I thought that conversions between F# functions and System.Func had to be done manually, but there appears to be a case where the compiler (sometimes) does it for you. And when it goes wrong the error message isn't accurate:
module Foo =
let dict = new System.Collections.Generic.Dictionary<string, System.Func<obj,obj>>()
let f (x:obj) = x
do
// Question 1: why does this compile without explicit type conversion?
dict.["foo"] <- fun (x:obj) -> x
// Question 2: given that the above line compiles, why does this fail?
dict.["bar"] <- f
The last line fails to compile, and the error is:
This expression was expected to have type
System.Func<obj,obj>
but here has type
'a -> obj
Clearly the function f doesn't have a signature of 'a > obj. If the F# 3.1 compiler is happy with the first dictionary assignment, then why not the second?
The part of the spec that should explain this is 8.13.7 Type Directed Conversions at Member Invocations. In short, when invoking a member, an automatic conversion from an F# function to a delegate will be applied. Unfortunately, the spec is a bit unclear; from the wording it seems that this conversion might apply to any function expression, but in practice it only appears to apply to anonymous function expressions.
The spec is also a bit out of date; in F# 3.0 type directed conversions also enable a conversion to a System.Linq.Expressions.Expression<SomeDelegateType>.
EDIT
In looking at some past correspondence with the F# team, I think I've tracked down how a conversion could get applied to a non-syntactic function expression. I'll include it here for completeness, but it's a bit of a strange corner case, so for most purposes you should probably consider the rule to be that only syntactic functions will have the type directed conversion applied.
The exception is that overload resolution can result in converting an arbitrary expression of function type; this is partly explained by section 14.4 Method Application Resolution, although it's pretty dense and still not entirely clear. Basically, the argument expressions are only elaborated when there are multiple overloads; when there's just a single candidate method, the argument types are asserted against the unelaborated arguments (note: it's not obvious that this should actually matter in terms of whether the conversion is applicable, but it does matter empirically). Here's an example demonstrating this exception:
type T =
static member M(i:int) = "first overload"
static member M(f:System.Func<int,int>) = "second overload"
let f i = i + 1
T.M f |> printfn "%s"
EDIT: This answer explains only the mysterious promotion to 'a -> obj. #kvb points out that replacing obj with int in OPs example still doesn't work, so that promotion is in itself insufficient explanation for the observed behaviour.
To increase flexibility, the F# type elaborator may under certain conditions promote a named function from f : SomeType -> OtherType to f<'a where 'a :> SomeType> : 'a -> OtherType. This is to reduce the need for upcasts. (See spec. 14.4.2.)
Question 2 first:
dict["bar"] <- f (* Why does this fail? *)
Because f is a "named function", its type is promoted from f : obj -> obj following sec. 14.4.2 to the seemingly less restrictive f<'a where 'a :> obj> : 'a -> obj. But this type is incompatible with System.Func<obj, obj>.
Question 1:
dict["foo"] <- fun (x:obj) -> x (* Why doesn't this, then? *)
This is fine because the anonymous function is not named, and so sec. 14.4.2 does not apply. The type is never promoted from obj -> obj and so fits.
We can observe the interpreter exhibit behaviour following 14.4.2:
> let f = id : obj -> obj
val f : (obj -> obj) (* Ok, f has type obj -> obj *)
> f
val it : ('a -> obj) = <fun:it#135-31> (* f promoted when used. *)
(The interpreter doesn't output constraints to obj.)
Related
Is there a way to do string formatting for Exceptions with kprintf (or similar) and still get warnings about redundant arguments? So far I have:
type MyException (s:string) =
inherit System.Exception(s)
static member Raise msg =
Printf.kprintf (fun s -> raise (MyException(s))) msg
do
MyException.Raise "boom %d" 9 1 // Gives NO warning about redundant arguments
failwithf "boom %d" 9 1 // Gives warning about redundant arguments
I know I could use the new $ formatting but would like to stay compatible with older versions of F#.
The problem does not come from kprintf, but from raise, which allows the return type of the function to be any type, including a curried function. The function failwithf has this warning as a special case.
If you return, let’s say, a string or a unit, you would actually get an error. The downside is that now you cannot raise in every position anymore, as then the return argument is not generic anymore.
You could fix that by forcing a type argument, but that would make it less than ideal for the caller.
For instance, this works, but if you change the return type of f to be a string, it fails:
open System
type MyException (s:string) =
inherit System.Exception(s)
static member Raise msg =
Printf.kprintf (fun s -> raise (MyException s) |> ignore) msg
module Test =
let f() : unit = // string won’t work
MyException.Raise "boom %d" 9 10 // error, as expected
Edit: workaround
Here's some kind of a workaround. You can use generic type constraints to limit the type, and use the fact that an F# function type is not comparable or equatable, does not have a proper value null.
Most types you'll use are equatable, because of F#'s powerful structural equality support. Other options for a type constraint are struct (only structs), null (this excludes record and DU types) or a certain inheritance constraint if that serves your use case.
Example:
type MyException (s:string) =
inherit System.Exception(s)
static member Raise<'T when 'T : equality> msg =
let raiser s =
raise (MyException s)
Unchecked.defaultof<'T> // fool the compiler
Printf.kprintf raiser msg
module Test =
let f() = MyException.Raise "boom %d" 9
let g() = MyException.Raise "boom %d" 9 10 // error
Downside, apart from the type restriction, is that the error you get will be a type error, not a warning as you requested.
The check for this warning in the compiler specifically looks for a number of known F# library functions. You can see the checks in the compiler source code and here (the functions are raise, failwith, failwithf, nullArg, invalidOp and invalidArg).
The reason for this is that, there is, in principle, nothing wrong with a generic function that returns 'T and is used so that 'T is inferred to be a function. For example, it is perfectly fine to do:
let id f = f
id sin 3.14
Here, we define a function id with one argument, but if I give it a function as an argument, it also becomes valid to call it with an extra argument as id sin 3.14.
The special functions are special in that they look like they return a value of a generic type, but they never actually return, because they raise an execption. The compiler cannot, in general, know whether that is the case for any custom function or method you yourself write - and so it only checks this for known special functions. (It would have to do more sophisticated analysis, possibly marking the return type of these functions as the bottom type, but that is beyond what F# does...).
given
[
1,"test2"
3,"test"
]
|> dict
// turn it into keyvaluepair sequence
|> Seq.map id
|> fun x -> x.ToDictionary<_,_,_>((fun x -> x.Key), fun x -> x.Value)
which fails to compile if I don't explicitly use the <_,_,_> after ToDictionary.
Intellisense works just fine, but compilation fails with the error: Lookup on object of indeterminate type based on information prior to this program point
So, it seems, Intellisense knows how to resolve the method call.
This seems to be a clue
|> fun x -> x.ToDictionary<_,_>((fun x -> x.Key), fun x -> x.Value)
fails with
Type constraint mismatch.
The type 'b -> 'c is not compatible with type IEqualityComparer<'a>
The type 'b -> 'c' is not compatible with the type 'IEqualityComparer<'a>'
(using external F# compiler)
x.ToDictionary((fun x -> x.Key), id)
works as expected as does
let vMap (item:KeyValuePair<_,_>) = item.Value
x.ToDictionary((fun x -> x.Key), vMap)
I've replicated the behavior in FSI and LinqPad.
As a big fan of and avid reader of Eric Lippert I really want to know
what overload resolution, (or possibly extension methods from different places) are conflicting here that the compiler is confused by?
Even though the types are known ahead, the compiler's getting confused between the overload which takes an element selector and a comparer. The lambda compiles to FSharpFunc rather than the standard delegate types in C# like Action or Func, and issues do come up translating from one to the other. To make it work, you can :
Supply a type annotation for the offending Func
fun x -> x.ToDictionary((fun pair -> pair.Key), (fun (pair : KeyValuePair<_, _>) -> pair.Value)) //compiles
or name the argument as a hint
fun x -> x.ToDictionary((fun pair -> pair.Key), elementSelector = (fun (pair) -> pair.Value))
or force it to pick the 3 argument version:
x.ToLookup((fun pair -> pair.Key), (fun (pair) -> pair.Value), EqualityComparer.Default)
Aside
In your example,
let vMap (item:KeyValuePair<_,_>) = item.Value
x.ToDictionary((fun x -> x.Key), vMap)
you would explicitly need to annotate vMap because the compiler cannot find out what type the property exists on without another pass. For example,
List.map (fun x -> x.Length) ["one"; "two"] // this fails to compile
This is one of the reasons why the pipe operator is so useful, because it allows you to avoid type annotations:
["one"; "two"] |> List.map (fun x -> x.Length) // works
List.map (fun (x:string) -> x.Length) ["one"; "two"] //also works
The short answer:
The extension method ToDictionary is defined like this:
static member ToDictionary<'TSource,_,_>(source,_,_)
but is called like this:
source.ToDictionary<'TSource,_,_>(_,_)
The long answer:
This is the F# type signature of the function you are calling from msdn.
static member ToDictionary<'TSource, 'TKey, 'TElement> :
source:IEnumerable<'TSource> *
keySelector:Func<'TSource, 'TKey> *
elementSelector:Func<'TSource, 'TElement> -> Dictionary<'TKey, 'TElement>
But I only specified two regular parameters: keySelector and elementSelector. How come this has a source parameter?!
The source parameter is actually not put in the parenthesis, but is passed in by saying x.ToDictionary, where x is the source parameter. This is actually an example of a type extension. These kinds of methods are very natural in a functional programming language like F#, but more uncommon in an object oriented language like C#, so if you're coming from the C# world, it will be pretty confusing. Anyway, if we look at the C# header, it is a little easier to understand what is going on:
public static Dictionary<TKey, TElement> ToDictionary<TSource, TKey, TElement>(
this IEnumerable<TSource> source,
Func<TSource, TKey> keySelector,
Func<TSource, TElement> elementSelector
)
So the method is defined with a "this" prefix on a first parameter even though it is technically static. It basically allows you to add methods to already defined classes without re-compiling or extending them. This is called prototyping. It's kinda rare if you're a C# programmer, but languages like python and javascript force you to be aware of this. Take this example from https://docs.python.org/3/tutorial/classes.html:
class Dog:
tricks = [] # mistaken use of a class variable
def __init__(self, name):
self.name = name
def add_trick(self, trick):
self.tricks.append(trick)
>>> d = Dog('Fido')
>>> e = Dog('Buddy')
>>> d.add_trick('roll over')
>>> e.add_trick('play dead')
>>> d.tricks # unexpectedly shared by all dogs
['roll over', 'play dead']
The method add_trick is defined with self as a first parameter, but the function is called as d.add_trick('roll over'). F# actually does this naturally as well, but in a way that mimics the way the function is called. When you declare:
member x.doSomething() = ...
or
member this.doSomething() = ...
Here, you are adding function doSomething to the prototype (or class definition) of "x"/"this". Thus, in your example, you actually have three type parameters, and three regular parameters, but one of them is not used in the call. All you have left is to declare the key selector function, and the element selector function, which you did. That's why it looks weird.
As I know, explicit type parameters in value definitions is a one way to overcome "value restriction" problem.
Is there another cases when I need to use them?
Upd: I mean "explicitly generic constructs", where type parameter is enclosed in angle brackets, i.e.
let f<'T> x = x
Polymorphic recursion is another case. That is, if you want to use a different generic instantiation within the function body, then you need to use explicit parameters on the definition:
// perfectly balanced tree
type 'a PerfectTree =
| Single of 'a
| Node of ('a*'a) PerfectTree
// need type parameters here
let rec fold<'a,'b> (f:'a -> 'b) (g:'b->'b->'b) : 'a PerfectTree -> 'b = function
| Single a -> f a
| Node t -> t |> fold (fun (a,b) -> g (f a) (f b)) g
let sum = fold id (+)
let ten = sum (Node(Node(Single((1,2),(3,4)))))
This would likely be rare, but when you want to prevent further generalization (§14.6.7):
Explicit type parameter definitions on value and member definitions can affect the process of type inference and generalization. In particular, a declaration that includes explicit generic parameters will not be generalized beyond those generic parameters. For example, consider this function:
let f<'T> (x : 'T) y = x
During type inference, this will result in a function of the following type, where '_b is a type inference variable that is yet to be resolved.
f<'T> : 'T -> '_b -> '_b
To permit generalization at these definitions, either remove the explicit generic parameters (if they can be inferred), or use the required number of parameters, as the following example shows:
let throw<'T,'U> (x:'T) (y:'U) = x
Of course, you could also accomplish this with type annotations.
Most obvious example: write a function to calculate the length of a string.
You have to write:
let f (a:string) = a.Length
and you need the annotation. Without the annotation, the compiler can't determine the type of a. Other similar examples exist - particularly when using libraries designed to be used from C#.
Dealing with updated answer:
The same problem applies - string becomes A<string> which has a method get that returns a string
let f (a:A<string>) = a.get().Length
I am trying to write a typed abstract syntax tree datatype that can represent
function application.
So far I have
type Expr<'a> =
| Constant of 'a
| Application of Expr<'b -> 'a> * Expr<'b> // error: The type parameter 'b' is not defined
I don't think there is a way in F# to write something like 'for all b' on that last line - am I approaching this problem wrongly?
In general, the F# type system is not expressive enough to (directly) define a typed abstract syntax tree as the one in your example. This can be done using generalized algebraic data types (GADTs) which are not supported in F# (although they are available in Haskell and OCaml). It would be nice to have this in F#, but I think it makes the language a bit more complex.
Technically speaking, the compiler is complaining because the type variable 'b is not defined. But of course, if you define it, then you get type Expr<'a, 'b> which has a different meaning.
If you wanted to express this in F#, you'd have to use a workaround based on interfaces (an interface can have generic method, which give you a way to express constraint like exists 'b which you need here). This will probably get very ugly very soon, so I do not think it is a good approach, but it would look something like this:
// Represents an application that returns 'a but consists
// of an argument 'b and a function 'b -> 'a
type IApplication<'a> =
abstract Appl<'b> : Expr<'b -> 'a> * Expr<'b> -> unit
and Expr<'a> =
// Constant just stores a value...
| Constant of 'a
// An application is something that we can call with an
// implementation (handler). The function then calls the
// 'Appl' method of the handler we provide. As this method
// is generic, it will be called with an appropriate type
// argument 'b that represents the type of the argument.
| Application of (IApplication<'a> -> unit)
To represent an expression tree of (fun (n:int) -> string n) 42, you could write something like:
let expr =
Application(fun appl ->
appl.Appl(Constant(fun (n:int) -> string n),
Constant(42)))
A function to evaluate the expression can be written like this:
let rec eval<'T> : Expr<'T> -> 'T = function
| Constant(v) -> v // Just return the constant
| Application(f) ->
// We use a bit of dirty mutable state (to keep types simpler for now)
let res = ref None
// Call the function with a 'handler' that evaluates function application
f { new IApplication<'T> with
member x.Appl<'A>(efunc : Expr<'A -> 'T>, earg : Expr<'A>) =
// Here we get function 'efunc' and argument 'earg'
// The type 'A is the type of the argument (which can be
// anything, depending on the created AST)
let f = eval<'A -> 'T> efunc
let a = eval<'A> earg
res := Some <| (f a) }
res.Value.Value
As I said, this is a bit really extreme workaround, so I do not think it is a good idea to actually use it. I suppose the F# way of doing this would be to use untyped Expr type. Can you write a bit more about the overall goal of your project (perhaps there is another good approach)?
I am trying to write a typed abstract syntax tree datatype that can represent
function application.
So far I have
type Expr<'a> =
| Constant of 'a
| Application of Expr<'b -> 'a> * Expr<'b> // error: The type parameter 'b' is not defined
I don't think there is a way in F# to write something like 'for all b' on that last line - am I approaching this problem wrongly?
In general, the F# type system is not expressive enough to (directly) define a typed abstract syntax tree as the one in your example. This can be done using generalized algebraic data types (GADTs) which are not supported in F# (although they are available in Haskell and OCaml). It would be nice to have this in F#, but I think it makes the language a bit more complex.
Technically speaking, the compiler is complaining because the type variable 'b is not defined. But of course, if you define it, then you get type Expr<'a, 'b> which has a different meaning.
If you wanted to express this in F#, you'd have to use a workaround based on interfaces (an interface can have generic method, which give you a way to express constraint like exists 'b which you need here). This will probably get very ugly very soon, so I do not think it is a good approach, but it would look something like this:
// Represents an application that returns 'a but consists
// of an argument 'b and a function 'b -> 'a
type IApplication<'a> =
abstract Appl<'b> : Expr<'b -> 'a> * Expr<'b> -> unit
and Expr<'a> =
// Constant just stores a value...
| Constant of 'a
// An application is something that we can call with an
// implementation (handler). The function then calls the
// 'Appl' method of the handler we provide. As this method
// is generic, it will be called with an appropriate type
// argument 'b that represents the type of the argument.
| Application of (IApplication<'a> -> unit)
To represent an expression tree of (fun (n:int) -> string n) 42, you could write something like:
let expr =
Application(fun appl ->
appl.Appl(Constant(fun (n:int) -> string n),
Constant(42)))
A function to evaluate the expression can be written like this:
let rec eval<'T> : Expr<'T> -> 'T = function
| Constant(v) -> v // Just return the constant
| Application(f) ->
// We use a bit of dirty mutable state (to keep types simpler for now)
let res = ref None
// Call the function with a 'handler' that evaluates function application
f { new IApplication<'T> with
member x.Appl<'A>(efunc : Expr<'A -> 'T>, earg : Expr<'A>) =
// Here we get function 'efunc' and argument 'earg'
// The type 'A is the type of the argument (which can be
// anything, depending on the created AST)
let f = eval<'A -> 'T> efunc
let a = eval<'A> earg
res := Some <| (f a) }
res.Value.Value
As I said, this is a bit really extreme workaround, so I do not think it is a good idea to actually use it. I suppose the F# way of doing this would be to use untyped Expr type. Can you write a bit more about the overall goal of your project (perhaps there is another good approach)?