I have a bunch of modules that export an IModule interface. So in the main program I have no problems
...
let mutable modules = Seq.empty
[<ImportMany>]
member x.Modules
with get():IEnumerable<Lazy<IModule, IModuleData>> = modules
and set(a) = modules <- a
...
But now I need to expose an interface back to those modules. So each module will import a single interface
...
let mutable parent:IParent = ?
[<Import>]
member x.Parent
with get():IParent = parent
and set(a) = parent <- a
...
So my problem is how do I go about creating my mutable "parent" when I have no initial value for it? Also, is this the appropriate way to expose an API back to component parts?
Using Unchecked.defaultof<_> should do the trick, but it means that you're circumventing the F# type system, which may be a dangerous thing to do - the system tries to prevent you from accidentally dereferencing null values (and getting NullReferenceException).
Types that are declared in F# don't have null as a proper value, which is an attempt to eliminate the usual errors caused by null. The clean F# approach is to use option types to represent the fact that a value is missing:
let mutable parent:option<IParent> = None
[<Import>]
member x.Parent
with get():IParent =
match parent with
| Some p -> p
| None -> failwith "TODO: Throw some reasonable exception here!"
and set(a) = parent <- Some(a)
If you just want to say that IParent can have a null value (perhaps because you need to use it in some C# code that will ignore the F# restriction anyway), then you can mark the type definition using a special attribute that allows using null with the type.
[<AllowNullLiteral>]
type IParent =
abstract DoStuff : unit -> unit
Then you can write let mutable parent:IParent = null. The benefit of this approach is that you can also easily check whether a value is null (using just if parent <> null then ...) which is not that obvious when you use Unchecked.defaultof<_>.
let mutable parent = Unchecked.defaultof<IParent>
should do the trick.
Following up on what Tomas explained, you should probably put your imports directly into your constructor. That will allow your code to be a bit more idiomatic.
Related
I've installed the package Topshelf.FSharp, and there is an example of how to use it here:
https://gist.github.com/haf/4252121
Part of the example defines an "Svc" (service) class like this:
type Svc() =
member x.Start() =
printfn "Started"
member x.Stop() =
printfn "Stopped"
With Owin selfhost you call one of the various static overloads of IDisposable WebApp.Start(...) to start a web server, and then dispose it to stop it. In C#, if you want to combine Topshelf and Owin, you can store the IDisposable result of Start() in a private local field of the Svc class when Start() is called, and then call Dispose() on it in the Stop() method.
In F# you can declare an unitialized field of type IDisposable using "explicit fields" (http://msdn.microsoft.com/en-us/library/dd469494.aspx), but this seems somewhat awkward, is there a better way?
You do not have to use an explicit field to produce an "uninitialized" IDisposable value. For example, to define a restartable service, you can use an implicit webApp field like this:
// Example 1: Using an implicit mutable IDisposable field.
type Svc() =
let mutable webApp = null
member __.Start() =
if webApp = null then webApp <- WebApp.Start<Startup> "http://localhost:12345"
member __.Stop() =
if webApp <> null then webApp.Dispose(); webApp <- null
// Example 2: Using an implicit mutable IDisposable option field.
type Svc'() =
let mutable webApp = None
member __.Start() =
match webApp with
| Some _ -> ()
| None -> webApp <- Some(WebApp.Start<Startup> "http://localhost:12345")
member __.Stop() =
match webApp with
| Some webAppValue -> webAppValue.Dispose(); webApp <- None
| None -> ()
If the service does not have to be restartable, I would use an (immutable) implicit lazy field instead, as #Tom suggested.
Could you make the member a Lazy<T> instead and have it automatically initialise when it is first accessed? I am not quite sure of the requirements of F# in this case i.e. why exactly this is a problem, but for deferred initialisation, Lazy would seem to be a good choice.
Updated below...
I recently started experimenting with ServiceStack in F#, so naturally I started with porting the Hello World sample:
open ServiceStack.ServiceHost
open ServiceStack.ServiceInterface
open ServiceStack.WebHost.Endpoints
[<CLIMutable; Route("/hello"); Route("/hello/{Name}")>]
type Hello = { Name : string }
[<CLIMutable>]
type HelloResponse = { Result : string }
type HelloService() =
inherit Service()
member x.Any(req:Hello) =
box { Result = sprintf "Hello, %s!" req.Name }
type HelloAppHost() =
inherit AppHostBase("Hello Web Services", typeof<HelloService>.Assembly)
override x.Configure container = ()
type Global() =
inherit System.Web.HttpApplication()
member x.Application_Start() =
let appHost = new HelloAppHost()
appHost.Init()
That works great. It's very concise, easy to work with, I love it. However, I noticed that the routes defined in the sample allow for the Name parameter to not be included. Of course, Hello, ! looks kind of lame as output. I could use String.IsNullOrEmpty, but it is idiomatic in F# to be explicit about things that are optional by using the Option type. So I modified my Hello type accordingly to see what would happen:
[<CLIMutable; Route("/hello"); Route("/hello/{Name}")>]
type Hello = { Name : string option }
As soon as I did this, the F# type system forced me to deal with the fact that Name might not have a value, so I changed HelloService to this to get everything to compile:
type HelloService() =
inherit Service()
member x.Any(req:Hello) =
box { Result =
match req.Name with
| Some name -> sprintf "Hello, %s!" name
| None -> "Hello!" }
This compiles, and runs perfectly when I don't supply a Name parameter. However, when I do supply a name...
KeyValueDataContractDeserializer: Error converting to type: Type
definitions should start with a '{', expecting serialized type
'FSharpOption`1', got string starting with: World
This wasn't a complete surprise of course, but it brings me to my question:
It would be trivial for me to write a function that can wrap an instance of type T into an instance of type FSharpOption<T>. Are there any hooks in ServiceStack that would let me provide such a function for use during deserialization? I looked, but I couldn't find any, and I'm hoping I was just looking in the wrong place.
This is more important for F# use than it might seem at first, because classes defined in F# are by default not allowed to be null. So the only (satisfying, non-hacky) way of having one class as an optional property of another class is with, you guessed it, the Option type.
Update:
I was able to sort-of get this working by making the following changes:
In the ServiceStack source, I made this type public:
ServiceStack.Text.Common.ParseFactoryDelegate
...and I also made this field public:
ServiceStack.Text.Jsv.JsvReader.ParseFnCache
With those two things public, I was able to write this code in F# to modify the ParseFnCache dictionary. I had to run this code prior to creating an instance of my AppHost - it didn't work if I ran it inside the AppHost's Configure method.
JsvReader.ParseFnCache.[typeof<Option<string>>] <-
ParseFactoryDelegate(fun () ->
ParseStringDelegate(fun s -> (if String.IsNullOrEmpty s then None else Some s) |> box))
This works for my original test case, but aside from the fact that I had to make brittle changes to the internals of ServiceStack, it sucks because I have to do it once for each type I want to be able to wrap in an Option<T>.
What would be better is if I could do this in a generic way. In C# terms, it would be awesome if I could provide to ServiceStack a Func<T, Option<T>> and ServiceStack would, when deserializing a property whose generic type definition matches that of the return type of my function, deserialize T and then pass the result into my function.
Something like that would be amazingly convenient, but I could live with the once-per-wrapped-type approach if it were actually part of ServiceStack and not my ugly hack that probably breaks something somewhere else.
So there are a couple of extensibility points in ServiceStack, on the framework level you can add your own Custom Request Binder this allows you to provide your own model binder that's used, e.g:
base.RequestBinders.Add(typeof(Hello), httpReq => {
var requestDto = ...;
return requestDto;
});
But then you would need to handle the model binding for the different Content-Types yourself, see CreateContentTypeRequest for how ServiceStack does it.
Then there are hooks at the JSON Serializer level, e.g:
JsConfig<Hello>.OnDeserializedFn = dto => newDto;
This lets you modify the instance of the type returned, but it still needs to be the same type but it looks like the F# option modifier changes the structural definition of the type?
But I'm open to adding any hooks that would make ServiceStack more palatable for F#.
What does the code look like to generically convert a normal Hello type to an F# Hello type with option?
The only thing I can think of is to replace the option type with your own type, one that has an implicit conversion from string to myOption, and anything else you need.
Not all that nice, but workable. Your type would probably also need to be serializable.
type myOption =
| None
| Some of string
static member public op_Implicit (s:string) = if s <> null then Some s else None
member public this.Value = match this with
| Some s -> s
| _ -> null
member this.Opt = match this with
| Some s -> Option.Some s
| None -> Option.None
Your record type would then be
[<CLIMutable>]
type Hello =
{ Name : myOption }
On the other hand, ServiceStack is open source, so maybe something could be done there.
Under normal circumstances, F# functions can be converted to delegates by calling new DelegateType and passing in the function as an argument. But when the delegate contains byref parameter, this is not possible directly. For example the code:
type ActionByRef<'a> = delegate of 'a byref -> unit
let f (x:double byref) =
x <- 6.0
let x = ref 42.0
let d = new ActionByRef<_>(f)
won't compile, giving the following error:
This function value is being used to construct a delegate type whose signature includes a byref argument. You must use an explicit lambda expression taking 1 arguments.
Following the error, modifying the code to use
let d = new ActionByRef<_>(fun x -> f(&x))
works. But my question is: why is this necessary? Why won't F# allow the conversion from named function to this delegate, but conversion from lambda is fine?
I came upon this behavior when researching another question. I realize byref is meant only for compatibility with other .Net languages.
I think the problem is that byref<'T> is not an actual type in F# - it looks like a type (to make the language simpler), but it gets compiled to a parameter marked with the out flag. This means that byref<'T> can be only used in a place where the compiler can actually use the out flag.
The problem with function values is that you can construct function e.g. by partial application:
let foo (n:int) (b:byref<int>) =
b <- n
When you pass foo as an argument to a delegate constructor, it is a specific case of partial application (with no arguments), but partial application actually needs to construct a new method and then give that to the delegate:
type IntRefAction = delegate of byref<int> -> unit
let ac = IntRefAction(foo 5)
The compiler could be clever and generate new method with byref parameter (or out flag) and then pass that by reference to the actual function, but in general, there will be other compiler-generated method when you don't use the fun ... -> ... syntax. Handling this would add complexity and I think that's a relatively rare case, so the F# compiler doesn't do that and asks you to be more explicit...
I've been struggling to get this to compile for about an hour. It must be something stupid. Can you spot it?
in my lib project:
namespace TravelerStuff
open System
type Traveler =
abstract GetData : unit -> unit
type public DeltaTraveler() =
interface Traveler with
member v.GetData () =
printf "hello"
and in my console test app:
[<EntryPoint>] let main _ =
let traveler = new TravelerStuff.DeltaTraveler()
traveler.GetData // this line won't compile: (The field, constructor or member 'GetData' is not defined)
As gradbot says, F# doesn't currently implicitly convert values to interfaces when searching for members. Also, F# only uses explicit interface implementation (as known from C#) and not implicit implementation where members are not only compiled as implementation of an interface, but also as ordinary (directly visible) members of the type.
Aside from casting, you can duplicate the member in the type definition:
type DeltaTraveler() =
member v.GetData () = printf "hello"
interface Traveler with
member v.GetData () = v.GetData()
Also, if you just want to implement an interface, but don't need to add any members, you can use F# object expressions (which are more lightweight):
let deltaTraveler() =
{ new Traveler with
member v.GetData () = printf "hello" }
// The function directly returns value of type 'Traveler'
let t = deltaTraveler()
t.GetData()
You need to upcast. F# currently won't do it for you in this situation.
(traveler :> TravelerStuff.Traveler).GetData()
// open the namespace to reduce typing.
open TravelerStuff
(traveler :> Traveler).GetData()
Snip from F# docs.
In many object-oriented languages,
upcasting is implicit; in F#, the
rules are slightly different.
Upcasting is applied automatically
when you pass arguments to methods on
an object type. However, for let-bound
functions in a module, upcasting is
not automatic, unless the parameter
type is declared as a flexible type.
For more information, see Flexible Types (F#).
In F# its a big deal that they do not have null values and do not want to support it. Still the programmer has to make cases for None similar to C# programmers having to check != null.
Is None really less evil than null?
The problem with null is that you have the possibility to use it almost everywhere, i.e. introduce invalid states where this is neither intended nor makes sense.
Having an 'a option is always an explicit thing. You state that an operation can either produce Some meaningful value or None, which the compiler can enforce to be checked and processed correctly.
By discouraging null in favor of an 'a option-type, you basically have the guarantee that any value in your program is somehow meaningful. If some code is designed to work with these values, you cannot simply pass invalid ones, and if there is a function of option-type, you will have to cover all possibilities.
Of course it is less evil!
If you don't check against None, then it most cases you'll have a type error in your application, meaning that it won't compile, therefore it cannot crash with a NullReferenceException (since None translates to null).
For example:
let myObject : option<_> = getObjectToUse() // you get a Some<'T>, added explicit typing for clarity
match myObject with
| Some o -> o.DoSomething()
| None -> ... // you have to explicitly handle this case
It is still possible to achieve C#-like behavior, but it is less intuitive, as you have to explicitly say "ignore that this can be None":
let o = myObject.Value // throws NullReferenceException if myObject = None
In C#, you're not forced to consider the case of your variable being null, so it is possible that you simply forget to make a check. Same example as above:
var myObject = GetObjectToUse(); // you get back a nullable type
myObject.DoSomething() // no type error, but a runtime error
Edit: Stephen Swensen is absolutely right, my example code had some flaws, was writing it in a hurry. Fixed. Thank you!
Let's say I show you a function definition like this:
val getPersonByName : (name : string) -> Person
What do you think happens when you pass in a name of a person who doesn't exist in the data store?
Does the function throw a NotFound exception?
Does it return null?
Does it create the person if they don't exist?
Short of reading the code (if you have access to it), reading the documentation (if someone was kindly enough to write it), or just calling the function, you have no way of knowing. And that's basically the problem with null values: they look and act just like non-null values, at least until runtime.
Now let's say you have a function with this signature instead:
val getPersonByName : (name : string) -> option<Person>
This definition makes it very explicit what happens: you'll either get a person back or you won't, and this sort of information is communicated in the function's data type. Usually, you have a better guarantee of handling both cases of a option type than a potentially null value.
I'd say option types are much more benevolent than nulls.
In F# its a big deal that they do not have null values and do not want to support it. Still the programmer has to make cases for None similar to C# programmers having to check != null.
Is None really less evil than null?
Whereas null introduces potential sources of run-time error (NullRefereceException) every time you dereference an object in C#, None forces you to make the sources of run-time error explicit in F#.
For example, invoking GetHashCode on a given object causes C# to silently inject a source of run-time error:
class Foo {
int m;
Foo(int n) { m=n; }
int Hash() { return m; }
static int hash(Foo o) { return o.Hash(); }
};
In contrast, the equivalent code in F# is expected to be null free:
type Foo =
{ m: int }
member foo.Hash() = foo.m
let hash (o: Foo) = o.Hash()
If you really wanted an optional value in F# then you would use the option type and you must handle it explicitly or the compiler will give a warning or error:
let maybeHash (o: Foo option) =
match o with
| None -> 0
| Some o -> o.Hash()
You can still get NullReferenceException in F# by circumventing the type system (which is required for interop):
> hash (box null |> unbox);;
System.NullReferenceException: Object reference not set to an instance of an object.
at Microsoft.FSharp.Core.LanguagePrimitives.IntrinsicFunctions.UnboxGeneric[T](Object source)
at <StartupCode$FSI_0021>.$FSI_0021.main#()
Stopped due to error