A tiny EDSL to write type-level-unit tests.
A small example:
import Test.TypeSpec
main :: IO ()
main = print spec0
spec :: Expect "Expect something..." (Int `Isn't` Bool)
spec = Valid
This will output:
Valid:
Expect something...
(✓ Different)
Using the operators from TypeSpecCrazy:
specCrazy ::
"Higher kinded assertions"
###########################
"ShouldBe accepts types of kind * -> *"
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ShouldBe Maybe Maybe
-* ShouldBe [] []
-* ShouldBe (->) (->)
specCrazy = Valid
main = print specCrazy
The output:
Valid:
Higher kinded assertions
ShouldBe accepts types of kind * -> *
(✓ Equal)
(✓ Equal)
(✓ Equal)
If you like Lisp, this might be for you:
type ALot = 1000
specAliases ::
(Explain "There are a variety aliases for the basic combinators."
(Context "Basic Combinators"
(Describe "Context"
(It "labels expectations using 'It'"
(Describe "Describe"
(It's "an alias for It, just like They"
(It's "time for the first assertion"
(1000 `Is` ALot))))))))
specAliases = Valid
main = print specAliases
This will output:
Valid:
There are a variety aliases for the basic combinators.
Basic Combinators
Context
labels expectations using 'It'
Describe
an alias for It, just like They
time for the first assertion
(✓ Equal)
The key feature is that the compiler checks the assertions and expectations made
in a ‘TypeSpec’ and right away rejects invalid types.
When compiling this example:
specFailing ::
TypeSpec
(It "counts the number of elements in a tuple"
(Count ((),(),()) `ShouldBe` 4))
specFailing = Valid
type family Count a :: Nat where
Count (a,b) = 2
Count (a,b,c) = 3
Count (a,b,c,d) = 4
The compiler complains:
error:
• counts the number of elements in a tuple
Expected type: 3
Actual type: 4
• In the expression: Valid
In an equation for ‘specFailing’: specFailing = Valid
After all, with TypeError
GHC is quite a test-runner.
If you accept to defer type checking and have invalid specs checked during test execution, use
(should-not-typecheck)[https://github.com/CRogers/should-not-typecheck].