A class for types with a default value.
An alternative to data-default package, that provides a lot of additional instances for Default
type class.
Default
type class provides a default value (def
) of a type. In case when a particluar type has an instance for Monoid
, then the default value should be mempty
, otherwise it can cause a great deal of confusion.
One must understand, that Default
type class, has no axioms associated with it. Its purpose is to minimize:
Number of imports.
Number of introduced definitions that are part of an API.
Complexity of building more complicated "default values" out of simpler "default values".
Cognitive overhead. (Reduction of cognitive overhead is partially a consequence of reducing above issues.)
As one may notice, most of the listed problems, that Default
type class tries to solve, to various degrees of success, are for human benefit, and not theoretically founded. Because of this, please always try hard to define sensible instances of Default
. Most importantly, document what def
means for particular type, and always check that def
is sensible, by testing it on a real world sample of human beings, one is not a big enough sample.
That said, using Default
may not always be a good idea. If it breaks people's mental model, or theoretical models with real axioms, then just don't use it.
data-default-extra
Description
An alternative to data-default package, but built on top of the same underlying package data-default-class. This package provides a lot of additional instances for Default
type class.
Default
type class provides a default value (def
) of a type. In case when a particluar type has an instance for Monoid
, then the default value should be mempty
, otherwise it can cause a great deal of confusion.
One must understand, that Default
type class, has no axioms associated with it. Its purpose is to minimize:
Number of imports.
Number of introduced definitions that are part of an API.
Complexity of building more complicated "default values" out of simpler "default values".
Cognitive overhead. (Reduction of cognitive overhead is partially a consequence of reducing above issues.)
As one may notice, most of the listed problems, that Default
type class tries to solve, to various degrees of success, are for human benefit, and not theoretically founded. Because of this, please always try hard to define sensible instances of Default
. Most importantly, document what def
means for particular type, and always check that def
is sensible, by testing it on a real world sample of human beings, one is not a big enough sample.
That said, using Default
may not always be a good idea. If it breaks people's mental models, or theoretical models with real axioms, then just don't use it.
This package, in most part, just reexports a set of packages that provide additional instances for data-default-class package.
GHC Generics
Create Default
instances using GHC Generics.
Usage example:
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DefaultSignatures #-}
import GHC.Generics (Generic)
import Data.Default (Default)
data MyType = MyType Int (Maybe String)
deriving (Generic, Show)
instance Default MyType
λ> def :: MyType
MyType 0 Nothing
One can also use DeriveAnyClass
language extension:
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DeriveAnyClass #-}
import GHC.Generics (Generic)
import Data.Default (Default)
data MyType = MyType Int (Maybe String)
deriving (Default, Generic, Show)
Be aware that DeriveAnyClass
doesn't work well together with GeneralizedNewtypeDeriving
on some versions of GHC. See GHC issue #10598
License
The BSD 3-Clause License, see LICENSE file for details.
Contributions
Contributions, pull requests and bug reports are welcome! Please don't be afraid to contact author using GitHub or by e-mail.