Relationships

We can link @Table in DBFlow via 1-1, 1-many, or many-to-many. For 1-1 we use @PrimaryKey, for 1-many we use @OneToMany, and for many-to-many we use the @ManyToMany annotation.

One To One

DBFlow supports multiple @ForeignKey right out of the box as well (and for the most part, they can also be @PrimaryKey).

@Table(database = AppDatabase::class)
class Dog(@PrimaryKey var name: String,
          @ForeignKey(tableClass = Breed::class)
          @PrimaryKey var breed: String,
          @ForeignKey var owner: Owner? = null)

@ForeignKey can only be a subset of types:

1. Model

2. Any field not requiring a TypeConverter. If not a Model or a table class, you must specify the tableClass it points to.

3. Cannot inherit @ForeignKey from non-model classes (see Inherited Columns)

If you create a circular reference (i.e. two tables with strong references to Model as @ForeignKey to each other), read on.

Stubbed Relationships

For efficiency reasons we recommend specifying @ForeignKey(stubbedRelationship = true). What this will do is only preset the primary key references into a table object. This only works with models that have fully mutable constructor properties.

All other fields will not be set. If you need to access the full object, you will have to call modelAdapter<MyStubbedModel>().load() .

From our previous example of Dog, instead of using a String field for breed we recommended by using a Breed. It is nearly identical, but the difference being we would then only need to call load() on the reference and it would query the Breed table for a row with the breed id. This also makes it easier if the table you reference has multiple primary keys, since DBFlow will handle the work for you.

Multiple calls to load() will query the DB every time, so call when needed. Also if you don't specify @Database(foreignKeyConstraintsEnforced = true), calling load() may not have any effect. Essentially without enforcing @ForeignKey at a SQLite level, you can end up with floating key references that do not exist in the referenced table.

In normal circumstances, for every load of a Dog object from the database, we would also do a load of related Owner. This means that even if multiple Dog say (50) all point to same owner we end up doing 2x retrievals for every load of Dog. Replacing that model field of Owner with a stubbed relationship prevents the extra N lookup time, leading to much faster loads of Dog.

Note: if you need more detailed information, you will still need to load the full data on each individual object.

Note: using stubbed relationships also helps to prevent circular references that can get you in a StackOverFlowError if two tables strongly reference each other in @ForeignKey.

Our modified example now looks like this:

@Table(database = AppDatabase::class)
class Dog(@PrimaryKey var name: String,
          @ForeignKey(stubbedRelationship = true)
          @PrimaryKey var breed: Breed? = null,
          @ForeignKey(stubbedRelationship = true)
          var owner: Owner? = null)

One To Many

In DBFlow, @OneToMany is an annotation that you provide to a method in your Model class that will allow management of those objects during CRUD operations. This can allow you to combine a relationship of objects to a single Model to happen together on load, save, insert, update, and deletion.

@Table(database = ColonyDatabase::class)
class Queen(@PrimaryKey(autoincrement = true)
            var id: Long = 0,
            var name: String? = null,
            @ForeignKey(saveForeignKeyModel = false)
            var colony: Colony? = null) : BaseModel() {

    @get:OneToMany
    val ants: List<Ant>? by oneToMany { select from Ant::class where Ant_Table.queen_id.eq(id) }

}

Note: This is not recommended to use heavily. It impacts performance exponentially and only recommended if you have a small set of parent objects that reference a subset of items in the DB.

Efficient Methods

When using @ManyToMany, by default we skip the Model methods in each retrieved Ant (in this example). If you have nested @ManyToMany (which should strongly be avoided), you can turn off the efficient operations. Call @OneToMany(efficientMethods = false) and it will instead loop through each model and perform save(), delete(), etc when the parent model is called.

Custom ForeignKeyReferences

When simple @ForeignKey annotation is not enough, you can manually specify references for your table:

@ForeignKey(saveForeignKeyModel = false,
references = {ForeignKeyReference(columnName = "colony", foreignKeyColumnName = "id")})
var colony: Colony? = null;

By default not specifying references will take each field and append "${foreignKeyFieldName}_${ForeignKeyReferenceColumnName}" to make the reference column name. So by default the previous example would use colony_id without references. With references it becomes colony.

Many To Many

In DBFlow many to many is done via source-gen. A simple table:

@Table(database = AppDatabase::class)
@ManyToMany(referencedTable = Follower::class)
class User(@PrimaryKey var id: Int = 0, @PrimaryKey var name: String = "")

Generates a @Table class named User_Follower, which DBFlow treats as if you coded the class yourself!:

@Table(
    database = TestDatabase.class
)
public final class User_Follower extends BaseModel {
  @PrimaryKey(
      autoincrement = true
  )
  long _id;

  @ForeignKey(
      saveForeignKeyModel = false
  )
  Follower follower;

  @ForeignKey(
      saveForeignKeyModel = false
  )
  User user;

  public final long getId() {
    return _id;
  }

  public final Followers getFollower() {
    return follower;
  }

  public final void setFollower(Follower param) {
    follower = param;
  }

  public final Users getUser() {
    return user;
  }

  public final void setUser(User param) {
    user = param;
  }
}

This annotation makes it very easy to generate "join" tables for you to use in the app for a ManyToMany relationship. It only generates the table you need. To use it you must reference it in code as normal.

Note: This annotation is only a helper to generate tables that otherwise you would have to write yourself. It is expected that management still is done by you, the developer.

Custom Column Names

You can change the name of the columns that are generated. By default they are simply lower case first letter version of the table name.

referencedTableColumnName -> Refers to the referenced table.

thisTableColumnName -> Refers to the table that is creating the reference.

Multiple ManyToMany

You can also specify @MultipleManyToMany which enables you to define more than a single @ManyToMany relationship on the table.

A class can use both:

@Table(database = TestDatabase::class)
@ManyToMany(referencedTable = TestModel1::class)
@MultipleManyToMany({@ManyToMany(referencedTable = TestModel2::class),
    @ManyToMany(referencedTable = TestModel3::class)})
class ManyToManyModel(
  @PrimaryKey var name: String = "",
  @PrimaryKey var id: Int = 0,
  @PrimaryKey var anotherColumn: Char? = null)

Last updated