Allow record wildcards for empty records ¶

Allow record wild cards for empty records. This removes and artificial restriction, making the language simpler and more predictable.

Motivation ¶

Currently this is allowed:

data Foo = Foo { a :: Int }

f (Foo {..}) = 1

x = Foo {..}
  where a = 1

But this is not:

data Foo = Foo { }

f (Foo {..}) = 1

x = Foo {..}

Prohibiting this doesn’t seem good:

It feels arbitrary and artificial. Does anyone expect this behavior?
It is a pitfall for generated code, which cannot uniformly use record wildcard syntax no matter how many fields there are.

Furthermore this is allowed

data Foo = Foo { a :: Int }

f (Foo { a = 1, ..}) = 1

x = Foo { a = 1, ..}

So we see .. binding or using no fields isn’t even consistently prohibited!

Proposed Change Specification ¶

Allow record wildcard syntax for empty variants. There are no (named) fields, so no variables are bound in patterns, and no variables are used in expressions.

Examples ¶

The second example in the motivation is allowed.

This is also allowed:

data Foo = Foo -- no {}

f (Foo {..}) = 1

x = Foo {..}

for it doesn’t matter today whether empty variants are declared with {}, and this should remain true.

Effect and Interactions ¶

Foo {..} has the same meaning as Foo {} for an empty record. Note that means that:

data Foo = Foo
f (Foo {..}) = 42

is newly accepted just as:

data Foo = Foo
f (Foo {}) = 42

is today

Costs and Drawbacks ¶

Cannot think of any.

Alternatives ¶

Do nothing. Can’t think of anything else.

Unresolved Questions ¶

None at this time.

Implementation Plan ¶

This should be very easy. Perhaps we should use it as a mentoring exercise for new contributors.

Endorsements ¶

There was positive feedback in https://github.com/ghc-proposals/ghc-proposals/issues/484 where this was previously brought up.