[RFC FS-1149 Discussion] - Support CallerArgumentExpression

[ijklam]

This is a discussion about RFC-1149 Supporting CallerArgumentExpression

[ijklam]

What about the computation expression situations? Is this okay?

open System.Runtime.CompilerServices

type Builder() =
    member self.Bind(
        x, f, 
        [<CallerArgumentExpression "x">] ?exp : string, 
        [<CallerArgumentExpression "f">] ?exp2 : string) =
        (f x, $"f={exp2.Value}, x={exp.Value}")

    member self.Return(x, [<CallerArgumentExpression "x">] ?exp : string) =
        (x, $"x={exp.Value}")

let b = Builder()

b { do! () } // (((), "x=do!"), "f=do!, x=()")
b { let! a = 123 in return a } // ((123, "x=a"), "f=return a, x=123")

b {
    let! a = 123
    let! b = 456
    return a + b
} (*
  (((579, "x=a + b"), "f=return a + b, x=456"),
   "f=let! b = 456
    return a + b, x=123")
*)

[T-Gro]

This looks sensible.

[ijklam]

Should we retain the original newline mark or make it all \n?

	A.assertEqual("\"abc\"\r\n#line 1\r\n  : string", Operators.DefaultArg(FSharpOption<string>.Some("\"abc\"\n#line 1\n  : string"), "no value"));
	A.assertEqual("(+) 1\r\n#line 1\r\n        123", Operators.DefaultArg(FSharpOption<string>.Some("(+) 1\n#line 1\n        123"), "no value"));

[T-Gro]

Is there a need to split the contents for their lines for some other processing? If not, I would retain the original contents as they are.

[ijklam]

okay