In Haskell 98 the name of a record field
is automatically also the name of a function which gets the value
of the according field.
E.g. if we have
data Pair a b = Pair {first :: a, second :: b}
then
first :: Pair a b -> a
second :: Pair a b -> b
However for setting or modifying a field value
we need to use some syntactic sugar, which is often clumsy.
modifyFirst :: (a -> a) -> (Pair a b -> Pair a b)
modifyFirst f r@(Pair {first=a}) = r{first = f a}
With this package you can define record field accessors
which allow setting, getting and modifying values easily.
The package clearly demonstrates the power of the functional approach:
You can combine accessors of a record and sub-records,
to make the access look like the fields of the sub-record belong to the main record.
Example:
*Data.Accessor.Example> (first^:second^=10) (('b',7),"hallo")
(('b',10),"hallo")
You can easily manipulate record fields in a Control.Monad.State.State monad,
you can easily code Show instances that use the Accessor syntax
and you can parse binary streams into records.
See Data.Accessor.Example for demonstration of all features.
It would be great if in revised Haskell versions the names of record fields
are automatically Data.Accessor.Accessors
rather than plain get functions.
For now, the package data-accessor-template provides Template Haskell functions
for automated generation of Data.Acesssor.Accessors.
See also the other data-accessor packages
that provide an Accessor interface to other data types.
The package enumset provides accessors to bit-packed records.
For similar packages see lenses and fclabel.
A related concept are editors
http://conal.net/blog/posts/semantic-editor-combinators/.
Editors only consist of a modify method
(and modify applied to a const function is a set function).
This way, they can modify all function values of a function at once,
whereas an accessor can only change a single function value,
say, it can change f 0 = 1 to f 0 = 2.
This way, editors can even change the type of a record or a function.
An Arrow instance can be defined for editors,
but for accessors only a Category instance is possible ((.) method).
The reason is the arr method of the Arrow class,
that conflicts with the two-way nature (set and get) of accessors.
