Copyright© 2019 Elias Castegren and Kiko Fernandez-Reyes
LicenseMIT
Stabilityexperimental
Portabilityportable
Safe HaskellNone

MultiError.Typechecker

Description

This module includes everything you need to get started type checking a program. To build the Abstract Syntax Tree (AST), please import and build the AST from MultiError.AST.

The main entry point to the type checker is the combinator tcProgram, which takes an AST and returns either a list of errors, or the typed program. For example, for the following program (using a made up syntax):

class C
  val f: Foo

should be parsed to generate this AST:

testClass1 =
 ClassDef {cname = "C"
          ,fields = [FieldDef {fmod = Val, fname = "f", ftype = ClassType "Foo"}]
          ,methods = []}

To type check the AST, run the tcProgram combinator as follows:

tcProgram testClass1

This is an increment on top of the Typechecker module, that refactors the type checker to be able to throw multiple errors.

Synopsis

Documentation

type TypecheckM a = forall m. (MonadReader Env m, MonadError TCErrors m) => m a Source #

The type checking monad. The type checking monad is the stacking of the Reader and Exception monads.

(<:>) :: Semigroup e => Either e a -> Either e b -> Either e (a, b) Source #

The function <:> takes two Either monads and returns an error if one of them is an error or aggregates both results. For example:

let error = Left "Error" <:> Right 42
let errors = Left "Error" <:> Left "Error2"
let valid = Right "42" <:> Right "0"

evaluates error = Left Error, errors = Left ErrorError2, and valid = 420.

(<&>) :: (Semigroup e, MonadError e m) => m a -> m b -> m (a, b) Source #

Forks two computations in the Except monad, and either returns both of their results, or aggregates the errors of one or both of the computations. For example:

    (fields', methods') <- forkM precheck fields <&>
                           forkM precheck methods

In this example, if the evaluation of forkM precheck fields and and forkM precheck methods return errors, we aggregate them using <:>. If only one of them fails, then return the single error. If both computation succeed, return a monad wrapped around the tuple with both results.

forkM :: (Semigroup e, MonadError e m) => (a -> m b) -> [a] -> m [b] Source #

Allows typechecking a list of items, collecting error messages from all of them.

newtype TCErrors Source #

Declaration of type checking errors. An error will (usually) be created using the helper function tcError. As an example:

tcError $ DuplicateClassError (Name "Foo")

throws an error that indicates that the class is defined multiple times.

Constructors

TCErrors (NonEmpty TCError) 
Instances
Show TCErrors Source # 
Instance details

Defined in MultiError.Typechecker

Methods

showsPrec :: Int -> TCErrors -> ShowS

show :: TCErrors -> String

showList :: [TCErrors] -> ShowS

Semigroup TCErrors Source # 
Instance details

Defined in MultiError.Typechecker

Methods

(<>) :: TCErrors -> TCErrors -> TCErrors

sconcat :: NonEmpty TCErrors -> TCErrors

stimes :: Integral b => b -> TCErrors -> TCErrors

data TCError Source #

Declaration of a type checking error, where Error represents the current type checking error and Backtrace the up-to-date backtrace.

Constructors

TCError

Type checking error value constructor

Fields

  • Error

    Current type checking error

  • Backtrace

    Backtrace of the type checker

Instances
Show TCError Source # 
Instance details

Defined in MultiError.Typechecker

Methods

showsPrec :: Int -> TCError -> ShowS

show :: TCError -> String

showList :: [TCError] -> ShowS

tcError :: Error -> TypecheckM a Source #

Throw a type checking Error

data Error Source #

Data declaration of available errors. Value constructors are used to create statically known errors. For example:

UnknownClassError  (Name c)

creates a UnknownClassError. This error should be created whenever there is a class whose declaration is unknown or inexistent.

Constructors

UnknownClassError Name

Reference of a class that does not exists

UnknownFieldError Name

Reference of a field that does not exists

UnknownMethodError Name

Reference of a method that does not exists

UnboundVariableError Name

Unbound variable

TypeMismatchError Type Type

Type mismatch error, the first Type refers to the formal type argument, the second Type refers to the actual type argument.

ImmutableFieldError Expr

Immutable field error, used when someone violates immutability

NonLValError Expr

Error to indicate that a one cannot assign a value to expression Expr

PrimitiveNullError Type

Error indicating that the return type cannot be Null

NonClassTypeError Type

Used to indicate that Type is not of a class type

NonArrowTypeError Type

Expecting a function (arrow) type but got another type instead.

ConstructorCallError Type

Tried to call a constructor outside of instantiation

UninferrableError Expr

Cannot infer type of Expr

Instances
Show Error Source # 
Instance details

Defined in MultiError.Typechecker

Methods

showsPrec :: Int -> Error -> ShowS

show :: Error -> String

showList :: [Error] -> ShowS

data Env Source #

Environment. The Env is used during type checking, and is updated as the type checker runs. Most likely, one uses the Reader monad to hide details of how the environment is updated, via the common local function.

Constructors

Env 

Fields

setConstructor :: Name -> Env -> Env Source #

Conditionally update the environment to track if we are in a constructor method.

emptyEnv :: Env Source #

Generates an empty environment.

lookupClass :: Name -> Env -> Maybe ClassDef Source #

Helper function to lookup a class given a Name and an Env. Usually it relies on the Reader monad, so that passing the Env can be omitted. For example:

findClass :: Type -> TypecheckM ClassDef
findClass ty@(ClassType c) = do
  cls <- asks $ lookupClass c
  case cls of
    Just cdef -> return cdef
    Nothing -> tcError $ UnknownClassError c
findClass ty = tcError $ NonClassTypeError ty

In this function (findClass), the Reader function asks injects the Reader monad as the last argument. More details in the paper.

lookupVar :: Name -> Env -> Maybe Type Source #

Look up a variable by its Name in the Env, returning an option type indicating whether the variable was found or not.

findClass :: Type -> TypecheckM ClassDef Source #

Find a class declaration by its Type

findMethod :: Type -> Name -> TypecheckM MethodDef Source #

Find a method declaration by its Type and method name m

findField :: Type -> Name -> TypecheckM FieldDef Source #

Find a field declaration by its Type (ty) and field name f

findVar :: Name -> TypecheckM Type Source #

Find a variable in the environment by its name x

genEnv :: Program -> Env Source #

Generates an environment (symbol's table) from a Program,

addVariable :: Name -> Type -> Env -> Env Source #

Add a variable name and its type to the environment Env.

addParameters :: [Param] -> Env -> Env Source #

Add a list of parameters, Param, to the environment.

tcProgram :: Program -> Either TCErrors Program Source #

Main entry point of the type checker. This function type checks an AST returning either a list of errors or a well-typed program. For instance, assuming the following made up language: > > class C > val f: Foo >

it should be parsed to generate the following AST:

testClass1 =
 ClassDef {cname = "C"
          ,fields = [FieldDef {fmod = Val, fname = "f", ftype = ClassType "Foo"}]
          ,methods = []}

To type check the AST, run the tcProgram combinator as follows:

tcProgram testClass1

which either returns a list of errors or the resulting typed AST.

class Typecheckable a where Source #

The type class defines how to type check an AST node.

Minimal complete definition

doTypecheck

Methods

doTypecheck :: a -> TypecheckM a Source #

Type check the well-formedness of an AST node.

typecheck :: Backtraceable a => a -> TypecheckM a Source #

Type check an AST node, updating the environment's backtrace.

hasType :: Expr -> Type -> TypecheckM Expr Source #

This combinator is used whenever a certain type is expected. This function is quite important. Here follows an example:

doTypecheck mdef@(MethodDef {mparams, mbody, mtype}) = do
  -- typecheck the well-formedness of types of method parameters
  mparams' <- mapM typecheck mparams
  mtype' <- typecheck mtype

  -- extend environment with method parameters and typecheck body
  mbody' <- local (addParameters mparams) $ hasType mbody mtype'
  ...

in the last line, because we are type checking a method declaration, it is statically known what should be the return type of the function body. In these cases, one should use the hasType combinator.

testClass1 :: ClassDef Source #

Class definition for didactic purposes. This AST represents the following class, which is named C, contains an immutable field f of type Foo:

class C:
  val f: Foo

This class is ill-typed, as there is no declaration of Foo anywhere. To check how to type checker catches this error, run:

tcProgram (Program [testClass1])

testClass2 :: ClassDef Source #

Test program with a class, field, method, and variable access. The class Bar does not exist in the environment. The variable access is unbound.

This program is the AST equivalent of the following syntax:

class D
  val g: Bar
  def m(): Int
    x

testSuite :: IO () Source #

Test suite that runs testProgram.