SynType with single slash and type

[nojaf]

Consider the following code and AST:

[<Measure>] type herth = / second
// Same AST
[<Measure>] type herth = 1 / second

AST:

Types
   ([SynTypeDefn
       (SynComponentInfo
          ([{ Attributes = [ ... ], None, [], [herth],
           PreXmlDoc ((1,0), FSharp.Compiler.Xml.XmlDocCollector),
           false, None, tmp.fsx (1,17--1,22)),
        Simple
          (TypeAbbrev
             (Ok,
              Tuple
                (false,
                 [(true,
                   StaticConstant (Int32 1, tmp.fsx (1,25--1,33)));
                  (false,
                   LongIdent (SynLongIdent ([second], [], [None])))], ...]
Types
   ([SynTypeDefn
       (SynComponentInfo
          ([{ Attributes = [ ... ]
              Range = tmp.fsx (3,0--3,11) }], None, [], [herth],
           PreXmlDoc ((3,0), FSharp.Compiler.Xml.XmlDocCollector),
           false, None, tmp.fsx (3,17--3,22)),
        Simple
          (TypeAbbrev
             (Ok,
              Tuple
                (false,
                 [(true,
                   StaticConstant (Int32 1, tmp.fsx (3,25--3,26)));
                  (false,
                   LongIdent (SynLongIdent ([second], [], [None])))],... ], tmp.fsx (3,0--3,35))

Expected behaviour

The AST should be different, there is no actual 1 in / second.

Actual behaviour

I noticed a shortcut in:

fsharp/src/Compiler/pars.fsy

Lines 5253 to 5255 in 4be3bb5

	| INFIX_STAR_DIV_MOD_OP tupleOrQuotTypeElements
	{ if $1 <> "/" then reportParseErrorAt (rhs parseState 1) (FSComp.SR.parsUnexpectedInfixOperator());
	SynType.Tuple(false, (true, SynType.StaticConstant (SynConst.Int32 1, lhs parseState)) :: $2, lhs parseState) }

Perhaps we need a new case for this in SynType.
I'm not quite sure if SynType.Tuple is quite correct to capture a / b in the first place.
How should this be modelled? Any Ideas @dsyme and @auduchinok?

//cc @edgarfgp

[auduchinok]

I'm not quite sure if SynType.Tuple is quite correct to capture a / b in the first place.

Yes, the tuple doesn't sound like the appropriate case here.

How should this be modelled? Any Ideas @dsyme and @auduchinok?

Are there any other type math cases that we could take into account?