Often we need a pair of conversion functions: one to encode a value as a string, and another corresponding function to decode a string back into the original type. Here we show a concise way to define these functions for datatypes whose constructors enumerate a small fixed set of possibilities (sometimes described as “enums”).
We use the deriving strategies extension to be explicit about which deriving mechanism is in use for each typeclass.
This code uses
In addition to GHC’s
Generic class, we’ll also need to import the
GEnum class from the
generic-deriving library. When a type has an instance of the
GEnum class, we can use
genum to obtain a list of all values of that type.
We’ll be using the
Map data structure from the
First we’ll define two enum datatypes to use for examples. Let’s imagine we’re writing code related to billing for a subscription service. The product comes in three varieties – basic, standard, and pro – and our users can choose to pay either monthly or annually.
For each type, we derive instances of the
GEnum classes. We will not be directly using the
Generic class, but it is a requirement for deriving the
GEnum class which we will be using.
We’ll also define a
Bill datatype, consisting of a product and a billing frequency. These two fields represent a summary of a subscriber’s invoice.
Here are some example encoding functions.
For our first example, we define a way to represent values of the
Product type as
Next we define a way to encode the
Bill type as an
Integer. Since a bill consists of a
Product (of which there are three) and a
Frequency (of which there are two), then there are 3 × 2 = 6 values of
Bill to encode.
Decoding is converting back in the other direction – for example,
Integer -> Maybe Bill. The return type includes
Maybe because not every integer represents a bill, so decoding can fail.
It could be rather tedious and redundant to also write to corresponding decoding functions. Fortunately, we can take a shortcut.
If the type
a we’re encoding has an instance of
GEnum and we have an encoding function
f :: a -> b such as
encodeBill, then this
invert function will infer the corresponding decoding function.
The first step is to construct
Map that contains, for every value of type
a, a mapping from the encoded form back to the original value. This is where we use
genum to list all values of type
b -> Maybe a decoding function, then, consists of a lookup from that map.
Now we can use the
invert function to generate the decoding functions for
Bill with minimal effort.
We demonstrate by printing the results of some encodings and decodings.
putStrLn (encodeProduct Basic) putStrLn (encodeProduct Standard) putStrLn (show (decodeProduct "p1")) putStrLn (show (decodeProduct "xyz")) putStrLn (show (encodeBill (Bill Basic Annual))) putStrLn (show (encodeBill (Bill Pro Monthly))) putStrLn (show (decodeBill 31)) putStrLn (show (decodeBill 50))