Hspec: A Testing Framework for Haskell

Writing Tests with Hspec

Using `describe` and `it`

Hspec provides an EDSL to describe tests.

describe and it are used to organize tests with Hspec.

main :: IO ()
main = hspec $ do
  describe "Prelude.read" $ do
    it "can parse integers" $ do
      read "10" `shouldBe` (10 :: Int)

    it "can parse floating-point numbers" $ do
      read "2.5" `shouldBe` (2.5 :: Float)

A spec can consist of multiple describes, and describes can be nested.

main :: IO ()
main = hspec $ do
  describe "Prelude" $ do
    describe "read" $ do
      it "can parse integers" $ do
        read "10" `shouldBe` (10 :: Int)

    describe "head" $ do
      it "returns the first element of a list" $ do
        head [23 ..] `shouldBe` 23

Using `context`

context is just an alias for describe. Use context with when or with, e.g.:

main :: IO ()
main = hspec $ do
  describe "parse" $ do
    context "when provided with invalid input" $ do
      it "returns a parse error" $ do
        parse "some invalid input" `shouldBe` Left "parse error"

Using hooks

before_ runs a custom IO action before every spec item. For example, if you have an action flushDb which flushes your database, you can run it before every spec item with:

main :: IO ()
main = hspec $ before_ flushDb $ do
  describe "/api/users/count" $ do
    it "returns the number of users" $ do
      post "/api/users/create" "name=Jay"
      get "/api/users/count" `shouldReturn` 1

    context "when there are no users" $ do
      it "returns 0" $ do
        get "/api/users/count" `shouldReturn` 0

Similarly, after_ runs a custom IO action after every spec item:

main :: IO ()
main = hspec $ after_ truncateDatabase $ do
  describe "createUser" $ do
    it "creates a new user" $ do
      let eva = User (UserId 3) (Name "Eva") (Age 28)
      createUser eva
      getUser (UserId 3) `shouldReturn` eva

  describe "countUsers" $ do
    it "counts all registered users" $ do
      countUsers `shouldReturn` 0

around_ is passed an IO action for each spec item so that it can perform whatever setup and teardown is necessary.

serveStubbedApi :: String -> Int -> IO Server
stopServer :: Server -> IO ()

withStubbedApi :: IO () -> IO ()
withStubbedApi action =
  bracket (serveStubbedApi "localhost" 80)
          stopServer
          (const action)

main :: IO ()
main = hspec $ around_ withStubbedApi $ do
  describe "api client" $ do
    it "should authenticate" $ do
      c <- newClient (Just ("user", "pass"))
      get c "/api/auth" `shouldReturn` status200

    it "should allow anonymous access" $ do
      c <- newClient Nothing
      get c "/api/dogs" `shouldReturn` status200

Hooks support passing values to spec items (for example, if you wanted to open a database connection before each item and pass the connection in). This can be done with before, around and after. Here's an example for how to use around:

openConnection :: IO Connection
openConnection = ...

closeConnection :: Connection -> IO ()
closeConnection = ...

withDatabaseConnection :: (Connection -> IO ()) -> IO ()
withDatabaseConnection = bracket openConnection closeConnection

spec :: Spec
spec = do
  around withDatabaseConnection $ do
    describe "createRecipe" $ do
      it "creates a new recipe" $ \c -> do
        let ingredients = [Eggs, Butter, Flour, Sugar]
        createRecipe c (Recipe "Cake" ingredients)
        getRecipe c "Cake" `shouldReturn` ingredients

Using `pending` and `pendingWith`

pending/pendingWith can be used as a kind of TODO list while you are writing your specs. If you think of some behaviour that your code should satisfy, but you are working on something else, you can add a spec item and just set it to pending.

main :: IO ()
main = hspec $ do
  describe "/login" $ do
    it "should use correct status codes" $ do
      pending

    it "should require basic authentication" $ do
      pendingWith "need to fix base64 first"

It's advisable to resolve all pending spec items before committing your changes.

The report produced at the end of a test run counts passed, failed, and pending spec items separately, and having unresolved pending spec items does not cause your test suite to fail; that is, it can still exit with a status code of 0 if there are unresolved pending spec items.