Prefix form for MkSolo#

ghc-proposals · Feb 21, 2024 · 646e1fb · 646e1fb
1 parent 6edec29
commit 646e1fb
Showing 1 changed file with 12 additions and 1 deletion.
diff --git a/proposals/0475-tuple-syntax.rst b/proposals/0475-tuple-syntax.rst
@@ -142,7 +142,7 @@ Proposed Change Specification
      data Unit# = (# #)
 
      type Solo# :: TYPE rep -> TYPE (TupleRep [rep])
-     data Solo# a = (# a #)
+     data Solo# a = MkSolo# a
 
      type Tuple0# = Unit#
      type Tuple1# = Solo#
@@ -275,6 +275,8 @@ Proposed Change Specification
 
    #. With ``-XUnboxedTuples``, an occurrence of ``(# ty #)`` in type-syntax is a synonym for ``GHC.Types.Solo# ty``.
 
+   #. With ``-XUnboxedTuples``, an occurrence of ``(# x #)`` in term-syntax is a synonym for ``GHC.Types.MkSolo# x``.
+
    #. With ``-XUnboxedTuples``, an occurrence of ``(#,,...,,#)`` where there are *n* commas (for *n* ≧ 1) in type-syntax
       is a synonym for ``GHC.Types.Tuple<n+1>#``.
 
@@ -331,6 +333,10 @@ Proposed Change Specification
       never types or type constructors. (Note that ``(...) =>`` is special syntax, not an occurrence of any of the types
       listed above. See `below <#constraints-special-syntax>`_.)
 
+   #. An occurrence of ``(# x #)`` is a synonym for ``GHC.Types.MkSolo# x`` in both term- and type-syntax.
+      (It is not valid in types at the time of writing, as GHC is unable to promote unboxed tuples, but this concern
+       is separate from syntax.)
+
    #. A use of ``(# ... | ... | ... #)``, where each of the ``...`` is filled in, (among other arities) is now disallowed.
 
    #. An occurrence of ``GHC.Tuple.Tuple<n> ty1 ty2 ... ty<n>`` is pretty-printed as ``Tuple (ty1, ty2, ..., ty<n>)``.
@@ -463,6 +469,11 @@ Effect and Interactions
    means either ``Tuple2# Int Bool`` (with ``-XNoListTuplePuns``) or ``Solo# (Tuple2 Int Bool)`` (with
    ``-XListTuplePuns``). Furthermore, ``Tuple# (Int#, Double#)`` would not kind-check with ``-XListTuplePuns``. So we avoid these problems by simply erroring in the 1-element case.
 
+#. The name of the data constructor for unboxed 1-tuples is ``MkSolo#`` rather
+   than ``(# #)`` to distinguish it from the data constructor for 0-tuples.
+   The ambiguity only arises when the name is used unapplied.
+   When applied to an argument, ``MkSolo# a`` can be written and pretty-printed as ``(# a #)``.
+
 Costs and Drawbacks
 -------------------
 1. This is one more feature to maintain, but the code would be pretty local.