6
comments
on 2/28/2018 12:11 AM

Writing reactive code in WebSharper.UI can result in the use of a lot of functions and combinators, and code that is more verbose than a static counterpart would be. However, a recent new feature can help reduce most of this overhead: the V property. Let's get started with it!

Note: examples in this article are given in F#; but all the concepts are applicable in C#, except the final part about V and lenses, which is only compatible with F# record types.

The working example

Here is a simple static form asking the user for their name. It has two input fields and a paragraph showing the value. Obviously this last part is not very useful yet since this is only a static form; it just always shows whatever was the initial value.

1
2
3
4
5
6
7
8
9
10
11
12
type FullName =
    {
        First: string
        Last: string
    }

let promptName (name: FullName) =
    form [] [
        input [attr.value name.First] []
        input [attr.value name.Last] []
        p [] [text (sprintf "Hello, %s %s!" name.First name.Last)]
    ]

See it in action

Over the course of this article, we'll transform it into a dynamic form where the paragraph reacts to the user's inputs. Spoiler alert: here's what our final code will look like.

1
2
3
4
5
6
let promptName (name: Var<FullName>) =
    form [] [
        Doc.Input [] (Lens name.V.First)
        Doc.Input [] (Lens name.V.Last)
        p [] [text (sprintf "Hello, %s %s!" name.V.First name.V.Last)]
    ]

See it in action

I would argue that the above is exactly as readable as the original static code, while providing extra reactive functionality. Now let's see how the sausage is made!

A first, unmodular attempt

Before we try to make our little function reactive, here's a refresher on how reactive code works in WebSharper. Fundamentally, it is based on the combination of two types:

  • Var<'T> is a mutable cell containing a value of type 'T. It's basically equivalent to a ref<'T>, except that it can be reactively observed by a...
  • View<'T>, which is a read-only reactive 'T.

A number of functions integrate these into the page. For example:

  • Doc.Input is similar to the regular input function that creates an <input> element, but it takes a Var<string> and binds itself to it. When the user types, the Var is updated, and when the Var changes due to other code, the input box is updated.
  • textView shows text content, similar to text, except it's reactive: the text content changes whenever the View<string> changes.

Armed with these, let's write our first attempt at a reactive form.

1
2
3
4
5
6
7
8
9
10
11
12
    let promptName (first: Var<string>) (last: Var<string>) =
        form [] [
            Doc.Input [] first
            Doc.Input [] last
            p [] [
                textView (
                    View.Map2
                        (fun f l -> sprintf "Hello, %s %s!" f l)
                        first.View last.View
                )
            ]
        ]

See it in action

View.Map2 creates a new View by applying a function to the values of two input Views.

This works well enough, but there are several improvements we could make.

  • First, we completely dropped our FullName record, and instead we are taking the two fields as separate Vars. This is pretty bad for the maintainability of our code. FullName is probably a type from our application's model: we pass FullName values around, maybe even between the server and the client. We don't want to have to convert it into its components and then put them back together whenever we want the user to interact with it.
  • Second, this is a bit verbose, isn't it? We have to call another function and write an anonymous function just to display two strings. That feels overkill.

Deriving Vars with Lenses

Let's attack the first problem. Ideally, we'd like our function to receive a Var<FullName> and be able to deal with it. We want each of our input boxes to transparently update the corresponding field of this Var. This is what lenses are for.

Lenses are a way to "zoom in" (hence the name) on a part of a Var, such as a field of a record. You provide a getter (a function that retrieves the field value from the full value) and a setter (a function that takes an old full value and a new field value, and returns the new full value) and it gives you back a new Var for your field.

1
2
3
let promptName (name: Var<FullName>) =
    let first = Var.Lens name (fun n -> n.First) (fun n f -> { n with First = f })
    // ...

The important aspect of this is that the reactivity works as you would expect: if you set the value of the full Var, the field Var's own View is updated, and vice-versa.

Armed with this new tool, we can write a better version of promptName that uses FullName.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
let promptName (name: Var<FullName>) =
    let first = Var.Lens name (fun n -> n.First) (fun n f -> { n with First = f })
    let last = Var.Lens name (fun n -> n.Last) (fun n l -> { n with Last = l })
    form [] [
        Doc.Input [] first
        Doc.Input [] last
        p [] [
            textView (
                View.Map2
                    (fun f l -> sprintf "Hello, %s %s!" f l)
                    first.View last.View
            )
        ]
    ]

See it in action

This works nicely! But... we've made the verbosity problem worse, haven't we? ☹️

Introducing V

We'd like to be able to compose together Views as naturally as we would compose simple values. Instead of this:

1
2
3
View.Map2
    (fun f l -> sprintf "Hello, %s %s!" f l)
    first.View last.View

We want to write this:

1
sprintf "Hello, %s %s!" first.View last.View

Of course, that isn't quite possible directly &mdash; the types don't even match! first.View has type View<string>, but sprintf wants a string.

This is where WebSharper introduces a bit of "magic". The type View<'T> has a field called V, of type 'T. Now, 'T is not a reactive type, so this field should only be able to return one value: the current value of our View at the moment of calling .V. But when you use .V in an argument to certain special functions, WebSharper automatically transforms this argument into the appropriate calls to View.Map, View.Map2, or whatever else is needed to build a reactive value from this expression.

text is one such special function. When its argument contains uses of .V, it is compiled into a call to textView and the argument is transformed to be properly reactive. So this:

1
2
3
// Assuming you have a value myView: View<string>

text ("myView's current value is " + myView.V)

is exactly equivalent to this:

1
textView (myView |> View.Map (fun x -> "myView's current value is " + x))

That's much more readable, don't you think? It allows us to simplify our display function tremendously:

1
2
3
4
5
6
7
8
9
10
let promptName (name: Var<FullName>) =
    let first = Var.Lens name (fun n -> n.First) (fun n f -> { n with First = f })
    let last = Var.Lens name (fun n -> n.Last) (fun n l -> { n with Last = l })
    form [] [
        Doc.Input [] first
        Doc.Input [] last
        p [] [
            text (sprintf "Hello, %s %s!" first.View.V last.View.V)
        ]
    ]

In fact, we can simplify it even further, as Var also has a field .V, which is equivalent to .View.V.

1
2
3
4
5
6
7
8
9
10
let promptName (name: Var<FullName>) =
    let first = Var.Lens name (fun n -> n.First) (fun n f -> { n with First = f })
    let last = Var.Lens name (fun n -> n.Last) (fun n l -> { n with Last = l })
    form [] [
        Doc.Input [] first
        Doc.Input [] last
        p [] [
            text (sprintf "Hello, %s %s!" first.V last.V)
        ]
    ]

See it in action

Side note: earlier I said that .V "should only be able to return one value: the current value of our View at the moment of calling .V". That's the only thing it could do, if it did anything; but in fact, it causes a compile error. That's because .V is really intended to simplify reactive code; so I made sure that people wouldn't end up with non-reactive code by mistake. Plus, there are some cases (in particular when using View.MapAsync) where it is simply not possible to extract the current value of a View.

Using .V with lenses

Now let's look into simplifying our lenses. They really feel like boilerplate: getting and setting a field in a record is a very uniform pattern. Can WebSharper's "magic" help us with it? As it turns out, yes it can.

As said before, Var also has a .V field. In addition to serving as a shortcut for .View.V, it can also be used to write lenses very concisely, in conjunction with the Lens function. So this:

1
Lens name.V.First

is exactly equivalent to this:

1
Var.Lens name (fun n -> n.First) (fun n f -> { n with First = f })

Note that, at least for the moment, Lens is only able to deal with record fields.

This allows us to simplify our code further:

1
2
3
4
5
6
7
8
9
10
let promptName (name: Var<FullName>) =
    let first = Lens name.V.First
    let last = Lens name.V.Last
    form [] [
        Doc.Input [] first
        Doc.Input [] last
        p [] [
            text (sprintf "Hello, %s %s!" first.V last.V)
        ]
    ]

Now, this final step is more subjective: I think that in the above code, the definitions of first and last have become short enough that it's a bit redundant to give them names at all. I would shorten it to the following:

1
2
3
4
5
6
let promptName (name: Var<FullName>) =
    form [] [
        Doc.Input [] (Lens name.V.First)
        Doc.Input [] (Lens name.V.Last)
        p [] [text (sprintf "Hello, %s %s!" name.V.First name.V.Last)]
    ]

See it in action

And here we are! Our reactive code is now clear and maintainable.

.

Sorry, if I want to follow this interesting article by doing each step in Visual Studio 2017, which WebSharper project type should I choose at first?

Here is a screenshot below

choose proj

I'm also getting an error:

The value or constructor 'Lens' is not defined.

Thank you!

By on 2/28/2018 3:12 AM ()

Updated nuget package! All very cool! Thank you!

nuget

By on 2/28/2018 3:56 AM ()

Single-Page Application is the best to write code that runs entirely on the client.

Just put your code in the Client module in Client.fs, and replace the body of the Main() function with the body of the Main value from TryWebSharper (the "See it in action" links throughout the article).

By on 2/28/2018 3:19 AM ()

Also, I need to put the content of the Markup from TryWebSharper into the html tag of index.html, is it right? Then, there seems to be an issue when I build and run the project from VisualStudio through iisexpress... I can't see the main dynamically replaced...

By on 2/28/2018 8:45 AM ()

Ok, found this, it needs the scripts too in the markup

1
2
3
4
5
6
7
8
9
10
11
12
<!DOCTYPE html>
<html lang="en">
<head>
    <title>Reactive UI</title>
    <script type="text/javascript" src="Content/ReactiveSPA.head.js"></script>
</head>
<body>
    <div id="main"></div>
    <!--[BODY]-->
    <script type="text/javascript" src="Content/ReactiveSPA.min.js"></script>
</body>
</html>
By on 2/28/2018 8:49 AM ()

Perfect, is the reactive approach with lensing applicable to a code like this and how? Thank you again!

By on 2/28/2018 7:14 AM ()
IntelliFactory Offices Copyright (c) 2011-2012 IntelliFactory. All rights reserved.
Home | Products | Consulting | Trainings | Blogs | Jobs | Contact Us | Terms of Use | Privacy Policy | Cookie Policy
Built with WebSharper