Version 13 (modified by igloo, 4 years ago) (diff) |
---|

# Proposal: ExplicitForall

Ticket | #133 |

Dependencies | |

Related | Rank2Types, RankNTypes, LiberalTypeSynonyms, PolymorphicComponents, ScopedTypeVariables |

## Compiler support

GHC | mostly (ExplicitForAll; still allows forall as an expression id; allows superfluous vars to be quantified) |

nhc98 | none |

Hugs | mostly (+98; still allows forall as an expression id; doesn't allow 0 tyvars to be quantified) |

UHC | mostly (none; still allows forall as an expression id; allows superfluous vars to be quantified; doesn't allow 0 tyvars to be quantified) |

JHC | none |

LHC | none |

## Description

ExplicitForAll enables the use of the keyword 'forall' to make a type explicitly polymorphic. Syntactically, it would mean the following change to Haskell 98:

- forall becomes a reserved word.
- . (dot) becomes a special (not reserved) operator.
- The following syntactic rule changes:

*type* → forall *tyvars* . *type*

|context=>type

|ftype

*ftype* → *btype* -> *type*

|btype

*gendecl* → *vars* :: *type*

It does not allow the use of explicitly polymorphic types in any way not already allowed by Haskell 98 for implicitly polymorphic types.

## Report Delta

[incomplete]

Changes relative to H2010 report.

In Section 2.4:

Replace:

```
| foreign | if | import | in | infix | infixl
```

with:

```
| forall | foreign | if | import | in | infix | infixl
```

In Section 3:

Replace:

```
exp → infixexp :: [context =>] type (expression type signature)
```

with:

```
exp → infixexp :: [forall tyvar1 … tyvark .] [context =>] type (expression type signature, k ≥ 0)
```

In Section 3.16:

Replace:

exp → exp :: [context =>] type Expression type-signatures have the form e :: t, where e is an expression and t is a type (Section 4.1.2); they are used to type an expression explicitly and may be used to resolve ambiguous typings due to overloading (see Section 4.3.4). The value of the expression is just that of exp. As with normal type signatures (see Section 4.4.1), the declared type may be more specific than the principal type derivable from exp, but it is an error to give a type that is more general than, or not comparable to, the principal type.

with:

exp → exp :: [forall tyvar1 … tyvark .] [context =>] type (k ≥ 0) Expression type-signatures have the form e :: t, where e is an expression and t is a type (Section 4.1.2); they are used to type an expression explicitly and may be used to resolve ambiguous typings due to overloading (see Section 4.3.4). The value of the expression is just that of exp. As with normal type signatures (see Section 4.4.1), the declared type may be more specific than the principal type derivable from exp, but it is an error to give a type that is more general than, or not comparable to, the principal type. The quantification [forall tyvar1 … tyvark .] is optional, but if given it must quantify exactly the set of type variables used in [context =>] type (with one exception: that of the distinguished type variable in a class declaration (Section 4.3.1), which must not be quantified).

In Section 4:

Replace:

```
gendecl → vars :: [context =>] type (type signature)
```

with:

```
gendecl → vars :: [forall tyvar1 … tyvark .] [context =>] type (type signature, k ≥ 0)
```

In Section 4.1.2:

Replace:

With one exception (that of the distinguished type variable in a class declaration (Section 4.3.1)), the type variables in a Haskell type expression are all assumed to be universally quantified; there is no explicit syntax for universal quantification [4]. For example, the type expression forall a . a -> a denotes the type ∀ a . a → a. For clarity, however, we often write quantification explicitly when discussing the types of Haskell programs. When we write an explicitly quantified type, the scope of the extends as far to the right as possible; for example, ∀ a . a → a means ∀ a . (a → a).

with:

The type variables in a type signature may be explicitly quantified with the forall keyword. All type variables used in the type must be quantified, with one exception (that of the distinguished type variable in a class declaration (Section 4.3.1), which must not be quantified). No additional type variables may be quantified. The scope of the quantifier extends as far to the right as possible. For example, the type expression a -> a denotes the type ∀ a . a → a. Quantifying the type variables is optional. For example, the type expression a -> a also denotes the type ∀ a . a → a. For convenience, we write forall vs . t even if there is no forall vs ..

In Section 4.1.4:

Replace:

```
(Eq a, Show a, Eq b) => [a] -> [b] -> String
```

with:

```
forall a b . (Eq a, Show a, Eq b) => [a] -> [b] -> String
```

In Section 4.3.1:

Replace:

for which there is an explicit type signature vi :: cxi => ti in cdecls.

with:

for which there is an explicit type signature vi :: forall vs . cxi => ti in cdecls.

In Section 4.4.4:

Replace:

```
gendecl → vars :: [context =>] type
```

with:

```
gendecl → vars :: [forall tyvar1 … tyvark .] [context =>] type (k ≥ 0)
```

Replace:

A type signature specifies types for variables, possibly with respect to a context. A type signature has the form: v1, …, vn :: cx => t which is equivalent to asserting vi :: cx => t for each i from 1 to n.

with:

A type signature specifies types for variables, possibly with respect to a context. A type signature has the form: v1, …, vn :: forall vs . cx => t which is equivalent to asserting vi :: cx => t for each i from 1 to n.

## References

http://www.haskell.org/pipermail/haskell-prime/2009-June/002786.html

## Pros

- Small and simple syntactic extension.
- Simplifies the later inclusion of semantic extensions that depend on it, e.g. Rank2Types.
- Easy to implement in tools that don't yet support the semantic extensions.
- The Report already mentions types using the explicit forall-quantified form, so only the grammar changes above are needed.

## Cons

- A small and incremental extension with little value of its own, only serving as a stepping stone for the various semantic extensions.