albums by that artist and published by that publisher. This isn't possible using just the association method approach, though you can combine the approaches: @artist.albums_dataset.where(:publisher=>@publisher) This doesn't just work for +many_to_one+ associations, it also works for the other associations: Album.one_to_one :album_info # The album related to that AlbumInfo instance Album.where(:album_info=>AlbumInfo[2]) Album.one_to_many :tracks # The album related to that Track instance Album.where(:tracks=>Track[3]) Album.many_to_many :tags # All albums related to that Tag instance Album.where(:tags=>Tag[4]) Album.one_through_one :tag # All albums related to that Tag instance Album.where(:tag=>Tag[4]) Note that for +one_to_many+ and +many_to_many+ associations, you still use the plural form even though only a single model object is given. You can also exclude by associations: Album.exclude(:artist=>@artist).all This will return all albums not by that artist. You can also provide an array with multiple model objects: Album.where(:artist=>[@artist1, @artist2]).all Similar to using an array of integers or strings, this will return all albums whose artist is one of those two artists. You can also use +exclude+ if you want all albums not by either of those artists: Album.exclude(:artist=>[@artist1, @artist2]).all If you are using a +one_to_many+ or +many_to_many+ association, you may want to return records where the records matches all of multiple records, instead of matching any of them. For example: Album.where(:tags=>[@tag1, @tag2]) This matches albums that are associated with either @tag1 or @tag2 or both. If you only want ones that you are associated with both, you can use separate filter calls: Album.where(:tags=>@tag1).where(:tags=>@tag2) Or the array form of condition specifiers: Album.where([[:tags, @tag1], [:tags, @tag2]]) These will return albums associated with both @tag1 and @tag2. You can also provide a dataset value when filtering by associations: Album.where(:artist=>Artist.where(Sequel.like(:name, 'A%'))).all This will return all albums whose artist starts with 'A'. Like the other forms, this can be inverted: Album.exclude(:artist=>Artist.where(Sequel.like(:name, 'A%'))).all This will return all albums whose artist does not start with 'A'. Filtering by associations even works for associations that have conditions added via the :conditions option or a block: Album.one_to_many :popular_tags, :clone=>:tags do |ds| ds.where{times_used > 1000} end Album.where(:popular_tags=>[@tag1, @tag2]) This will return all albums that whose popular tags would include at least one of those tags. Note that filtering by associations does not work for associations that use blocks with instance-specific code. == Name Collisions Because associations create instance methods, it's possible to override existing instance methods if you name an association the same as an existing method. For example, values and associations would be bad association names. == Database Schema Creating an association doesn't modify the database schema. Sequel assumes your associations reflect the existing database schema. If not, you should modify your schema before creating the associations. === many_to_one/one_to_many For example, for the following model code: class Album many_to_one :artist end class Artist one_to_many :albums end You probably want the following database schema: # albums artists # :id /--> :id # :artist_id --/ :name # :name Which could be created using the following Sequel code: DB.create_table(:artists) do # Primary key must be set explicitly primary_key :id String :name end DB.create_table(:albums) do primary_key :id # Table that foreign key references needs to be set explicitly # for a database foreign key reference to be created. foreign_key :artist_id, :artists String :name end If you already had a schema such as: # Database schema: # albums artists # :id :id # :name :name Then you just need to add the column: DB.alter_table(:albums) do add_foreign_key :artist_id, :artists end === many_to_many With many_to_many associations, the default join table for the association uses the sorted underscored names of both model classes. For example, with the following model code: class Album many_to_many :artists end class Artist many_to_many :albums end The default join table name would be albums_artists, not artists_albums, because: ["artists", "albums"].sort.join('_') # "albums_artists" Assume you already had the albums and artists tables created, and you just wanted to add an albums_artists join table to create the following schema: # Database schema: # albums # :id <----\ # :name \ albums_artists # \---- :album_id # artists /---- :artist_id # :id <-----/ # :name You could use the following Sequel code: DB.create_join_table(:album_id=>:albums, :artist_id=>:artists) # or DB.create_table(:albums_artists) do foreign_key :album_id, :albums foreign_key :artist_id, :artists end == Association Scope If you nest your Sequel::Model classes inside modules, then you should know that Sequel will only look in the same module for associations by default. So the following code will work fine: module App class Artist < Sequel::Model one_to_many :albums end class Album < Sequel::Model many_to_one :artist end end However, if you enclose your model classes inside two different modules, things will not work by default: module App1 class Artist < Sequel::Model one_to_many :albums end end module App2 class Album < Sequel::Model many_to_one :artist end end To fix this, you need to specify the full model class name using the :class option: module App1 class Artist < Sequel::Model one_to_many :albums, :class=>"App2::Album" end end module App2 class Album < Sequel::Model many_to_one :artist, :class=>"App1::Artist" end end If both classes are in the same module, but the default class name used is not correct, you need to specify the full class name with the :class option: module App1 class AlbumArtist < Sequel::Model one_to_many :albums end class Album < Sequel::Model many_to_one :artist, :class=>"App1::AlbumArtist" end end == Method Details In all of these methods, _association_ is replaced by the symbol you pass to the association. === _association_(reload = false) (e.g. albums) For +many_to_one+ and +one_to_one+ associations, the _association_ method returns either the single object associated, or nil if no object is associated. @artist = @album.artist For +one_to_many+ and +many_to_many+ associations, the _association_ method returns an array of associated objects, which may be empty if no objects are currently associated. @albums = @artist.albums === _association_=(object_to_associate) (e.g. artist=) [+many_to_one+ and +one_to_one+] The _association_= method sets up an association of the passed object to the current object. For +many_to_one+ associations, this sets the foreign key for the current object to point to the associated object's primary key. @album.artist = @artist For +one_to_one+ associations, this sets the foreign key of the associated object to the primary key value of the current object. For +many_to_one+ associations, this does not save the current object. For +one_to_one+ associations, this does save the associated object. === add_association(object_to_associate) (e.g. add_album) [+one_to_many+ and +many_to_many+] The add_association method associates the passed object to the current object. For +one_to_many+ associations, it sets the foreign key of the associated object to the primary key value of the current object, and saves the associated object. For +many_to_many+ associations, this inserts a row into the join table with the foreign keys set to the primary key values of the current and associated objects. Note that the singular form of the association name is used in this method. @artist.add_album(@album) In addition to passing an actual associated object, you can pass a hash, and a new associated object will be created from them: @artist.add_album(:name=>'RF') # creates Album object The add_association method returns the new associated object: @album = @artist.add_album(:name=>'RF') Note that the add_* methods for +one_to_many+ persist the changes by saving the passed in (or newly created) object. However, to avoid silent failures of these methods, they explicitly raise exceptions even when raise_on_save_failure is false for the associated model. You can disable this behavior (i.e. return nil instead of raising exceptions on a save failure) by setting the :raise_on_save_failure=>false option for the association. === remove_association(object_to_disassociate) (e.g. remove_album) [+one_to_many+ and +many_to_many+] The remove_association method disassociates the passed object from the current object. For +one_to_many+ associations, it sets the foreign key of the associated object to NULL, and saves the associated object. For +many_to_many+ associations, this deletes the matching row in the join table. Similar to the add_association method, the singular form of the association name is used in this method. @artist.remove_album(@album) Note that this does not delete @album from the database, it only disassociates it from the @artist. To delete @album from the database: @album.destroy The add_association and remove_association methods should be thought of as adding and removing from the association, not from the database. In addition to passing the object directly to remove_association, you can also pass the associated object's primary key: @artist.remove_album(10) This will look up the associated object using the key, and remove that album. The remove_association method returns the now disassociated object: @album = @artist.remove_album(10) === remove_all_association (e.g. remove_all_albums) [+one_to_many+ and +many_to_many+] The remove_all_association method disassociates all currently associated objects. For +one_to_many+ associations, it sets the foreign key of all associated objects to NULL in a single query. For +many_to_many+ associations, this deletes all matching rows in the join table. Unlike the add_association and remove_association method, the plural form of the association name is used in this method. The remove_all_association method returns the number of rows updated for +one_to_many+ associations and the number of rows deleted for +many_to_many+ associations: @rows_modified = @artist.remove_all_albums === association_dataset (e.g. albums_dataset) The association_dataset method returns a dataset that represents all associated objects. This dataset is like any other Sequel dataset, in that it can be filtered, ordered, etc.: ds = @artist.albums_dataset.where(Sequel.like(:name, 'A%')).order(:copies_sold) Unlike most other Sequel datasets, association datasets have a couple of added methods: ds.model_object # @artist ds.association_reflection # same as Artist.association_reflection(:albums) For a more info on Sequel's reflection capabilities see the {Reflection page}[rdoc-ref:doc/reflection.rdoc]. == Overriding Method Behavior Sequel is designed to be very flexible. If the default behavior of the association modification methods isn't what you desire, you can override the methods in your classes. However, you should be aware that for each of the association modification methods described, there is a private method that is preceeded by an underscore that does the actual modification. The public method without the underscore handles caching and callbacks, and shouldn't be overridden by the user. In addition to overriding the private method in your class, you can also use association options to change which method Sequel defines. The only difference between the two is that if you use an association option to change the method Sequel defines, you cannot call super to get the default behavior. === _association= (:setter option) Let's say you want to set a specific field whenever associating an object using the association setter method. For example, let's say you have a file_under column for each album to tell you where to file it. If the album is associated with an artist, it should be filed under the artist's name and the album's name, otherwise it should just use the album's name. class Album < Sequel::Model many_to_one :artist private def _artist=(artist) if artist self.artist_id = artist.id self.file_under = "#{artist.name}-#{name}" else self.artist_id = nil self.file_under = name end end end The above example is contrived, as you would generally use a before_save model hook to handle such a modification. However, if you only modify the album's artist using the artist= method, this approach may perform better. === \_add_association (:adder option) Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#add_album method: class Artist < Sequel::Model one_to_many :albums private def _add_album(album) album.update(:artist_id => id, :file_under=>"#{name}-#{album.name}") end end === \_remove_association (:remover option) Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#remove_album method: class Artist < Sequel::Model one_to_many :albums private def _remove_album(album) album.update(:artist_id => nil, :file_under=>album.name) end end === \_remove_all_association (:clearer option) Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#remove_all_albums method: class Artist < Sequel::Model one_to_many :albums private def _remove_all_albums # This is Dataset#update, not Model#update, so the :file_under=>:name # ends up being "SET file_under = name" in SQL. albums_dataset.update(:artist_id => nil, :file_under=>:name) end end == Association Options Sequel's associations mostly share the same options. For ease of understanding, they are grouped here by section === Association Dataset Modification Options ==== block All association defining methods take a block that is passed the default dataset and should return a modified copy of the dataset to use for the association. For example, if you wanted an association that returns all albums of an artist that went gold (sold at least 500,000 copies): Artist.one_to_many :gold_albums, :class=>:Album do |ds| ds.where{copies_sold > 500000} end The result of the block is cached as an optimization. One of the side effects of that is that if your block depends on external state, it won't work correctly unless you setup a delayed evaluation. For example: Artist.one_to_many :gold_albums, :class=>:Album do |ds| ds.where{copies_sold > $gold_limit} end In this case if you change $gold_limit later, the changes won't effect the association. If you want to pick up changes to $gold_limit, you need to setup a delayed evaluation: Artist.one_to_many :gold_albums, :class=>:Album do |ds| ds.where{copies_sold > Sequel.delay{$gold_limit}} end ==== :class This is the class of the associated objects that will be used. It's one of the most commonly used options. If it is not given, it guesses based on the name of the association. If a *_to_many association is used, uses the singular form of the association name. For example: Album.many_to_one :artist # guesses Artist Artist.one_to_many :albums # guesses Album However, for more complex associations, especially ones that add additional filters beyond the foreign/primary key relationships, the default class guessed will be wrong: # guesses GoldAlbum Artist.one_to_many :gold_albums do |ds| ds.where{copies_sold > 500000} end You can specify the :class option using the class itself, a Symbol, or a String: Album.many_to_one :artist, :class=>Artist # Class Album.many_to_one :artist, :class=>:Artist # Symbol Album.many_to_one :artist, :class=>"Artist" # String ==== :key For +many_to_one+ associations, this is the foreign_key in the current model's table that references the associated model's primary key as a symbol. Defaults to :association_id. Can use an array of symbols for a composite key association. Album.many_to_one :artist # :key=>:artist_id For +one_to_one+ and +one_to_many+ associations, is the foreign key in associated model's table that references current model's primary key, as a symbol. Defaults to :"#{self.name.underscore}_id". Artist.one_to_many :albums # :key=>:artist_id In both cases an array of symbols can be used for a composite key association: Apartment.many_to_one :building # :key=>[:city, :address] ==== :conditions The conditions to use to filter the association, can be any argument passed to +where+. If you use a hash or an array of two element arrays, this will also be used as a filter when using eager_graph to load the association. Artist.one_to_many :good_albums, :class=>:Album, :conditions=>{:good=>true} @artist.good_albums # SELECT * FROM albums WHERE ((artist_id = 1) AND (good IS TRUE)) ==== :order The column(s) by which to order the association dataset. Can be a singular column or an array. Artist.one_to_many :albums_by_name, :class=>:Album, :order=>:name Artist.one_to_many :albums_by_num_tracks, :class=>:Album, :order=>[:num_tracks, :name] ==== :select The columns to SELECT when loading the association. For most associations, it defaults to nil, so * is used. For +many_to_many+ associations, it defaults to the associated class's table_name.*, which means it doesn't include the columns from the join table. This is to prevent the common issue where the join table includes columns with the same name as columns in the associated table, in which case the joined table's columns would usually end up clobbering the values in the associated table. If you want to include the join table attributes, you can use this option, but beware that the join table columns can clash with columns from the associated table, so you should alias any columns that have the same name in both the join table and the associated table. Example: Artist.one_to_many :albums, :select=>[:id, :name] Album.many_to_many :tags, :select=>[Sequel.expr(:tags).*, :albums_tags__number] ==== :limit Limit the number of records to the provided value: Artist.one_to_many :best_selling_albums, :class=>:Album, :order=>:copies_sold, :limit=>5 # LIMIT 5 Use an array with two arguments for the value to specify a limit and an offset. Artist.one_to_many :next_best_selling_albums, :class=>:Album, :order=>:copies_sold, :limit=>[10, 5] # LIMIT 10 OFFSET 5 This probably doesn't make a lot of sense for *_to_one associations, though you could use it to specify an offset. ==== :join_table [+many_to_many+, +one_through_one+] Name of table that includes the foreign keys to both the current model and the associated model, as a symbol. Defaults to the name of current model and name of associated model, pluralized, underscored, sorted, and joined with '_'. Here's an example of the defaults: Artist.many_to_many :albums # :join_table=>:albums_artists Album.many_to_many :artists # :join_table=>:albums_artists Person.many_to_many :colleges # :join_table=>:colleges_people ==== :left_key [+many_to_many+, +one_through_one+] Foreign key in join table that points to current model's primary key, as a symbol. Defaults to :"#{model_name.underscore}_id". Album.many_to_many :tags # :left_key=>:album_id Can use an array of symbols for a composite key association. ==== :right_key [+many_to_many+, +one_through_one+] Foreign key in join table that points to associated model's primary key, as a symbol. Defaults to :"#{association_name.singularize}_id" for +many_to_many+ and :"#{association_name}_id" for +one_through_one+. Album.many_to_many :tags # :right_key=>:tag_id Can use an array of symbols for a composite key association. ==== :distinct Use the DISTINCT clause when selecting associating object, both when lazy loading and eager loading via eager (but not when using eager_graph). This is most useful for many_to_many associations that use join tables that contain more than just the foreign keys, where you are storing additional information. For example, if you have a database of people, degree types, and colleges, and you want to return all people from a given college, you may want to use :distinct so that if a person has two separate degrees from the same college, they won't show up twice. ==== :clone The :clone option clones an existing association, taking the options you specified for that association, and making a copy of them for this association. Other options provided by this association are then merged into the cloned options. This is commonly used if you have a bunch of similar associations that you want to DRY up: one_to_many :english_verses, :class=>:LyricVerse, :key=>:lyricsongid, :order=>:number, :conditions=>{:languageid=>1} one_to_many :romaji_verses, :clone=>:english_verses, :conditions=>{:languageid=>2} one_to_many :japanese_verses, :clone=>:english_verses, :conditions=>{:languageid=>3} Note that for the final two asociations, you didn't have to specify the :class, :key, or :order options, as they were copied by the :clone option. By specifying the :conditions option for the final two associations, it overrides the :conditions option of the first association, it doesn't attempt to merge them. In addition to the options hash, the :clone option will copy a block argument from the existing situation. If you want a cloned association to not have the same block as the association you are cloning from, specify the :block=>nil option in additon to the :clone option. ==== :dataset This is generally only specified for custom associations that aren't based on primary/foreign key relationships. It should be a proc that is instance_execed to get the base dataset to use before the other options are applied. If the proc accepts an argument, it is passed the related association reflection. For best performance, it's recommended that custom associations call the +associated_dataset+ method on the association reflection as the starting point for the dataset to return. The +associated_dataset+ method will return a dataset based on the associated class with most of the association options already applied, and the proc should return a modified copy of this dataset. Here's an example of an association of songs to artists through lyrics, where the artist can perform any one of four tasks for the lyric: Album.one_to_many :songs, :dataset=>(proc do |r| r.associated_dataset.select_all(:songs). join(Lyric, :id=>:lyricid, id=>[:composer_id, :arranger_id, :vocalist_id, :lyricist_id]) end) Artist.first.songs_dataset # SELECT songs.* FROM songs # INNER JOIN lyrics ON # lyrics.id = songs.lyric_id AND # 1 IN (composer_id, arranger_id, vocalist_id, lyricist_id) ==== :extend A module or array of modules to extend the dataset with. These are used to set up association extensions. For more information , please see the {Advanced Associations page}[rdoc-ref:doc/advanced_associations.rdoc]. ==== :primary_key [+many_to_one+, +one_to_one+, +one_to_many+] The column that the :key option references, as a symbol. For +many_to_one+ associations, this column is in the associated table. For +one_to_one+ and +one_to_many+ associations, this column is in the current table. In both cases, it defaults to the primary key of the table. Can use an array of symbols for a composite key association. Artist.set_primary_key :arid Artist.one_to_many :albums # :primary_key=>:arid Album.one_to_many :artist # :primary_key=>:arid ==== :left_primary_key [+many_to_many+, +one_through_one+] Column in current table that :left_key option points to, as a symbol. Defaults to primary key of current table. Album.set_primary_key :alid Album.many_to_many :tags # :left_primary_key=>:alid Can use an array of symbols for a composite key association. ==== :right_primary_key [+many_to_many+, +one_through_one+] Column in associated table that :right_key points to, as a symbol. Defaults to primary key of the associated table. Tag.set_primary_key :tid Album.many_to_many :tags # :right_primary_key=>:tid Can use an array of symbols for a composite key association. ==== :join_table_block [+many_to_many+, +one_through_one+] A proc that can be used to modify the dataset used in the add/remove/remove_all methods. It's separate from the association block, as that is called on a join of the join table and the associated table, whereas this option just applies to the join table. It can be used to make sure that filters are used when deleting. Artist.many_to_many :lead_guitar_albums, :join_table_block=>proc do |ds| ds.where(:instrument_id=>5) end === Callback Options All callbacks can be specified as a Symbol, Proc, or array of both/either specifying a callback to call. Symbols are interpreted as instance methods that are called with the associated object. Procs are called with the receiver as the first argument and the associated object as the second argument. If an array is given, all of them are called in order. Before callbacks are often used to check preconditions, they can return false to signal Sequel to abort the modification. If any before callback returns false, the remaining before callbacks are not called and modification is aborted. Before callbacks are also commonly used to modify the current object or the associated object. After callbacks are often used for notification (logging, email) after a successful modification has been made. ==== :before_add [+one_to_many+, +many_to_many+] Called before adding an object to the association: class Artist # Don't allow adding an album to an artist if it has no tracks one_to_many :albums, :before_add=>proc{|ar, al| false if al.num_tracks == 0} end ==== :after_add [+one_to_many+, +many_to_many+] Called after adding an object to the association: class Artist # Log all associations of albums to an audit logging table one_to_many :albums, :after_add=>:log_add_album private def log_add_album(album) DB[:audit_logs].insert(:log=>"Album #{album.inspect} associated to #{inspect}") end end ==== :before_remove [+one_to_many+, +many_to_many+] Called before removing an object from the association using remove_association: class Artist # Don't allow removing a self-titled album one_to_many :albums, :before_remove=>proc{|ar, al| false if al.name == ar.name} end This is not called when using remove_all_association. ==== :after_remove [+one_to_many+, +many_to_many+] Called after removing an object from the association using remove_association: class Artist # Log all disassociations of albums to an audit logging table one_to_many :albums, :after_remove=>:log_remove_album private def log_remove_album(album) DB[:audit_logs].insert(:log=>"Album #{album.inspect} disassociated from #{inspect}") end end This is not called when using remove_all_association. ==== :before_set [+many_to_one+, +one_to_one+] Called before the _association= method is called to modify the objects: class Album # Don't associate the album with an artist if the year the album was # released is less than the year the artist/band started. many_to_one :artist, :before_set=>proc{|al, ar| false if al.year < ar.year_started} end ==== :after_set [+many_to_one+, +one_to_one+] Called after the _association= method is called to modify the objects: class Album # Log all disassociations of albums to an audit logging table many_to_one :artist, :after_set=>:log_artist_set private def log_artist_set(artist) DB[:audit_logs].insert(:log=>"Artist for album #{inspect} set to #{artist.inspect}") end end ==== :after_load Called after retrieving the associated records from the database. class Artist # Cache all album names to a single string when retrieving the # albums. one_to_many :albums, :after_load=>:cache_album_names attr_reader :album_names private def cache_album_names(albums) @album_names = albums.map{|x| x.name}.join(", ") end end Generally used if you know you will always want a certain action done when retrieving the association. For +one_to_many+ and +many_to_many+ associations, both the argument to symbol callbacks and the second argument to proc callbacks will be an array of associated objects instead of a single object. ==== :uniq [+many_to_many+] Adds a after_load callback that makes the array of objects unique. In many cases, using the :distinct option is a better approach. === Eager Loading via eager (query per association) Options ==== :eager The associations to eagerly load via eager when loading the associated object(s). This is useful for example if you always want to eagerly load dependent associations when loading this association. For example, if you know that any time that you want to load an artist's albums, you are also going to want access to the album's tracks as well: # Eager load tracks when loading the albums Artist.one_to_many :albums, :eager=>:tracks You can also use a hash or array to specify multiple dependent associations to eagerly load: # Eager load the albums' tracks and the tracks' tags when loading the albums Artist.one_to_many :albums, :eager=>{:tracks=>:tags} # Eager load the albums' tags and tracks when loading the albums Artist.one_to_many :albums, :eager=>[:tags, :tracks] # Eager load the albums' tags, tracks, and tracks' tags when loading the albums Artist.one_to_many :albums, :eager=>[:tags, {:tracks=>:tags}] ==== :eager_loader A custom loader to use when eagerly load associated objects via eager. For many details and examples of custom eager loaders, please see the {Advanced Associations guide}[rdoc-ref:doc/advanced_associations.rdoc]. ==== :eager_loader_key A symbol for the key column to use to populate the key hash for the eager loader. Generally does not need to be set manually, defaults to the key method used. Can be set to nil to not populate the key hash (better for performance if a custom eager loader does not use the key_hash). ==== :eager_block If given, should be a proc to use instead of the association method block when eagerly loading. To not use a block when eager loading when one is used normally, should to nil. It's very uncommon to need this option. === Eager Loading via eager_graph (one query with joins) Options ==== :eager_graph The associations to eagerly load via eager_graph when loading the associated object(s). This is useful for example if you always want to eagerly load dependent associations when loading this association, but you want to filter or order the association based on dependent associations: Artist.one_to_many :albums_with_short_tracks, :class=>:Album, :eager_graph=>:tracks do |ds| ds.where{tracks__seconds < 120} end Artist.one_to_many :albums_by_track_name, :class=>:Album, :eager_graph=>:tracks do |ds| ds.order(:tracks__name) end You can also use a hash or array of arguments for :eager_graph, similar to what the :eager option accepts. ==== :graph_conditions The additional conditions to use on the SQL join when eagerly loading the association via eager_graph. Should be a hash or an array of two element arrays. If not specified, the :conditions option is used if it is a hash or array of two element arrays. Artist.one_to_many :active_albums, :class=>:Album, :graph_conditions=>{:active=>true} Note that these conditions on the association are in addition to the default conditions specified by the foreign/primary keys. If you want to replace the conditions specified by the foreign/primary keys, you need the :graph_only_conditions options. ==== :graph_block The block to pass to Dataset#join_table when eagerly loading the association via eager_graph. This is useful to specify conditions that can't be specified in a hash or array of two element arrays. Artist.one_to_many :gold_albums, :class=>:Album, :graph_block=>proc{|j,lj,js| Sequel.qualify(j, :copies_sold) > 500000} ==== :graph_join_type The type of SQL join to use when eagerly loading the association via eager_graph. Defaults to :left_outer. This is useful if you want to ensure that only artists that have albums are returned: Artist.one_to_many :albums, :graph_join_type=>:inner # Will exclude artists without an album Artist.eager_graph(:albums).all ==== :graph_select A column or array of columns to select from the associated table when eagerly loading the association via eager_graph. Defaults to all columns in the associated table. ==== :graph_only_conditions The conditions to use on the SQL join when eagerly loading the association via eager_graph, instead of the default conditions specified by the foreign/primary keys. This option causes the :graph_conditions option to be ignored. This can be useful if the keys you are using are strings and you want to do a case insensitive comparison. For example, let's say that instead of integer keys, you used string keys based on the album or artist name, and that the album was associated to the artist by name. However, you weren't enforcing case sensitivity between the keys, so you still want to return albums where the artist's name differs in case: Artist.one_to_many :albums, :key=>:artist_name, :graph_only_conditions=>nil, :graph_block=>proc{|j,lj,js| {Sequel.function(:lower, Sequel.qualify(j, :artist_name))=> Sequel.function(:lower, Sequel.qualify(lj, :name))}} Note how :graph_only_conditions is set to nil to ignore any existing conditions, and :graph_block is used to set up the case insensitive comparison. Another case where :graph_only_conditions may be used is if you want to use a JOIN USING or NATURAL JOIN for the graph: # JOIN USING Artist.one_to_many :albums, :key=>:artist_name, :graph_only_conditions=>[:artist_name] # NATURAL JOIN Artist.one_to_many :albums, :key=>:artist_name, :graph_only_conditions=>nil, :graph_join_type=>:natural ==== :graph_alias_base The base name to use for the table alias when eager graphing. Defaults to the name of the association. If the alias name has already been used in the query, Sequel will create a unique alias by appending a numeric suffix (e.g. alias_0, alias_1, ...) until the alias is unique. This is mostly useful if you have associations with the same name in many models, and you want to be able to easily tell which table alias corresponds to which association when eagerly graphing multiple associations with the same name. You can override this option on a per-graph basis by specifying the association as an SQL::AliasedExpression instead of a symbol: Album.eager_graph(Sequel.as(:artist, :a)) ==== :eager_grapher Sets up a custom grapher to use when eager loading the objects via eager_graph. This is the eager_graph analogue to the :eager_loader option. This isn't generally needed, as one of the other eager_graph related association options is usually sufficient. If specified, should be a proc that accepts a single hash argument, which will contain at least the following keys: :callback :: A callback proc used to dynamically modify the dataset to graph into the current dataset, before such graphing is done. This is nil if no callback proc is used. :implicit_qualifier :: The alias that was used for the current table (since you can cascade associations). :join_type :: Override the join type to use when graphing. :limit_strategy :: The limit strategy symbol to use when graphing (for limited associations only) :self :: The dataset that is doing the eager loading :table_alias :: An alias to use for the table to graph for this association. Example: Artist.one_to_many :self_title_albums, :class=>:Album, :eager_grapher=>(proc do |eo| eo[:self].graph(Album, {:artist_id=>:id, :name=>:name}, :table_alias=>eo[:table_alias], :implicit_qualifier=>eo[:implicit_qualifier]) end) ==== :order_eager_graph Whether to add the order to the dataset's order when graphing via eager_graph. Defaults to true, so set to false to disable. Sequel has to do some guess work when attempting to add the association's order to an eager_graphed dataset. In most cases it does so correctly, but if it has problems, you'll probably want to set this option to false. ==== :graph_order Override the order added when using eager_graph, instead of using the one defined in :order. This is useful if :order contains qualified identifiers, as the qualifiers may not match the aliases automatically used by eager_graph. This should contain unqualified identifiers, and eager_graph will automatically qualify them with the appropriate alias. ==== :graph_join_table_conditions [+many_to_many+, +one_through_one+] The additional conditions to use on the SQL join for the join table when eagerly loading the association via eager_graph. Should be a hash or an array of two element arrays. Let's say you have a database of people, colleges, and a table called degrees_received that includes a string field specifying the name of the degree, and you want to eager load all colleges for people where the person has received a specific degree: Person.many_to_many :bs_degree_colleges, :class=>:College, :join_table=>:degrees_received, :graph_join_table_conditions=>{:degree=>'BS'} ==== :graph_join_table_block [+many_to_many+, +one_through_one+] The block to pass to join_table for the join table when eagerly loading the association via eager_graph. This is used for similar reasons as :graph_block, but is only used for +many_to_many+ associations when graphing the join table into the dataset. It's used in the same place as :graph_join_table_conditions but like :graph_block, is needed for situations where the conditions can't be specified as a hash or array of two element arrays. Let's say you have a database of people, colleges, and a table called degrees_received that includes a string field specifying the name of the degree, and you want to eager load all colleges for people where the person has received a bachelor's degree (degree starting with B): Person.many_to_many :bachelor_degree_colleges, :class=>:College, :join_table=>:degrees_received, :graph_join_table_block=>proc{|j,lj,js| Sequel.qualify(j, :degree).like('B%')} This should be done when graphing the join table, instead of when graphing the final table, as :degree is a column of the join table. ==== :graph_join_table_join_type [+many_to_many+, +one_through_one+] The type of SQL join to use for the join table when eagerly loading the association via eager_graph. Defaults to the :graph_join_type option or :left_outer. This exists mainly for consistency in the unlikely case that you want to use a different join type when JOINing to the join table then you want to use for JOINing to the final table ==== :graph_join_table_only_conditions [+many_to_many+, +one_through_one+] The conditions to use on the SQL join for the join table when eagerly loading the association via eager_graph, instead of the default conditions specified by the foreign/primary keys. This option causes the :graph_join_table_conditions option to be ignored. This is only useful if you want to replace the default foreign/primary key conditions that Sequel would use when eagerly graphing. === Associations Based on SQL Expressions Options Sequel's associations can work not just with columns, but also with arbitrary SQL expressions. For example, on PostgreSQL, you can store foreign keys to other tables in hstore, json, or jsonb columns, and Sequel can automatically work with such constructs, including full support for eager loading. There's actually two parts to supporting associations based on SQL expressions. First is you must have an instance method in the model that returns the value that the SQL expression would return. Second is you must have an SQL expression object. If Sequel has access to a model instance and needs to get the value of the expression, it calls the method to get the value. If Sequel does not have access to a model instance, but needs to use the SQL expression in a query, it will use the SQL expression object. Below is an example storing foreign keys to other tables in a PostgreSQL hstore column, using the +pg_hstore+ and +pg_hstore_ops+ extensions. # Example schema: # albums artists # :id /---> :id # :meta ---/ :name # :name class Album < Sequel::Model many_to_one :artist, :key_column=>Sequel.cast(Sequel.hstore(:meta)['artist_id'], Integer) def artist_id meta['artist_id'].to_i end end class Artist < Sequel::Model one_to_many :albums, :key=>Sequel.cast(Sequel.hstore(:meta)['artist_id'], Integer), :key_method=>:artist_id end # Example schema: # albums albums_artists artists # :id <----- :meta -------> :id # :name :name class Album < Sequel::Model many_to_many :artists, :left_key=>Sequel.cast(Sequel.hstore(:meta)['album_id'], Integer), :right_key=>Sequel.cast(Sequel.hstore(:meta)['artist_id'], Integer) end class Artist < Sequel::Model many_to_many :albums, :left_key=>Sequel.cast(Sequel.hstore(:meta)['artist_id'], Integer), :right_key=>Sequel.cast(Sequel.hstore(:meta)['album_id'], Integer) end ==== :key_column [+many_to_one+] Like the :key option, but :key references the method name, while :key_column references the underlying column/expression. ==== :primary_key_method [+many_to_one+] Like the :primary_key option, but :primary_key references the column/expression name, while :primary_key_method references the method name. ==== :primary_key_column [+one_to_many+, +one_to_one+] Like the :primary_key option, but :primary_key references the method name, while :primary_key_column references the underlying column/expression. ==== :key_method [+one_to_many+, +one_to_one+] Like the :key option, but :key references the column/expression name, while :key_method references the method name. ==== :left_primary_key_column [+many_to_many+, +one_through_one+] Like the :left_primary_key option, but :left_primary_key references the method name, while :left_primary_key_column references the underlying column/expression. ==== :right_primary_key_method [+many_to_many+, +one_through_one+] Like the :right_primary_key option, but :right_primary_key references the column/expression name, while :right_primary_key_method references the method name. === Private Method Overriding Options These options override the private methods that Sequel defines to do the actual work of associating and deassociating objects. ==== :setter [*_to_one associations] This overrides the default behavior when you call an association setter method. Let's say you want to set a specific field whenever associating an object using the association setter method. class Album < Sequel::Model many_to_one :artist, :setter=>(proc do |artist| if artist self.artist_id = artist.id self.file_under = "#{artist.name}-#{name}" else self.artist_id = nil self.file_under = name end end) end ==== :adder [*_to_many associations] Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#add_album method: class Artist < Sequel::Model one_to_many :albums, :adder=>(proc do |album| album.update(:artist_id => id, :file_under=>"#{name}-#{album.name}") end) end ==== :remover [*_to_many associations] Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#remove_album method: class Artist < Sequel::Model one_to_many :albums, :remover=>(proc do |album| album.update(:artist_id => nil, :file_under=>album.name) end) end ==== :clearer [*_to_many associations] Continuing with the same example, here's how you would handle the same case if you also wanted to handle the Artist#remove_all_albums method: class Artist < Sequel::Model one_to_many :albums, :clearer=>(proc do albums_dataset.update(:artist_id => nil, :file_under=>:name) end) end === Advanced Options ==== :reciprocal The symbol name of the reciprocal association, if it exists. By default, Sequel will try to determine it by looking at the associated model's associations for a association that matches the current association's key(s). Set to nil to not use a reciprocal. Reciprocals are used in Sequel to modify the matching cached associations in associated objects when calling association methods on the current object. For example, when you retrieve objects in a one_to_many association, it'll automatically set the matching many_to_one association in the associated objects. The result of this is that code that does this: @artist.albums.each{|album| album.artist.name} only does one database query, because when the @artist's albums are retrieved, the cached artist association for each album is set to @artist. In addition to the one_to_many retrieval case, the association modification methods affect the reciprocals as well: # Sets the cached artist association for @album to @artist @artist.add_album(@album) # Sets the cached artist association for @album to nil @artist.remove_album(@album) # Sets the cached artist association to nil for the @artist's # cached albums association @artist.remove_all_albums # Remove @album from the artist1's cached albums association, and add @album # to @artist2's cached albums association. @album.artist # @artist1 @album.artist = @artist2 Sequel can usually guess the correct reciprocal, but if you have multiple associations to the same associated class that use the same keys, you may want to specify the :reciprocal option manually to ensure the correct one is used. ==== :read_only For +many_to_one+ and +one_to_one+ associations, do not add a setter method. For +one_to_many+ and +many_to_many+, do not add the add_association, remove_association, or remove_all_association methods. If the default modification methods would not do what you want, and you don't plan on overriding the internal modification methods to do what you want, it may be best to set this option to true. ==== :validate Set to false to not validate when implicitly saving any associated object. When using the +one_to_many+ association modification methods, the +one_to_one+ setter method, or creating a new object by passing a hash to the add_association method, Sequel will automatically save the object. If you don't want to validate objects when these implicit saves are done, the validate option should be set to false. ==== :raise_on_save_failure [+one_to_many+ associations] Set to false to not raise an exception when validation or a before hook fails when implicitly saving an associated object in the add_* or remove_* methods. This mirrors the raise_on_save_failure model setting, which these methods do not respect (by design, since the setting is dangerous). If you use this option, you must explicitly check all add_* and remove_* return values to see if they were successful. ==== :allow_eager If set to false, you cannot load the association eagerly via eager or eager_graph. Artist.one_to_many :albums, :allow_eager=>false Artist.eager(:albums) # Raises Sequel::Error This is usually used if the association dataset depends on specific values in model instance that would not be valid when eager loading for multiple instances. ==== :instance_specific This allows you to override the setting of whether the dataset contains instance specific code. For example, if you are passing a block to the association, Sequel sets this to true by default, which disables some optimizations that would be invalid if the association is instance specific. If you know that the block does not contain instance specific code, you can set this to false to reenable the optimizations. ==== :cartesian_product_number The number of joins completed by this association that could cause more than one row for each row in the current table (default: 0 for *_one associations, 1 for *_to_many associations). This should only be modified in specific cases. For example, if you have a one_to_one association that can actually return more than one row (where the default association method will just return the first), or a many_to_many association where there is a unique index in the join table so that you know only one object will ever be associated through the association. ==== :methods_module The module that the methods created by the association will be placed into. Defaults to the module containing the model's columns. This is not included in the model's class, so you are responsible for doing that manually. This is only useful in rare cases, such as when a plugin that adds associations depends on another plugin that defines instance methods of the same name. In that case, the instance methods of the dependent plugin would override the association methods created by the main plugin. ==== :eager_limit_strategy This setting determines what strategy to use for eager loading the associations that use the :limit setting to limit the number of returned records. You can't use LIMIT directly, since you want a limit for each group of associated records, not a LIMIT on the total number of records returned by the dataset. In general, Sequel picks an appropriate strategy, so it is not usually necessary to specify a strategy. You can specify true for this option to have Sequel choose which strategy to use (this is the default). You can specify a symbol to manually choose a strategy. The available strategies are: :union :: Uses one or more UNION queries with a subquery for each record you are eagerly loading for (this is the default strategy). :distinct_on :: Uses DISTINCT ON to ensure only the first matching record is loaded (only used for one_*_one associations without offsets on PostgreSQL). :window_function :: Uses a ROW_NUMBER window functions to ensure the correctly limited/offset records are returned. :ruby :: Uses ruby array slicing to emulate database limiting/offsetting. ==== :subqueries_per_union The number of subqueries per union query to use when eager loading for a limited association using a union strategy. This defaults to 40, but the optimum number depends on the database in use and the latency between the database and the application. ==== :filter_limit_strategy The strategy to use when filtering by limited associations. In general Sequel will choose either a :distinct_on, :window_function, or :correlated_subquery strategy based on the association type and what the database supports, but you can override that if necessary using this option.