persistent-2.1: Type-safe, multi-backend data serialization.

Safe HaskellNone




data Checkmark Source

A Checkmark should be used as a field type whenever a uniqueness constraint should guarantee that a certain kind of record may appear at most once, but other kinds of records may appear any number of times.

NOTE: You need to mark any Checkmark fields as nullable (see the following example).

For example, suppose there's a Location entity that represents where a user has lived:

    user    UserId
    name    Text
    current Checkmark nullable

    UniqueLocation user current

The UniqueLocation constraint allows any number of Inactive Locations to be current. However, there may be at most one current Location per user (i.e., either zero or one per user).

This data type works because of the way that SQL treats NULLable fields within uniqueness constraints. The SQL standard says that NULL values should be considered different, so we represent Inactive as SQL NULL, thus allowing any number of Inactive records. On the other hand, we represent Active as TRUE, so the uniqueness constraint will disallow more than one Active record.

Note: There may be DBMSs that do not respect the SQL standard's treatment of NULL values on uniqueness constraints, please check if this data type works before relying on it.

The SQL BOOLEAN type is used because it's the smallest data type available. Note that we never use FALSE, just TRUE and NULL. Provides the same behavior Maybe () would if () was a valid PersistField.



When used on a uniqueness constraint, there may be at most one Active record.


When used on a uniqueness constraint, there may be any number of Inactive records.

data WhyNullable Source

The reason why a field is nullable is very important. A field that is nullable because of a Maybe tag will have its type changed from A to Maybe A. OTOH, a field that is nullable because of a nullable tag will remain with the same type.

newtype DBName Source




unDBName :: Text

data FieldDef Source




fieldHaskell :: !HaskellName

name of the field

fieldDB :: !DBName
fieldType :: !FieldType
fieldSqlType :: !SqlType
fieldAttrs :: ![Attr]

user annotations for a field

fieldStrict :: !Bool

a strict field in the data type. Default: true

fieldReference :: !ReferenceDef

data ReferenceDef Source

There are 3 kinds of references 1) composite (to fields that exist in the record) 2) single field 3) embedded


ForeignRef !HaskellName !FieldType

A ForeignRef has a late binding to the EntityDef it references via HaskellName and has the Haskell type of the foreign key in the form of FieldType

EmbedRef EmbedEntityDef 
CompositeRef CompositeDef 

data EmbedEntityDef Source

An EmbedEntityDef is the same as an EntityDef But it is only used for fieldReference so it only has data needed for embedding

data EmbedFieldDef Source

An EmbedFieldDef is the same as a FieldDef But it is only used for embeddedFields so it only has data needed for embedding

type ForeignFieldDef = (HaskellName, DBName) Source

Used instead of FieldDef to generate a smaller amount of code

data PersistValue Source

A raw value which can be stored in any backend and can be marshalled to and from a PersistField.


PersistText Text 
PersistByteString ByteString 
PersistInt64 Int64 
PersistDouble Double 
PersistRational Rational 
PersistBool Bool 
PersistDay Day 
PersistTimeOfDay TimeOfDay 
PersistUTCTime UTCTime 
PersistList [PersistValue] 
PersistMap [(Text, PersistValue)] 
PersistObjectId ByteString

Intended especially for MongoDB backend

PersistDbSpecific ByteString

Using PersistDbSpecific allows you to use types specific to a particular backend For example, below is a simple example of the PostGIS geography type:

data Geo = Geo ByteString

instance PersistField Geo where
  toPersistValue (Geo t) = PersistDbSpecific t

  fromPersistValue (PersistDbSpecific t) = Right $ Geo $ Data.ByteString.concat ["'", t, "'"]
  fromPersistValue _ = Left "Geo values must be converted from PersistDbSpecific"

instance PersistFieldSql Geo where
  sqlType _ = SqlOther "GEOGRAPHY(POINT,4326)"

toPoint :: Double -> Double -> Geo
toPoint lat lon = Geo $ Data.ByteString.concat ["'POINT(", ps $ lon, " ", ps $ lat, ")'"]
  where ps = Data.Text.pack . show

If Foo has a geography field, we can then perform insertions like the following:

insert $ Foo (toPoint 44 44)

data SqlType Source

A SQL data type. Naming attempts to reflect the underlying Haskell datatypes, eg SqlString instead of SqlVarchar. Different SQL databases may have different translations for these types.


SqlNumeric Word32 Word32 

Always uses UTC timezone

SqlOther Text

a backend-specific name

data Update record Source

updataing a database entity

Persistent users use combinators to create these


forall typ . PersistField typ => Update 
BackendUpdate (BackendSpecificUpdate (PersistEntityBackend record) record) 

type family BackendSpecificUpdate backend record Source

data SelectOpt record Source

query options

Persistent users use these directly


forall typ . Asc (EntityField record typ) 
forall typ . Desc (EntityField record typ) 
OffsetBy Int 
LimitTo Int 

data Filter record Source

Filters which are available for select, updateWhere and deleteWhere. Each filter constructor specifies the field being filtered on, the type of comparison applied (equals, not equals, etc) and the argument for the comparison.

Persistent users use combinators to create these


forall typ . PersistField typ => Filter 


filterField :: EntityField record typ
filterValue :: Either typ [typ]
filterFilter :: PersistFilter
FilterAnd [Filter record]

convenient for internal use, not needed for the API

FilterOr [Filter record] 
BackendFilter (BackendSpecificFilter (PersistEntityBackend record) record) 

type family BackendSpecificFilter backend record Source

data Entity record Source

Datatype that represents an entity, with both its Key and its Haskell record representation.

When using a SQL-based backend (such as SQLite or PostgreSQL), an Entity may take any number of columns depending on how many fields it has. In order to reconstruct your entity on the Haskell side, persistent needs all of your entity columns and in the right order. Note that you don't need to worry about this when using persistent's API since everything is handled correctly behind the scenes.

However, if you want to issue a raw SQL command that returns an Entity, then you have to be careful with the column order. While you could use SELECT Entity.* WHERE ... and that would work most of the time, there are times when the order of the columns on your database is different from the order that persistent expects (for example, if you add a new field in the middle of you entity definition and then use the migration code -- persistent will expect the column to be in the middle, but your DBMS will put it as the last column). So, instead of using a query like the one above, you may use rawSql (from the Database.Persist.GenericSql module) with its /entity selection placeholder/ (a double question mark ??). Using rawSql the query above must be written as SELECT ?? WHERE ... Then rawSql will replace ?? with the list of all columns that we need from your entity in the right order. If your query returns two entities (i.e. (Entity backend a, Entity backend b)), then you must you use SELECT ??, ?? WHERE ..., and so on.


PersistEntity record => Entity 


entityKey :: Key record
entityVal :: record


(PersistEntity record, Eq (Key record), Eq record) => Eq (Entity record) 
(PersistEntity record, Ord (Key record), Ord record) => Ord (Entity record) 
(PersistEntity record, Read (Key record), Read record) => Read (Entity record) 
(PersistEntity record, Show (Key record), Show record) => Show (Entity record) 
(PersistEntity record, PersistField record, PersistField (Key record)) => PersistField (Entity record) 
(PersistField record, PersistEntity record) => PersistFieldSql (Entity record) 
(PersistEntity a, (~) * (PersistEntityBackend a) SqlBackend) => RawSql (Entity a)