ploeh blog danish software design

In F#, Discriminated Unions provide a good alternative to overloading.

When you're learning a new programming language, you may experience these phases:

1. [language x] is the coolest thing ever :D
2. [language x] sucks, because it can't do [xyz], like [language y], which I know really well :(
3. Now I finally understand that in [language x] I don't need [xyz], because it has different idioms o_O

From C#, I'm used to method overloading as a design technique to provide an API with easy-to-learn default methods, and more complicated, but more flexible, methods - all in the same family of methods. Alas, in F#, function overloading isn't possible. (Well, method overloading is still possible, because you can create .NET classes in F#, but if you're creating a module with free-standing functions, you can't have two functions with the same name, but different parameters.)

This really bothered me until I realized that F# has other language constructs that enable you to approach that problem differently. Discriminated Unions is one such language construct.

Multiple related functions #

Recently, I was working with a little module that enabled me to list a range of dates. As an example, I wanted to be able to list all dates in a given year, or all dates in a given month.

My first attempt looked like this:

module Dates =
    let InitInfinite (date : DateTime) =
        date |> Seq.unfold (fun d -> Some(d, d.AddDays 1.0))
 
    let InYear year =
        DateTime(year, 1, 1)
        |> InitInfinite
        |> Seq.takeWhile (fun d -> d.Year = year)
 
    let InMonth year month =
        DateTime(year, month, 1)
        |> InitInfinite
        |> Seq.takeWhile (fun d -> d.Month = month)

These functions did what I wanted them to do, but I found the names awkward. The InYear and InMonth functions are closely related, so I would have liked to call them both In, as I would have been able to do in C#. That would have enabled me to write code such as:

let actual = Dates.In year

and

let actual = Dates.In year month

where year and month are integer values. Alas, that's not possible, so instead I had to settle for the slightly more clumsy InYear:

let actual = Dates.InYear year

and InMonth:

let actual = Dates.InMonth year month

That is, until I realized that I could model this better with a Discriminated Union.

One function #

After a day or so, I had one of those small revelations described in Domain-Driven Design: implicitly, I was working with a concept of a period. Time to make the implicit concept explicit:

type Period =
    | Year of int
    | Month of int * int
 
module Dates =
    let InitInfinite (date : DateTime) =
        date |> Seq.unfold (fun d -> Some(d, d.AddDays 1.0))
 
    let private InYear year =
        DateTime(year, 1, 1)
        |> InitInfinite
        |> Seq.takeWhile (fun d -> d.Year = year)
 
    let private InMonth year month =
        DateTime(year, month, 1)
        |> InitInfinite
        |> Seq.takeWhile (fun d -> d.Month = month)
 
    let In period =
        match period with
        | Year(y) -> InYear y
        | Month(y, m) -> InMonth y m

Notice that I defined a Period Discriminated Union outside the module, because it enables me to write client code like:

let actual = Dates.In(Year(year))

and

let actual = Dates.In(Month(year, month))

This syntax requires slightly more characters than the previous alternative, but is (subjectively) more elegant.

If you prefer, you can refactor the In function to:

let In = function
    | Year(y) -> InYear y
    | Month(y, m) -> InMonth y m

You may have noticed that this implementation still relies on the two private functions InYear and InMonth, but it's easy to refactor the In function to a single function:

let In period =
    let generate dt predicate =
        dt |> InitInfinite |> Seq.takeWhile predicate
    match period with
    | Year(y) -> generate (DateTime(y, 1, 1)) (fun d -> d.Year = y)
    | Month(y, m) -> generate (DateTime(y, m, 1)) (fun d -> d.Month = m)

As you can see, the introduction of a Period Discriminated Union enabled me to express the API in a way that closely resembles what I originally envisioned.

A richer API #

Once I made the change to a Discriminated Union, I discovered that I could make my API richer. Soon, I had this Dates module:

type Period =
    | Year of int
    | Month of int * int
    | Day of int * int * int
 
module Dates =
    let InitInfinite (date : DateTime) =
        date |> Seq.unfold (fun d -> Some(d, d.AddDays 1.0))
 
    let In period =
        let generate dt predicate =
            dt |> InitInfinite |> Seq.takeWhile predicate
        match period with
        | Year(y) -> generate (DateTime(y, 1, 1)) (fun d -> d.Year = y)
        | Month(y, m) -> generate (DateTime(y, m, 1)) (fun d -> d.Month = m)
        | Day(y, m, d) -> DateTime(y, m, d) |> Seq.singleton
 
    let BoundariesIn period =
        let getBoundaries firstTick (forward : DateTime -> DateTime) =
            let lastTick = forward(firstTick).AddTicks -1L
            (firstTick, lastTick)
        match period with
        | Year(y) -> getBoundaries (DateTime(y, 1, 1)) (fun d -> d.AddYears 1)
        | Month(y, m) -> getBoundaries (DateTime(y, m, 1)) (fun d -> d.AddMonths 1)
        | Day(y, m, d) -> getBoundaries (DateTime(y, m, d)) (fun d -> d.AddDays 1.0)

Notice that I added a Day case. Originally, I didn't think it would be valuable, as Dates.In(Day(year, month, day)) seemed like quite a convoluted way of saying DateTime(year, month, day). However, it turned out that the In abstraction was valuable, also with a single day - simply because it's an abstraction.

Additionally, I then discovered the utility of a function called BoundariesIn, which gives me the boundaries of a Period - that is, the very first and last tick of the Period.

Summary #

It's easy to become frustrated while learning a new programming language. In F#, I was annoyed by the lack of function overloading, until I realized that a single function taking a Discriminated Union might actually be a richer idiom.

With F#, it's easy to create DTOs for use with the ASP.NET Web API, using record types.

When writing HTTP (or RESTful) web services with the ASP.NET Web API, the most normal approach is to define Data Transfer Objects (DTOs), which represents the data structures that go on the wire in the form of JSON or XML. While I usually call these boundary objects for renditions (a bit punny), a more normal terminology is, indeed, DTOs.

To enable the default .NET serializers (and particularly, deserializers) to do their magic, these DTOs must be mutable and have a default constructor. Not particularly something that seems to fit nicely with F#.

Until recently, I've been declaring such DTOs like this in F#:

type HomeRendition() =
    [<DefaultValue>] val mutable Message : string
    [<DefaultValue>] val mutable Time : string

However, then I discovered the [<CLIMutable>] attribute. This effectively turn a standard F# record type into something that can be serialized with the normal .NET serializers. This means that it's possible to redefine the above HomeRendition like this:

[<CLIMutable>]
type HomeRendition = {
    Message : string
    Time : string }

This is much better, because this looks like a proper immutable record type from the perspective of the F# compiler. However, from a CLI perspective, the HomeRendition class has a default constructor, and mutable properties.

A DTO defined like this still works with the ASP.NET Web API, although by default, the serialization looks a bit strange:

<HomeRendition>
  <Message_x0040_>
    This message is served by a pure F# ASP.NET Web API implementation.
  </Message_x0040_>
  <Time_x0040_>2013-10-15T23:32:42.6725088+02:00</Time_x0040_>
</HomeRendition>

The reason for that is that the [<CLIMutable>] attribute causes the record type to be compile with auto-generated internal mutable fields, and these are named by appending an @ character - in this case, the field names become Message@ and Time@. Since the unicode value for the @ character is x0040, these field names become Message_x0040_ and Time_x0040_.

Wait a minute! Did I say internal fields? Yes, I did. Then why are the internal fields being serialized instead of the public properties? Well, what can I say, other than the DataContractSerializer once again proves to be a rather poor choice of default serializer. Yet another reason to use a sane XML serializer.

No doubt, one of my readers can come up with a good solution for the DataContractSerializer too, but since I always switch my Web API services to use a proper XML serializer, I don't really care:

GlobalConfiguration.Configuration.Formatters.XmlFormatter.UseXmlSerializer <- true

Now all is well again:

<HomeRendition>
  <Message>
    This message is served by a pure F# ASP.NET Web API implementation.
  </Message>
  <Time>2013-10-15T23:50:06.9025322+02:00</Time>
</HomeRendition>

That's it: much easier, and more robust, Web API DTOs with F# record types. Just apply the [<CLIMutable>] attribute.

Comments

config.Formatters.JsonFormatter.SerializerSettings.ContractResolver <-
    Newtonsoft.Json.Serialization.CamelCasePropertyNamesContractResolver()

type HomeRecord = {
    Message : string
    Time : string
}

type InputModel = {
    Message : string
}

type HomeController() =
    inherit ApiController()
    member this.Get() =
        this.Ok({ Message = "Hello from F#!"; Time = DateTime.Now.ToString() })
    member this.Post(input : InputModel) =
        this.Ok(input.Message)

You can run a validation task for every commit in a Git branch in order to verify the integrity of a Git branch's history.

Just like the Soviet Union, I occasionally prefer to rewrite history... although I limit myself to rewriting the history of one or more Git branches. When I do that, I'd like to verify that I didn't break anything after an interactive rebase. It actually turns out that this can easily happen. Some commits are left in a state where they either don't compile, or tests fail. Git is happy, but the code isn't. To deal with such situations, I wanted a tool that could verify each and every commit in a branch like this:

1. Check out a commit.
2. Compile the code.
3. Run all tests.
4. Repeat for the next commit, or stop if there's a failure.

Motivation #

If you don't care about my motivation for doing this, you can skip this section and move on to the solution. However, some people might like to point out to me that I shouldn't rewrite the history of my Git repositories - and particularly not of the master branch. I agree, for all repositories where I actually collaborate with other programmers.

However, I also use Git locally to produce teaching materials. As an example, if you download the example code for my Pluralsight courses, you'll see that it's all available as Git repositories. I think that's an added benefit for the student, because not only can you see the final result, but you can also see all the steps that lead me there. I even occasionally write commit messages longer than a single line, explaining my thinking as I check in code.

Like everyone else, I'm fallible, so often, when I prepare such materials, I end up taking some detours that I feel will confuse students more than it will help. Thus, I have a need to be able to edit my Git repositories. Right now, as an example, I'm working with a repository with 120 commits on the master branch, and I need to make some changes in the beginning of the history.

Solution #

With much help from Darrell Hamilton, who ended up providing this gist, I was able to piece together this bash script:

#!/bin/sh
COMMITS=$(git --git-dir=BookingApi/.git --work-tree=BookingApi log --oneline  --reverse | cut -d " " -f 1)
CODE=0

git --git-dir=BookingApi/.git --work-tree=BookingApi reset --hard
git --git-dir=BookingApi/.git --work-tree=BookingApi clean -xdf
 
for COMMIT in $COMMITS
do
    git --git-dir=BookingApi/.git --work-tree=BookingApi checkout $COMMIT
    # run-tests          
    ./build.sh

    if [ $? -eq 0 ]
    then
        echo $COMMIT - passed
    else
        echo $COMMIT - failed
        exit
    fi

    git --git-dir=BookingApi/.git --work-tree=BookingApi reset --hard
    git --git-dir=BookingApi/.git --work-tree=BookingApi clean -xdf
done
 
git --git-dir=BookingApi/.git --work-tree=BookingApi checkout master

As you can tell, it follows the algorithm outlined above. First, it does a git log to get all the commits in the branch, and then it loops through all of them, one by one. For each commit, it compiles and runs all tests by calling out to a separate build.sh script. If the build and test step succeeds, it cleans up the working directory and moves on to the next step. If the verification step fails, it stops, so that I can examine the problem.

(The reason for the use of --git-dir and --work-tree is that I need to run the script from outside the Git repository itself; otherwise, the git clean -xdf step would delete the script files!)

This has turned out to work beautifully, and has already caught quite a number of commits that Git could happily rebase, but afterwards either couldn't compile, or had failing tests.

Running those tests in a tight loop has also finally provided some work for my multi-core processor:

Each run through those 120 commits takes about 20 minutes, so, even though today we have fast compilers, once again, we have an excuse for slacking off.

Comments

My book Dependency Injection in .NET has received a Jolt Productivity Award!

It turns out that my book Dependency Injection in .NET has received a Jolt Productivity Award 2013 in the Best Books category. This is a completely unexpected (but obviously very pleasant) surprise :D

The motivation for giving the award states that it:

"provide[s] a brilliant and remarkably readable explanation of DI"

and concludes that

"It's the definitive guide to DI and the first place architects and implementers should turn for a truly deep and complete understanding of the pattern — regardless of whether .NET is part of their toolkit.""

Although it's already been selling well, the book is still available either directly from Manning, or via Amazon.com - and, in case you didn't know, every time you buy a Manning book, you automatically also get the electronic version, which (in the case of my book), includes a Kindle version, as well as a PDF. You can also skip the physical copy and buy only the electronic version, but this option is (AFAIK) only available from Manning.

What an honor!

Note to future self: how I got Karma running from bash in Windows.

Yesterday, I spent quite a bit of time getting Karma to run from bash on my Windows 7 (x64) laptop. Since, a month from now, I'll have absolutely no recollection of how I achieved this, I'm writing it down here for the benefit of my future self, as well as any other people who may find this valuable.

Add karma to PATH #

My first problem was that 'karma' wasn't a recognized command. Although I'd used Chocolatey to install node.js, used npm to install Karma, and 'karma' was a recognized command from PowerShell, bash didn't recognize it. While I don't yet know if the following is the 100 % correct solution, I managed to make bash recognize karma by adding my karma directory to my bash path:

PATH=$PATH:~/appdata/roaming/npm/node_modules/karma/bin

This works in the session itself, but will be forgotten by bash unless you add it to ~/.bash_profile. BTW, that file doesn't exist by default in a new Git Bash installation on Windows, so you'll have to add it yourself.

Chrome path #

My next problem was that, while karma could be found, I got the error message that it couldn't find Chrome, and that I should set the CHROME_BIN environment variable. After much experimentation, I managed to make it work by setting:

CHROME_BIN=/c/Program\ Files\ \(x86\)/Google/Chrome/Application/chrome.exe
export CHROME_BIN

That looks really simple, so why did I waste so much time shaving that yak? Well, setting the environment variable was one thing, but it didn't work before I also figured out to export it.

Once again, I added these lines to my .bash_profile file. Now I can run Karma from bash on Windows.

With Reactive Extensions, and a bit of composition, you can publish and subscribe to events in a structurally safe way.

Previously, in my series about Dependency Injection and events, you learned how to connect a publisher and a subscriber via a third party (often the Composition Root).

The problem with that approach is that while it's loosely coupled, it's too easy to forget to connect the publisher and the subscriber. It's also possible to forget to unsubscribe. In neither case can the compiler help you.

However, the advantage of using Reactive Extensions over .NET events is that IObserver<T> is composable. That turns out to be quite an important distinction!

The problem with IObservable<T> #

While I consider IObserver<T> to be an extremely versatile interface, I consider IObservable<T> to be of limited usefulness. Consider its definition:

public interface IObservable<out T>
{
    IDisposable Subscribe(IObserver<T> observer);
}

The idea is that the publisher (the Observable) receives a subscriber (the Observer) via Method Injection. When the method completes, the subscriber is considered subscribed to the publisher's events, until the subscriber disposes of the returned subscription reference.

There's a couple of problems with this design:

• It's too easy to forget to invoke the Subscribe method. This is not a problem if you're writing a system in which publishers dynamically subscribe to event streams, but it's problematic if your system relies on certain publishers and subscribers to be connected.
• It implies mutation in the publisher, because the publisher must somehow keep a list of all its subscribers.
• It breaks Command Query Separation (CQS).
• Since it implies mutation, it's not thread-safe by default.

From Method Injection to Constructor Injection #

As you learned in the last couple of articles, the subscriber should not require any dependency in order to react to events. Yet, if Method Injection via IObservable<T> isn't a good approach either, then what's left?

Good old Constructor Injection.

The important realization is that it's not the subscriber (NeedyClass, in previous examples) that requires a dependency - it's the publisher!

Imagine that until now, you've had a publisher implementing IObservable<T>. In keeping with the running example throughout this series, this was the publisher that originally implemented IDependency. Thus, it's still called RealDependency. For simplicity's sake, assume that its implementation is as simple as this:

public class RealDependency : IObservable<Unit>
{
    private readonly Subject<Unit> subject;
 
    public RealDependency()
    {
        this.subject = new Subject<Unit>();
    }
 
    public void MakeItHappen()
    {
        this.subject.OnNext(Unit.Default);
    }
 
    public IDisposable Subscribe(IObserver<Unit> observer)
    {
        return this.subject.Subscribe(observer);
    }
}

What if, instead of implementing IObservable<Unit>, this class would use Constructor Injection to request an IObserver<Unit>? Then it would look like this:

public class RealDependency
{
    private readonly IObserver<Unit> observer;
 
    public RealDependency(IObserver<Unit> observer)
    {
        this.observer = observer;
    }
 
    public void MakeItHappen()
    {
        this.observer.OnNext(Unit.Default);
    }
}

That's much simpler, and you just got rid of an entire type (IObservable<Unit>)! Even better, you've also eliminated all use of IDisposable. Oh, and it also conforms to CQS, and is thread-safe.

Connection #

The names of the concrete classes are completely off by now, but you can connect publisher (RealDependency) with its subscriber (NeedyClass) from a third party (Composition Root):

var subscriber = new NeedyClass();
var publisher = new RealDependency(subscriber);

Not only is this easy, the statically typed structure of both classes helps you do the right thing: the compiler will issue an error if you don't supply a subscriber to the publisher.

But wait, you say: now the publisher is forced to have a single observer. Isn't the whole idea about publish/subscribe that you can have an arbitrary number of subscribers for a given publisher? Yes, it is, and that's still possible.

Composition #

More than a single subscriber is easy if you introduce a Composite:

public class CompositeObserver<T> : IObserver<T>
{
    private readonly IEnumerable<IObserver<T>> observers;
 
    public CompositeObserver(IEnumerable<IObserver<T>> observers)
    {
        this.observers = observers;
    }
 
    public void OnCompleted()
    {
        foreach (var observer in this.observers)
            observer.OnCompleted();
    }
 
    public void OnError(Exception error)
    {
        foreach (var observer in this.observers)
            observer.OnError(error);
    }
 
    public void OnNext(T value)
    {
        foreach (var observer in this.observers)
            observer.OnNext(value);
    }
}

While it looks like a bit of work, this class is so reusable that I wonder why it's not included in the Rx library itself... It enables you to subscribe any number of subscribers to the publisher, e.g. two:

var sub1 = new NeedyClass();
var sub2 = new AnotherObserver();
var publisher = 
    new RealDependency(
        new CompositeObserver<Unit>(
            new IObserver<Unit>[] { sub1, sub2 }));

I'll leave it as an exercise to the reader to figure out how to implement the scenario with no subscribers :)

Conclusion #

Sticking to IObserver<T> and simply injecting it into the publishers is much simpler than any other alternative I've described so far. Nonetheless, keep in mind that the reason this simplification works so well is because it assumes that you know all subscribers when you compose your object graph.

There's a reason the IObservable<T> interface exists, and that's to support scenarios where publishers and subscribers come and go during the lifetime of an application. The simplification described here doesn't handle that scenario, but if you don't need that flexibility, you can greatly simplify your eventing infrastructure by disposing of IObservable<T> ;)

Comments

With Reactive Extensions, you can convert event subscription to something that looks more object-oriented, but all is still not quite as it should be...

In my series about Dependency Injection and events, you previously saw how to let a third party connect publisher and subscriber. I think that approach is valuable in all those cases where you connect publishers and subscribers in a narrow and well-defined area of your application, such as in the Composition Root. However, it's not a good fit if you need to connect and disconnect publishers and subscribers throughout your application's code base.

This article examines alternatives based on Reactive Extensions.

Re-design #

For the rest of this article, I'm going to assume that you're in a position where you can change the design - particularly the design of the IDependency interface. If not, you can always convert a standard .NET event into the appropriate IObservable<T>.

Using small iterations, you can first make IDependency implement IObservable<Unit>:

public interface IDependency : IObservable<Unit>
{
    event EventHandler ItHappened;
}

This enables you to change the implementation of NeedyClass to subscribe to IObservable<Unit> instead of the ItHappened event:

public class NeedyClass : IObserver<Unit>, IDisposable
{
    private readonly IDisposable subscription;
 
    public NeedyClass(IDependency dependency)
    {
        if (dependency == null)
            throw new ArgumentNullException("dependency");
 
        this.subscription = dependency.Subscribe(this);
    }
 
    public void OnCompleted()
    {
    }
 
    public void OnError(Exception error)
    {
    }
 
    public void OnNext(Unit value)
    {
        // Handle event here
    }
 
    public void Dispose()
    {
        this.subscription.Dispose();
    }
}

Because IObservable<T>.Subscribe returns an IDisposable, NeedyClass still needs to store that object in a field and dipose of it when it's done, which also means that it must implement IDisposable itself. Thus, you seem to be no better off than with Constructor Subscription. Actually, you're slightly worse off, because now NeedyClass gained three new public methods (OnCompleted, OnError, and OnNext), compared to a single public method with Constructor Subscription.

What's even worse is that you're still violating Nikola Malovic's 4th law of IoC: Injection Constructors should perform no work.

This doesn't seem promising.

Simplification #

While you seem to have gained little from introducing Reactive Extensions, at least you can simplify IDependency a bit. No classes should have to subscribe to the old-fashioned .NET event, so you can remove that from IDependency:

public interface IDependency : IObservable<Unit>
{
}

That leaves IDependency degenerate, so you might as well dispense entirely with it and let NeedyClass subscribe directly to IObservable<Unit>:

public class NeedyClass : IObserver<Unit>, IDisposable
{
    private readonly IDisposable subscription;
 
    public NeedyClass(IObservable<Unit> dependency)
    {
        if (dependency == null)
            throw new ArgumentNullException("dependency");
 
        this.subscription = dependency.Subscribe(this);
    }
 
    public void OnCompleted()
    {
    }
 
    public void OnError(Exception error)
    {
    }
 
    public void OnNext(Unit value)
    {
        // Handle event here
    }
 
    public void Dispose()
    {
        this.subscription.Dispose();
    }
}

At least you got rid of a custom type in favor of a well-known abstraction, so that will have to count for something.

Injection disconnect #

If you refer back to the discussion about Constructor Subscription, you may recall an inkling that NeedyClass requests the wrong type of dependency via the constructor. If it's saving an IDisposable as its class field, then why is it requesting an IObservable<Unit>? Shouldn't it simply request an IDisposable?

This sounds promising, but in the end turns out to be a false lead. Still, this actually compiles:

public class NeedyClass : IObserver<Unit>, IDisposable
{
    private readonly IDisposable subscription;
 
    public NeedyClass(IDisposable subscription)
    {
        if (subscription == null)
            throw new ArgumentNullException("subscription");
 
        this.subscription = subscription;
    }
 
    public void OnCompleted()
    {
    }
 
    public void OnError(Exception error)
    {
    }
 
    public void OnNext(Unit value)
    {
        // Handle event here
    }
 
    public void Dispose()
    {
        this.subscription.Dispose();
    }
}

The problem with this is that while it compiles, it doesn't work. Considering the implementation, you should also be suspicious: it's basically a degenerate Decorator of IDisposable. That subscription field doesn't seem to add anything to NeedyClass...

Examining why it doesn't work should be enlightening, though. A third party can attempt to connect NeedyClass with the observable. One attempt might look like this:

var observable = new FakyDependency();
var nc = new NeedyClass(observable.Subscribe());

However, this doesn't work, because that overload of the Subscribe method only exists to evaluate an event stream for side effects. The overload you'd need is the Subscribe(IObserver<T>) overload. However, the IObserver<Unit> you'd like to supply is an instance of NeedyClass, but you can't supply an instance of NeedyClass before you've supplied an IDisposable to it (a Catch-22)!

Third-party Connect #

Once more, this suggests that NeedyClass really shouldn't have a dependency in order to react to events. You can simply let it be a stand-alone implementation of IObserver<Unit>:

public class NeedyClass : IObserver<Unit>
{
    public void OnCompleted()
    {
    }
 
    public void OnError(Exception error)
    {
    }
 
    public void OnNext(Unit value)
    {
        // Handle event here
    }
}

Once again, you have a nice, stand-alone class you can connect to a publisher:

var nc = new NeedyClass();
var subscription = observable.Subscribe(nc);

That pretty much puts you back at the Third-party Connect solution, so that didn't seem to buy you much.

(Preliminary) conclusion #

So far, using Reactive Extensions instead of .NET events seems to have provided little value. You're able to replace a custom IDependency interface with a BCL interface, and that's always good. Apart from that, there seems to be little value to be gained from Reactive Extensions in this scenario.

The greatest advantage gained so far is that, hopefully you've learned something. You've learned (through two refactoring attempts) that NeedyClass isn't needy at all, and should not have a dependency in order to react to events.

The last advantage gained by using Reactive Extensions may seem like a small thing, but actually turns out to the most important of them all: by using IObservable<T>/IObserver<T> instead of .NET events, you've converted your code to work with objects. You know, .NET events are not real objects, so they can't be passed around, but IObservable<T> and IObserver<T> instances are real objects. This means that now that you know that NeedyClass shouldn't have a dependency, perhaps some other class should. Remember what I originally said about Inversion of Inversion of Control? In the next article in the series, I'll address that issue, and arrive at a more idiomatic DI solution.

Instead of using Constructor Injection to subscribe to events on a dependency, you can let a third party connect the subscriber to the publisher.

In the previous article in my series about Dependency Injection and events, you saw an example of how injecting a dependency that raises events violates Nikola Malovic's 4th law of IoC: Injection Constructors should perform no work.

In this article, you'll see the first of several alternatives.

Third-party Connect #

Take a step back and recall why you're using Dependency Injection in the first place. Hopefully, you use Dependency Injection because it provides the decoupling necessary to make your code more maintainable (and thus, you and your colleagues more productive). However, events are already a mechanism for decoupling. In .NET, events are simply a (limited) baked-in implementation of the Observer pattern.

Perhaps it's helpful if we consider alternative names for Dependency Injection. Nat Pryce prefers the term Third-party Connect, and I think that there's much sense in that name. Instead of focusing on the injection part, or even Inversion of Control, the term Third-party Connect focuses on the fact that when you decouple two objects, you need a third party to connect them with each other. In a well-designed application, this third party will often be the Composition Root.

If you already have a third party connecting NeedyClass with IDependency, must you use Constructor Injection?

Further decoupling #

Apparently, if you consider the code in the previous article, NeedyClass is required to do something whenever a particular IDependency instance raises its ItHappened event. What if, instead of injecting IDependency and subscribing a private event handler, you were to expose a public method that implements the same logic as that private event handler?

public class NeedyClass
{
    public void DoSomethingInteresting()
    {
        // Handle event here
    }
}

Notice how much simpler this implementation is, compared with the previous version. Nothing is injected, there are no interfaces in play. Such a class is very easy to unit test as well, so I think this looks very promising.

Doesn't this design break encapsulation? Not more than before. Remember, in the previous implementation, you could always inject an implementation of IDependency that enabled you to raise the ItHappened event, and thereby invoke the private event handler in NeedyClass. This new design just makes it a bit easier to invoke the method. Keep in mind that encapsulation isn't about public versus private members; encapsulation is about invariants.

This version of NeedyClass doesn't actually expose a public 'event handler', since the DoSomethingInteresting method doesn't match the event handler signature. Thus, if you use a static code analysis tool, it's not going to complain about a public event handler. The DoSomethingInteresting method is a method just like any other method. This design is much more decoupled than before, because NeedyClass knows nothing about IDependency. Hopefully, it makes sense as a stand-alone class with its own API. This makes it more reusable.

At this point, NeedyClass is no longer an appropriate name, but let's keep it for now.

Subscription #

In order to connect NeedyClass with IDependency, the third party (e.g. the Composition Root) can subscribe the DoSomethingInteresting method to the ItHappened event:

var nc = new NeedyClass();
dependency.ItHappened += (s, e) => nc.DoSomethingInteresting();

The advantage of this design is that it's much more decoupled than before. NeedyClass knows nothing about IDependency, and IDependency knows nothing about NeedyClass. However, one disadvantage is that it's easy to forget to connect these two objects with each other; the compiler no longer offers any help.

If both objects are long-lived objects (i.e. have the Singleton lifetime style), you probably only need to write and invoke that connection code once, in the Composition Root. In this case, one or two Smoke Tests should prevent any regressions.

However, if one or both of these objects have shorter lifetimes, you may want to encapsulate the creation of NeedyClass in some sort of Factory. Still, unless you make the NeedyClass constructor internal or private, programmers may forget to use the Factory, so this can still be a fragile solution.

Unsubscription #

Another problem that this Third-party Connect solution shares with Constructor Subscription is that you still need to think about disconnecting the two objects, once you're done with them. This isn't hard to do.

EventHandler handler = (s, e) => nc.DoSomethingInteresting();
dependency.ItHappened += handler;
 
// Do something interesting here
 
dependency.ItHappened -= handler;

While not particularly difficult, it does require you to take the extra step of assigning the event handler to a variable you can later use to unsubscribe. Still, the worst part of this attempt at a solution is probably that, just like you'll need to remember to subscribe NeedyClass to the event, you must also remember to unsubscribe it. At least, in this case, there's a better symmetry, because you must remember to both subscribe and unsubscribe, whereas with Constructor Subscription, you only needed to remember to unsubscribe (or dispose, as it were).

Conclusion #

Using Third-party Connect leads to a simpler, but more Fragile design. Still, I often find that the extreme simplicity of the involved classes trumps the fragility of the design; if I had to choose between Third-party Connect and Constructor Subscription, I'd select Third-party Connect. Fortunately, these aren't the only options available to you; in future articles, I'll approach a better alternative.

Using Constructor Injection, you can subscribe to an event within the constructor, but should you?

This post is the first in a series of articles about Dependency Injection and events.

Imagine that you have an interface that defines an event:

public interface IDependency
{
    event EventHandler ItHappened;
}

In order to keep the example simple, the ItHappened event carries no information; it just raises an event with an EventArgs instance. Thus, subscribers can react to the fact that this particular event happened, but there's no data contained in the event. However, the following discussion doesn't change if the event carries information.

A class must react to the events raised by IDependency implementations. A common approach is to subscribe to the event in the constructor. We can call this pattern Constructor Subscription:

public class NeedyClass
{
    private readonly IDependency dependency;
 
    public NeedyClass(IDependency dependency)
    {
        if (dependency == null)
            throw new ArgumentNullException("dependency");
 
        this.dependency = dependency;
        this.dependency.ItHappened += this.OnItHappened;
    }
 
    private void OnItHappened(object sender, EventArgs e)
    {
        // Handle event here
    }
}

The concern is that, by subscribing to an event, the constructor violates Nikola Malovic's 4th law of IoC: Injection Constructors should perform no work.

This rule about Dependency Injection explains why you can compose big object graphs with confidence. Still, the most compelling reason for conforming strictly to the rule is most likely performance considerations. So, what if you subscribe to a single event or two during construction? Will it adversely (and noticeably) affect performance of your overall system?

As always, the answer to such questions is: measure. However, I'd be quite surprised if it turns out that a single event subscription has a huge impact on performance...

Smell #

Consider the implementation of NeedyClass: it contains a design smell. Can you spot it?

Given the definition of the IDependency interface, the ItHappened event is the only member defined by the interface. If this is the case, then why is NeedyClass holding on to a reference of the interface?

You can remove the dependency field from NeedyClass, and nothing is going to break:

public class NeedyClass
{
    public NeedyClass(IDependency dependency)
    {
        if (dependency == null)
            throw new ArgumentNullException("dependency");
 
        dependency.ItHappened += this.OnItHappened;
    }
 
    private void OnItHappened(object sender, EventArgs e)
    {
        // Handle event here
    }
}

From the perspective of an outside observer, that's really strange. Why are we required to pass in an object that's not going to be used as is? Just like in a previous discussion about the implications of injecting IEnumerable<T>, if you're injecting an abstraction, and the constructor then starts querying, modifying, examining, or otherwise fidget with the injected object, then you're probably injecting the wrong object.

Keep in mind that Constructor Injection is a declaration of requirements. It's the declaring class that advertises to the world: these are (some of) my invariants. If the class can't use the injected dependency as is, it suggests that it requested the wrong object from its clients. The class with the Injection Constructor should declare what it really needs. In this case, it needs to react to events.

In the next post, I'll show you a better (i.e. simpler) design for reacting to events.

Unsubscription (Update, September 9, 2013, 11:48 UTC) #

Some readers have commented that NeedyClass must keep the reference to IDependency in order to unsubscribe from the event. This is true, and something I overlooked when I originally banged together the sample code yesterday evening. Apparently, three unit tests were at least a test too few... :$

From a perspective of basic correctness, NeedyClass works appropriately, but as Geert van Horrik and Claus Nielsen point out, this could easily lead to resource leaks (although in practice, that depends on the effective lifetime of NeedyClass).

What happens when we start taking resource management into account?

Well, NeedyClass must be sure to unsubscribe from the event when it goes out of scope. The most correct way of making sure this happens is to implement IDisposable:

public class NeedyClass : IDisposable
{
    private readonly IDependency dependency;
 
    public NeedyClass(IDependency dependency)
    {
        if (dependency == null)
            throw new ArgumentNullException("dependency");
 
        this.dependency = dependency;
        this.dependency.ItHappened += this.OnItHappened;
    }
 
    private void OnItHappened(object sender, EventArgs e)
    {
        // Handle event here
    }
 
    public void Dispose()
    {
        this.dependency.ItHappened -= this.OnItHappened;
    }
}

While this weakens the argument that NeedyClass is requesting the wrong kind of dependency, this is a lot of work just to consume an event. Furthermore, it isn't particularly safe, because you'll have to remember to dispose of all of your NeedyClass instances.

Thus, a better approach is still warranted.

Comments

Overview of articles about Dependency Injection and events.

One of my readers, Kenny Pflug, asks me about subscribing to events in the constructor of a class using Constructor Injection:

I want to register to a plain old .NET event of a dependency that is supplied via constructor injection. I read your book "DI in .NET" and in it you mention that injection constructors should be simple and not perform any other work than guarding for null values and assigning the dependencies to (read-only) fields. I also read on your blog that this is only a pattern and might be broken gently if the need would arise.

[...] do you know and prefer any patterns to solve this "registering to events of dependencies in a constructor" problem?

In short, I think that it's a code smell if a design requires you to subscribe to an event in the constructor; it's a smell that the dependency chain is slightly inverted. Inversion of Inversion of Control - how about that?

In the next series of posts, I'll attempt to provide various perspectives on this situation, starting with the original context.

• DI and events: Constructor Subscription.
• DI and events: Third-party Connect.
• DI and events: Reactive Extensions.
• DI and events: Composition.

In summary, it's probably not going to be a big problem if you subscribe to an event in a constructor, but it's a code smell, because it betrays a design issue.