[wg-camlp4] Against the use of syntactically-valid OCaml code for syntax extension purposes

[Leo White]

I would just like to say that I mostly agree with Gabriel. Where possible I 
think we should try to choose syntaxes that are obviously syntax 
extensions.

For example, I don't think that:

let rec expr =
  <:parser <
      If; x = expr; Then; y = expr; Else; z = expr -> "if"
    | Let; Ident x; Equal; x = expr; In; y = expr -> "let" >>

is really worse than

let rec expr =
  parser
    [< 'If; x = expr; 'Then; y = expr; 'Else; z = expr >] -> "if"
  | [< 'Let; 'Ident x; 'Equal; x = expr; 'In; y = expr >] -> "let"

and it makes it very clear that it is using an extension.

In some cases it seems sensible to use an obviously wrong syntax (e.g. 
!"asdf"), but in general I think we should try to use new syntax for 
extensions.

In particular I think that the following forms can cover many cases:

* Camlp4-style quotations: <:foo < arg >> 

* Some kind of quotation that looks like a type constructor and is quietly 
  ignored by the main compiler. I personally like the syntax "%foo" and 
  "%bar(arg)".

* Type-conv style annotations: type t = int with foo, bar(arg) 

* Allowing let and maybe some other keywords to be suffixed by some symbols 
  (e.g. let*)

* Allowing top-level style directives in ml files (e.g. "#ifdef ;;").