Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
Naked type aliases can add and name constraints
// This F# language suggestion wants a way to name collections of constraints:
// http://fslang.uservoice.com/forums/245727-f-language/suggestions/8509687-add-constraints-as-a-language-construct
//
// This is a type alias X<T> = T, so X<int> = int etc.
type X<'T> = 'T
// This is a type alias X<T> = T which adds a constraint
type WithStruct<'T when 'T : struct> = 'T
// Use it like this:
let f1 (x: WithStruct<'T>) : 'T = x
// This shows a type alias can imply multiple named constraintsm which gives a way of naming collections of constraints:
type WithFG< ^T when ^T : (member F : int -> unit)
and ^T : (member G : int -> unit) > = ^T
// Use it like this:
let inline f2 (x: WithFG< ^T >) =
(^T : (member F : int -> unit) (x, 1))
(^T : (member G : int -> unit) (x, 2))
@smoothdeveloper

This comment has been minimized.

Copy link

@smoothdeveloper smoothdeveloper commented Feb 28, 2016

they can name, but don't seem to allow to use without restating the constraint:

type HasName< ^T when ^T : (member Name : unit -> string)> = ^T

type Foo(n:string) =
  member x.Name = n

let inline name (named: HasName< 't >) = named.Name
//                                       ^^^  Lookup on object of indeterminate type based on information prior to this program point. A type annotation may be needed prior to this program point to constrain the type of the object. This may allow the lookup to be resolved.

So I'm still looking for how it is helping?

@cloudRoutine

This comment has been minimized.

Copy link

@cloudRoutine cloudRoutine commented Feb 28, 2016

What's the advantage of using "constraint aliases" over

let inline fnF x n = (^T : (member F : int -> unit) (x, n)) 
let inline fnG x n = (^T : (member G : int -> unit) (x, n))

let inline f3 x =
    fnF x 1
    fnG x 2 

every time (x: WithFG< ^T >) is used in an argument the function body will be polluted with (^T : (member F : int -> unit) (x, 1)) to take advantage of the SRTP,

@blumu

This comment has been minimized.

Copy link

@blumu blumu commented May 2, 2016

Thanks Don for sharing the Gist but unfortunately the type alias trick does not solve the problem for purely static member constraints. In the following example for instance I want to express the same set of constraints for both ^T1 and ^T2 without having to repeat them twice.

type Combine< ^T1, ^T2 when 
    ^T1 : (static member print : string -> unit) 
and ^T1 : (static member flush : unit -> unit) 
and ^T2 : (static member print : string -> unit) 
and ^T2 : (static member flush : unit -> unit) 
> = 
static member inline printAndFlushBoth m = 
  (^T1:(static member print : string -> unit) m) 
  (^T2:(static member print : string -> unit) m) 
  (^T1:(static member flush : unit -> unit) ()) 
  (^T2:(static member flush : unit -> unit) ())

Following your trick I could define a type alias encoding the two constraints (write and flush) but my constraints being entirely static I have no way to refer to the type alias it in a type annotation since there exist no variable of static type.

Do you have another trick up your sleeve for the above example?
Another benefit of named constraints is that the 'comparison' and 'equality' constraints could be defined in the F# core library instead of being hard-coded in the language.

An alternative suggestion would be to add support for generic type constraints of the form ^T :> GenericType<_>. Together with the above trick that would allow me to write:

type Printer< ^T when 
    ^T : (static member print : string -> unit) 
and ^T : (static member flush : unit -> unit) > = ^T

type Combine< ^T1, ^T2 when ^T1 :> Printer<_> and ^T2 :> Printer<_> > = 
  static member inline printAndFlushBoth m = 
    (^T1:(static member print : string -> unit) m) 
    (^T2:(static member print : string -> unit) m) 
    (^T1:(static member flush : unit -> unit) ()) 
    (^T2:(static member flush : unit -> unit) ())

which currently gives me an error:

error FS0698: Invalid constraint: the type used for the constraint is sealed, which means the constraint could only be satisfied by at most one solution
@Luiz-Monad

This comment has been minimized.

Copy link

@Luiz-Monad Luiz-Monad commented Mar 22, 2017

This would greatly help me.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment