generic-arbitrary
Generic implementation for QuickCheck's Arbitrary
http://github.com/typeable/generic-arbitrary#readme
LTS Haskell 22.36: | 1.0.1 |
Stackage Nightly 2024-10-03: | 1.0.1 |
Latest on Hackage: | 1.0.1 |
generic-arbitrary-1.0.1@sha256:677fca5829b5691e3527bcfe8499300c393ec5d10ec2d4cffe8e63a1ca647c0b,3459
Module documentation for 1.0.1
- Test
- Test.QuickCheck
- Test.QuickCheck.Arbitrary
- Test.QuickCheck
generic-arbitrary
What?
Package for deriving Arbitrary
via Generic
.
import GHC.Generics (Generic)
import Test.QuickCheck
import Test.QuickCheck.Arbitrary.Generic
data Expr
= Lit Int
| Add Expr Expr
| Mul Expr Expr
deriving (Eq, Show, Generic)
deriving Arbitrary via (GenericArbitrary Expr)
Older versions of this package had a problem with hanging arbitrary
method. Since 1.0.0
this problem almost solved.
For QuickCheck
older than 2.14.0
the GenericArbitrary
is not available, so
you will need to write instances more verbosely
data Expr
= Lit Int
| Add Expr Expr
| Mul Expr Expr
deriving (Eq, Show, Generic)
instance Arbitrary Expr where
arbitrary = genericArbitrary
shrink = genericShrink
Which is generally the same.
Infinite terms problem
The generic-arbitrary
can partially handle the problem with recursive
types. Assume the type R
data R = R R
deriving Generic
there is no instance
instance Arbitrary R where
arbitrary = genericArbitrary
shrink = genericShrink
If you try to compile this you will get a type level error
• R refers to itself in all constructors
Which means that there is no finite term for R
because it is recursive in all
it’s constructors. But, if you correct the definition of R
like this.
data R = R R | F
deriving Generic
Then it will compile. And the arbitrary
generated will not hang forever,
because it respects the size
parameter.
Limitation
There is a limitation of recursion detection:
data R1 = R1 R2
deriving (Eq, Ord, Show, Generic)
deriving anyclass NFData
deriving Arbitrary via (GenericArbitrary R1)
data R2 = R2 R1
deriving (Eq, Ord, Show, Generic)
deriving anyclass NFData
deriving Arbitrary via (GenericArbitrary R2)
This code will compile and the arbitrary
generated will always hang. Yes,
there is a problem with mutually recursive types.
Type parameters
Now let’s see an example of datatype with parameters
data A a = A a
deriving (Eq, Ord, Show)
deriving anyclass NFData
deriving (Generic)
instance (Arbitrary a) => Arbitrary (A a) where
arbitrary = genericArbitrary
shrink = genericShrink
It should work from first glance, but when compile it will throw an error:
• Could not deduce (Test.QuickCheck.Arbitrary.Generic.GArbitrary
(A a)
(GHC.Generics.D1
('GHC.Generics.MetaData "A" "ParametersTest" "main" 'False)
(GHC.Generics.C1
('GHC.Generics.MetaCons "A" 'GHC.Generics.PrefixI 'False)
(GHC.Generics.S1
('GHC.Generics.MetaSel
'Nothing
'GHC.Generics.NoSourceUnpackedness
'GHC.Generics.NoSourceStrictness
'GHC.Generics.DecidedLazy)
(GHC.Generics.Rec0 a))))
(TypesDiffer (A a) a))
arising from a use of ‘genericArbitrary’
Here the TypesDiffer
is a type familty dealing with recursive types and
helping us to eliminate inproper instances. To convince the compiller, that the
a
parameter is not an A a
we must fix the instance with additional
constraint Arg (A a) a
instance (Arg (A a) a, Arbitrary a) => Arbitrary (A a) where
arbitrary = genericArbitrary
shrink = genericShrink
Now everything compiles and works as expected.
Changes
1.0.1
- Compillability by GHC-9.2.4
- Fix cabal
tested-with
field
1.0.0
- Fixed issue with too big terms in case of recursive types
- Recpect the
size
argument - Types with parameters require
Arg
now
0.2.2
- Resolved an issue where the size of the generators could become negative
0.2.1
- Added compatibility with GHC 9.2.1
0.2.0
- Added
GenericArbitrary
for use with the DerivingVia compiler extension. - Minimum bound on QuickCheck changed to 2.14.
0.1.0
- Public release