Database.Beam.Schema

Allows introspection into database types.

All database types must be of kind '(Type -> Type) -> Type'. If the type parameter is named f, each field must be of the type of f applied to some type for which an IsDatabaseEntity instance exists.

The be type parameter is necessary so that the compiler can ensure that backend-specific entities only work on the proper backend.

Entities are documented under the corresponding section and in the manual

When parameterized by this entity tag, a database type will hold meta-information on the Haskell mappings of database entities. Under the hood, each entity type is transformed into its DatabaseEntityDescriptor type. For tables this includes the table name as well as the corresponding TableSettings, which provides names for each column.

Represents a meta-description of a particular entityType. Mostly, a wrapper around 'DatabaseEntityDescriptor be entityType', but carries around the IsDatabaseEntity dictionary.

The big Kahuna! All beam tables implement this class.

The kind of all table types is '(Type -> Type) -> Type'. This is because all table types are actually table type constructors. Every table type takes in another type constructor, called the column tag, and uses that constructor to instantiate the column types. See the documentation for Columnar.

This class is mostly Generic-derivable. You need only specify a type for the table's primary key and a method to extract the primary key given the table.

An example table:

data BlogPostT f = BlogPost
                 { _blogPostSlug    :: Columnar f Text
                 , _blogPostBody    :: Columnar f Text
                 , _blogPostDate    :: Columnar f UTCTime
                 , _blogPostAuthor  :: PrimaryKey AuthorT f
                 , _blogPostTagline :: Columnar f (Maybe Text)
                 , _blogPostImageGallery :: PrimaryKey ImageGalleryT (Nullable f) }
                   deriving Generic
instance Beamable BlogPostT
instance Table BlogPostT where
   data PrimaryKey BlogPostT f = BlogPostId (Columnar f Text) deriving Generic
   primaryKey = BlogPostId . _blogPostSlug
instance Beamable (PrimaryKey BlogPostT)

We can interpret this as follows:

• The _blogPostSlug, _blogPostBody, _blogPostDate, and _blogPostTagline fields are of types Text, Text, UTCTime, and 'Maybe Text' respectfully.

• Since _blogPostSlug, _blogPostBody, _blogPostDate, _blogPostAuthor must be provided (i.e, they cannot contain Nothing), they will be given SQL NOT NULL constraints. _blogPostTagline is declared Maybe so Nothing will be stored as NULL in the database. _blogPostImageGallery will be allowed to be empty because it uses the Nullable tag modifier.

• blogPostAuthor references the AuthorT table (not given here) and is required.

• blogPostImageGallery references the ImageGalleryT table (not given here), but this relation is not required (i.e., it may be Nothing. See Nullable).

Provides a number of introspection routines for the beam library. Allows us to "zip" tables with different column tags together. Always instantiate an empty Beamable instance for tables, primary keys, and any type that you would like to embed within either. See the manual for more information on embedding.

Return a TableSettings for the appropriate table type where each column has been given its default name. See the manual for information on the default naming convention.

Synonym for primaryKey

A type family that we use to "tag" columns in our table datatypes.

This is what allows us to use the same table type to hold table data, describe table settings, derive lenses, and provide expressions.

The basic rules are

Columnar Identity x = x

Thus, any Beam table applied to Identity will yield a simplified version of the data type, that contains just what you'd expect.

The Nullable type is used when referencing PrimaryKeys that we want to include optionally. For example, if we have a table with a PrimaryKey, like the following

data BeamTableT f = BeamTableT
                  { _refToAnotherTable :: PrimaryKey AnotherTableT f
                  , ... }

we would typically be required to provide values for the PrimaryKey embedded into BeamTableT. We can use Nullable to lift this constraint.

data BeamTableT f = BeamTableT
                  { _refToAnotherTable :: PrimaryKey AnotherTableT (Nullable f)
                  , ... }

Now we can use just_ and nothing_ to refer to this table optionally. The embedded PrimaryKey in _refToAnotherTable automatically has its fields converted into Maybe using Nullable.

The last Columnar rule is

Columnar f x = f x

Use this rule if you'd like to parameterize your table type over any other functor. For example, this is used in the query modules to write expressions such as 'TableT QExpr', which returns a table whose fields have been turned into query expressions.

The other rules are used within Beam to provide lenses and to expose the inner structure of the data type.

A short type-alias for Columnar. May shorten your schema definitions

If you declare a function 'Columnar f a -> b' and try to constrain your function by a type class for f, GHC will complain, because f is ambiguous in 'Columnar f a'. For example, 'Columnar Identity (Maybe a) ~ Maybe a' and 'Columnar (Nullable Identity) a ~ Maybe a', so given a type 'Columnar f a', we cannot know the type of f.

Thus, if you need to know f, you can instead use Columnar'. Since its a newtype, it carries around the f paramater unambiguously. Internally, it simply wraps 'Columnar f a'

Support for NULLable Foreign Key references.

data MyTable f = MyTable
               { nullableRef :: PrimaryKey AnotherTable (Nullable f)
               , ... }
                deriving (Generic, Typeable)

See Columnar for more information.

Metadata for a field of type ty in table.

Essentially a wrapper over the field name, but with a phantom type parameter, so that it forms an appropriate column tag.

Usually you use the defaultDbSettings function to generate an appropriate naming convention for you, and then modify it with withDbModification if necessary. Under this scheme, the field n be renamed using the IsString instance for TableField, or the fieldNamed function.

Van Laarhoven lens to retrieve or set the field name from a TableField.

Represents a table that contains metadata on its fields. In particular, each field of type 'Columnar f a' is transformed into 'TableField table a'. You can get or update the name of each field by using the fieldName lens.

The regular Haskell version of the table. Equivalent to 'tbl Identity'

Automatically provide names for tables, and descriptions for tables (using defTblFieldSettings). Your database must implement Generic, and must be auto-derivable. For more information on name generation, see the manual

Like tableLenses but for types that are instances of Database. Instead of pattern matching on LensFor, pattern match on TableLens.

Automatically deduce lenses for a table over any column tag. lenses at global level by doing a top-level pattern match on tableLenses, replacing every column in the pattern with `LensFor nameOfLensForField'. The lenses are generated per-column, not per field in the record. Thus if you have nested Beamable types, lenses are generated for each nested field.

For example,

data AuthorT f = AuthorT
               { _authorEmail     :: Columnar f Text
               , _authorFirstName :: Columnar f Text
               , _authorLastName  :: Columnar f Text }
                 deriving Generic

data BlogPostT f = BlogPost
                 { _blogPostSlug    :: Columnar f Text
                 , _blogPostBody    :: Columnar f Text
                 , _blogPostDate    :: Columnar f UTCTime
                 , _blogPostAuthor  :: PrimaryKey AuthorT f
                 , _blogPostTagline :: Columnar f (Maybe Text) }
                   deriving Generic
instance Table BlogPostT where
   data PrimaryKey BlogPostT f = BlogPostId (Columnar f Text)
   primaryKey = BlogPostId . _blogPostSlug
instance Table AuthorT where
   data PrimaryKey AuthorT f = AuthorId (Columnar f Text)
   primaryKey = AuthorId . _authorEmail

BlogPost (LensFor blogPostSlug
         (LensFor blogPostBody)
         (LensFor blogPostDate)
         (AuthorId (LensFor blogPostAuthorEmail))
         (LensFor blogPostTagLine) = tableLenses

Note: In order to have GHC deduce the right type, you will need to turn off the monomorphism restriction. This is a part of the Haskell standard that specifies that top-level definitions must be inferred to have a monomorphic type. However, lenses need a polymorphic type to work properly. You can turn off the monomorphism restriction by enabling the NoMonomorphismRestriction extension. You can do this per-file by using the {-# LANGUAGE NoMonomorphismRestriction #-} pragma at the top of the file. You can also pass the -XNoMonomorphismRestriction command line flag to GHC during compilation.