The Art of Unit Testing

After I read Tim Barcz’s review of Roy Osherove’s “The Art of Unit Testing” and I knew I had to get a copy right away. It just arrived and I read it in one sitting.  I am so pleased that I did. I’ll quarrel with it … but do not let that deter you from rushing to buy your own copy.

Let me say that again. I highly recommend this book – five stars -, especially to folks like me who are not deep into unit testing. This review is full of my grumpy disagreements. That’s how I engage with a good book. Don’t be dissuaded.

Warning: Long post ahead. The short of it: buy the book. Everything else is commentary.

There is no point in recapping the book’s main points as Tim Barcz did that for us. I’m coming at it from a different angle. I’m coming at it from the angle of a guy who wishes he wrote more tests, wishes he was good at testing, even wishes he practiced (or at least gave a serious, sustained effort at trying) TDD. A guy who doesn’t.

A guy much like the vast multitude of developers out there … who is embarrassed by being “old school”, is looking for an opportunity to catch up, but isn’t going to take crap from an obnoxious TDD fan-boy.

I’ve had plenty of success over the years, thank you very much. I’ve written good programs (and bad) that still work. And I can mop the floor with legions of developers who think TDD/BDD/WTF experience yields greatness. They remind me of newly minted MBAs who believe with unshakeable certainty that they’re entitled to a management position. Think again.

Do I sound defensive? Yup. Enough already. My point is this ...

One of Roy’s goals is to reach people like me. We’re experienced developers who may have mucked around with unit testing but aren’t doing it regularly and may have had some rough experiences. We believe … but we don’t practice. Can he do something for us that makes us want to try again or try harder. Can he keep it simple and approachable and be respectful and non-dogmatic.

Yes he can.

He extends olive branches aplenty throughout. Out the gate he writes: “One of the biggest failed projects I worked on had unit tests. … The project was a miserable failure because we let the tests we wrote do more harm than good.” 

Thank you. I don’t believe it for a second. Oh I believe the tests were every bit as unmaintainable. I’m just not buying that the project failed because of the tests. They contributed perhaps, but in my experience, projects fail for other deeper reasons. That, however, is another post.

What I applaud is that he opens empathetically. He goes straight to the dark heart of our limited test-mania experience: when brittle, inscrutable tests became so onerous that they had to be abandoned. Been there. Seen it several times.

I appreciate that a similarly open and self-critical sensibility shines throughout. I’m particularly fond of the section on alternatives to “Design for Testability” in appendix A. There he notes that the uncomfortable coding-style changes required to support testing are an artifact of the statically typed languages we use today.  “The main problems with non-testable designs is their inability to replace dependencies at runtime. That’s why we need to create interfaces, make methods, virtual, and do many other related things. [266]”

Dynamic languages, for example, don’t require such gymnastics. Perhaps with better tools and language extensions (Aspect Oriented Programming comes to mind) we can make testing easier for the statically typed languages.

Here he acknowledges that testing is just too darned hard, harder than it should be, and this difficulty – not resistance to new-ish ideas by crusty old farts like me – is a genuine obstacle.

Until then, we have to accept that incorporating unit testing in our practice requires more than an act of will. You will need hard won skills and experience and you will have to contort your code to get the benefits of unit testing. This is not your fault. You will pay a bigger price than you should have to pay. It may be rational to say “I can’t pay that price today, on this project.”

It may be rational. It may also be wrong. In any case, Roy’s goal is to reduce that price as best he can (a) with a progressive curriculum yielding skills you can use at each step and (b) by introducing you to tools that cover for language deficiencies.

Roy succeeds for me on both fronts. Each step was a small enough to grasp and big enough to be useful. The tools survey was thin … but at least he has one – with opinions – that gives you places to look and an appreciation of their place in a complete testing regime.

Part 1 - Basics

This part is so important for readers like me. Overall, I thought it was grand. I’m about to freak out about a few of Roy’s choices but before I do I want to say “(mostly) well done!”

My biggest disappointment is Roy’s scant mention of IoC. There is brief treatment of Dependency Injection [62-64] and a listing of IoC offerings in the appendix. That’s it. There is not a single example of IoC usage.

Testing is one of the primary justifications for using IoC. Such short shrift could leave the reader wondering what all the fuss is about. Wrongly, in my opinion. I was really looking forward to guidance on proper use of IoC in unit testing.

The omission felt consequential in Roy’s discussion of test super classes [152ff] where he takes a couple of classes that do logging and refactors their test classe to derive from a BaseTestClass [155] whose only contribution derived classes is its StubLogger. What a waste of inheritance. Injecting a logger is the IoC equivalent of “Hello, World”. What am I missing?

I realize (from painful experience) that it’s easy to create an IoC configuration rat’s nest in your test environment. That’s why I was hoping Roy would propose some best practices. Instead, I believe we are served an anti-pattern.

I must also say I was shocked to see favorable mention of using compiler directives [79 – 80]. He urges caution; I would ban the technique outright.

I was not fond of Roy’s preference for the AAA (Arrange-act-assert) style of test coding. This style facilitates brittle tests because it brings the “arrange” moment into the test class and this has been a source of trouble for me.

“Arrange” code is distracting and bloats the test, making it too hard to see what is going on and leading to test methods that do too many things at once. When I was using this style, I couldn’t stop putting multiple asserts in each method [a “no-no” discussed 199-205]; it was too painful to make separate methods.

His associated test naming convention tends to say more about how the test works than what it is trying to achieve … and I think it is easier to find and understand tests when the names express intent.

Since I adopted more of the Context/Specification style espoused by BDD fans (see, for example, Dan North’s 2006 essay and a more recent manifesto by Scott Bellware), I’ve written smaller tests that are easier to read and easier to maintain. Roy can’t be faulted too much for this; Context/Specification is starting to take hold only this year (2009) and we don’t have the years of experience that go with AAA.

Two caveats: As I made clear at the beginning, I don’t do enough unit testing to be taken seriously as a guide. Second, test regimes falter in year #2 as the long term maintenance of actually-existing-unit-test-implementations overwhelm the development effort; that’s why Roy’s book is important. But the Context/Specification style hasn’t been around long enough to prove its worth in the field. It will take a couple of years to find out.

Part 2 – Core Techniques

Discussion of difference between Stubs and Mocks was brilliant. "If it’s used to check an interaction (asserted against), it’s a mock object. Otherwise , it’s a stub” [90]

Loved that he handwrote mocks before introducing mocking frameworks (he prefers to call them “Isolation Frameworks”). This is a crucial pedagogical move. Many of us are stunned by the mocking framework syntax (e.g., Rhino Mocks) and our instinct is to run away and only use state-based testing.

Those of you who know better will smile knowingly as I confess to the awful mess I made for myself by hand rolling my own mocks for fear of frameworks. There is a reason and it is the sheer ugliness of mocking framework APIs.

Roy gets it. That’s why he sneaks up on Rhino Mocks.

“One Mock Per Test” [94]. I like the sound of it. I like Roy’s reasoning. It’s the kind of clear, unambiguous advice that novices like me need. I’m sure there are times when it is smart to set it aside but it has the whiff of hard-earned wisdom.

I much appreciated the “traps to avoid” section at the end of the Mocking chapter 5. It’s easy to say “if it looks complicated, stop”. We should say it again anyway. Roy goes one better and identifies the tell-tale signs of too much mock framework fascination.

Part 3 – Test Code

I tend to agree with Tim Barcz: Chapter 7, “The Pillars of good tests” is essential and some of it feels like it belongs early in the book … not here, 100 pages in. On the other hand, the reader isn’t ready for a review in depth of test smells and maintainability until they know the basics. On balance, the timing of this chapter feels right.

The passages on “trustworthy tests” overflow with good sense. How to fix a broken test … which includes breaking the production code to ensure the test still catches the failure … that’s a step you overlook at your peril.

It’s proof again, if proof is needed, that you don’t automate unit tests and you can’t do it by rote. Junk testing is hardly better than no testing … and Roy has an iron grip on this fact.

Chapter 6 concerns build automation, code organization, and conventions … crucial blocking and tackling.

This is the place I mentioned earlier at which Roy speaks favorably of test class inheritance where I feel IoC techniques are more appropriate. I don’t think much of overriding a virtual setup method either; I think the Template Pattern is much preferred. With Template Pattern  – in which derived classes override an empty virtual method that is called by the base class – you ensure that base behavior is always invoked and you don’t trouble the developer with knowing when the base method should be called.

Roy describes something he calls the “Test Template Pattern” [158] which sounds like Template Pattern but isn’t. His Test Template Pattern consists of abstract test methods which, perforce, must be implemented by derived test classes. The intention is to ensure that all derived test classes implement specific tests – not, as in Template Pattern, to provide a well-managed base class extension point.

The Context/Specification approach employs the Template Pattern (in the form of a virtual Context() method) as the preferred means by which a derived Specification class makes arrangements (adds “context”) that are particular to its needs.

Speaking of Context/Specification, if you prefer that style, you’ll need to adjust Roy’s recommendation from “One Test Class Per Class Under Test (CUT)” [149] to “One Test File Per Class Under Test”. That’s because Context/Specification yields many test classes, each dedicated to a different “context” in which the CUT is revealed. It is typical of the examples I’ve seen that these many classes can be found in the same physical file, named after the CUT.

I have a feeling that BDD practitioners go farther and argue that you build tests around scenarios, not classes. They could say that it’s a category mistake to force a correlation between CUTs and test files. I just don’t know. Such correlation seem convenient but it may distort the design process. I lack the experience necessary to weigh the tradeoffs. I wish Roy had explored this avenue.

Part 4 – Design and Process

Chapter 8 is about the politics of implementing a testing regime where none exists, a hugely important topic. I enjoyed this chapter immensely. Unfortunately, Roy is utterly unpersuasive.

To summarize, a team that writes tests takes twice as long to deliver the first implementation as the team that doesn’t [232], there are no studies proving that unit tests improve quality [234] even though we believe it anecdotally, there is strong evidence that programmers who write tests won’t do a good job of testing for bugs despite their best intentions [235], and finally, it appears most defects stem not from poor code quality but rather from misunderstanding the application domain [237] . This litany is not the way to management’s heart.

I will expand on each of these observations.

Time to Market

In the “Tough questions and answers” section Roy’s prepares an answer to the #1 question on your manager’s mind: “How much time will this add to the current process?”

Roy’s frank answer is “it doubles your initial implementation time …” [232]

That’s a conversation stopper. Management prizes an early delivery date and it is extremely difficult for management to distinguish the first implementation from “the” implementation.

Roy hastens to add “… the overall release date for the product may actually be reduced.”

That may re-open the conversation … because you’re talking about the delivery date again. You’re making the case that the project won’t be considered delivered until it passes some quality bar … that the savings in the mandatory testing phase may compensate for the slower start.

The equivocation - “may” – will be noticed. Management has heard too many stories about Total Cost of Ownership and Reduced Maintenance. It’s going to be tough.

Here’s the worst part. It is often true that to be in the lead at the first turn means you win the race. It means you get resource commitments that won’t be available without a (ridiculous) early delivery. This is so even if we finish much later than the conscientious, test driven developers. Too bad, because they never get the shot. And by the time the technology debt comes due, there are sunk costs (real and political) that management will be loathe to abandon.

This is just how it is. So, while I applaud Roy’s honesty, this is a tough sell. He needs another plan. He needs a way to shift the definition of “delivered” to an implementation that passes a measurable quality bar. He needs to talk about short cycles so that the evidence is experienced on this project and registers in management’s short term memory.

Roy shows some grasp of this dynamic. In his example – a tale of two projects – the debugged release time is 26 days in the worst (no-testing) case. You can win a month to prove your point … but not much longer.

Does unit testing improve code quality?

Roy is his typical honest self here. Unfortunately, what he reports is not likely to advance his cause.

He draws proper attention to code coverage. There are lovely charts. There is just one flaw: you have to convince the skeptics that you’re measuring something that matters.

You think that’s a good metric because it measures unit testing activity. The skeptic doesn’t care about your activity. Activity – expenditure of effort – is irrelevant. The skeptic cares about delivering the system that “works acceptably” as quickly as possible. The skeptic suspects your polishing one apple, while he wants many apples, perhaps less polished.

A devastating admission: “There aren’t any specific studies I can point to on whether unit testing helps achieve better code quality.” [234] Ouch! That has to be fixed.

Here’s another groaner: “A study by Glenford Myres showed that developers writing tests were not really looking for bugs, and so found only half to two-thirds of the bugs in an application.” [235]

Here’s another citation that Roy interprets as strengthening the case for unit tests although I think it does the opposite: “A study held by … showed that most defects don’t come from the code itself, but result from miscommunication between people, requirements that keep changing, and a lack of application domain knowledge.”[237]

It is not self-evident how unit testing alleviates these sources of error. The best he can say is that, as you correct course, the unit tests provide some assurance that the other things you still think are true are still tested. That’s valuable … but weak beer at best.

This chapter made him wonder again if I should be so ashamed of my test-less oeuvre.

Nah. We may lack the proof but absence of proof is not proof of absence. Where would we be if we only followed rigorously proven practices? Show me the study that proves “GoTo”s are bad.

There was a prolonged and super-heated argument in the ‘70s and ‘80 about the (de)merits of GoTo. Steve McConnell covers it in an article from his Code Complete where he makes reference to a Ben Shneiderman “literature survey”. I suspect a literature survey would yield comparable support for unit testing. Literature survey’s perhaps reflect the “wisdom of the field”; they are not evidence.

The fact is, we have very little social science on any development practices. The objection that unit testing and TDD are unproven could be raised about almost any practice. The anecdotal support for unit testing remains strong.

We shouldn’t leave it there. We need real studies. I’d like to see some of my former colleagues in economic sociology jump in. There’s at least a masters thesis here.

It’s also possible that the limited studies to which Roy refers (he does not cite them) produce inconclusive results because they don’t account for test quality. Noise from botched test regimes may be hiding the good news. Roy established early that (a) a poor bad unit testing can be worse than no unit testing and (b) it’s easy to make a mess.

If this interpretation is correct, we are challenged to improve testing as actually practiced in the wild. We lose the argument – and we should lose it -  if proper unit testing remains a rare skill, difficult to acquire. If Roy’s book becomes widely read and as more developers learn to write better tests, we could hope for a positive swing in the statistics.

Finally, I’ve heard Steve McConnell claim in a DotNet Rocks Show (0:28)  that “40% to 80% of its effort on unplanned defect correction work … in other words, low quality is the single largest cost driver for the average project.” I don’t know how Steve came by these statistics (and “40% – to 80%” is huge swag). It is Steve’s business to measure and track this stuff. And if you’re doing something that attacks the “single largest cost driver” … and you’re not disproportionately increasing costs with your remedy [!] … then you’re making business sense.

Chapter 9 on testing legacy code is a welcome introduction with good advice … but no substitute for Michael Feather’s Working Effectively with Legacy Code. Feather’s book is expensive ($47 on Amazon); perhaps Roy’s chapter and his enthusiasm for Feather’s book will encourage sales.

Appendices

Appendix “A”, ostensibly about design and testability, is mostly about design for testability. That’s no small leap. Testing code heavily is one thing. It is another to distort your design to satisfy inadequacies in the language that make testing difficult.

I’ve deliberately expressed this point in the most contentious way possible to dramatize the implications of exchanging “for” for “and”.

I hasten to express my enthusiasm for the contribution of “unit testing” to design. Expressing your expectations in code clarifies the design and casts a strong light on otherwise dark edge cases. Many of the test disciplines, loosening dependencies in particular, promote SOLID design principles (especially Single Responsibility) that are beneficial in their own right. Roy is excellent on these points.

The problem is that at least one of Roy’s recommendations, “Make methods virtual by default” [258], reduces design quality in order to make testing easier. Testability and Good Design are at cross purposes.

“Make methods virtual by default” is a terrible idea in my opinion. I explore that opinion in a separate post. My argument in brief is that a virtual method is an invitation to extension everywhere. Extensibility is not a frill. You have doors in your house for a reason; that’s where people are expected to enter. They aren’t expected to come through the windows. You don’t punch orifices into every wall. A plethora of virtual methods invites violation of the Open / Closed Principle (“Liskov Substitution Principle” to be more precise) and makes delivery, maintenance, and support of a system pointlessly more difficult.

This aside, the chapter, although brief, is clear and persuasive.

Appendix “B” enumerates helpful tools and test frameworks. Each merits only a brief blurb but I was pleased to have an annotated list of Roy-approved choices.

Conclusion

This is a wonderful book for the experienced developer who is open to unit testing while having limited experience of it. I suspect it will help technical managers of a certain age … managers who’s programming days are behind them, who’ve heard the fuss, been through a few fads, and want a serious, honest, warts-and-all look at unit testing.

I’m told also that it has earned the respect and admiration of many with deep unit testing experience. That’s a confidence builder for me.

Get it.

Drag & Drop Debate on Herding Code

I just listened to a Herding Code podcast in which “Drag and Drop” development was attacked and defended (thanks to Tim Heuer for the pointer).  That wasn’t the offical topic but there was a golden 10 minutes on this subject about 35 minutes into the program.

I don’t believe they do transcripts of Herding Code podcasts. With apologies to those guys, I transcribed those 10 minutes for your delectation. Go hear the whole thing.

Speakers

The guests (and bio extracts from their web sites) are

G. Andrew Duthie (GAD): ”G. Andrew Duthie, aka .net DEvHammer, is the Developer Evangelist for Microsoft’s Mid-Atlantic States district, where he provides support and education for developers working with the .net development platform.”

Alan Stevens (AS): “Alan Stevens is a ... software artisan living in Knoxville, TN. Alan is an Open Space Technology facilitator. Alan is a Microsoft Most Valuable Professional (MVP) in C#. Alan is a member of ASP Insiders”

The Exchange

GAD: Alan likes to beat me up on Drag and Drop (d&d) development … and I’ll be the first to say that I’ve done my fair share of d&d demos and some of the code I write I use d&d and I’m not ashamed to admit it.

AS: Will you please promise here now never to do that again!

GAD: No, I won’t actually. Here’s my take on the whole drag and drop thing. … [I enjoy] the concept of “Technical Debt” and d&d is an example of something that can lead to Technical Debt and I’m perfectly willing to accept some of that Technical Debt and sometimes the reality is that I pay for it later …

AS: No!, No. _I_ pay for it later. You go on and do a demo of something else that will be released next year. And I have to clean up behind these poor slobs that all they know about .NET development is what you showed them in the PowerPoint deck and the d&d demo … and they don’t have a _clue_ what’s going on. And it’s just _garbage_. I walk into these steaming piles of poo that are called “business applications” and have to clean this mess up. You just aren’t doing anybody any good by making them ignorant.

Unknown_1: So what you’re saying, Alan, is that your upset that MS is providing you with a job.

AS: I would say that there are other skills that my clients could exploit

GAD: So let me give you my take on this. This is a place where I think you and I disagree pretty vehemently.

AS: Yeah, you’re wrong!

GAD: My take is that if we are out there doing demos of any rapid application development feature that has the potential for accruing technical debt that we probably ought to be saying that. … So … If all that we’re doing is teaching people how to … spin the knobs … we’re not teaching them enough, but at the same time I think we have an obligation … to show off those new features. I’d love to see us do a better job of giving the caveats and letting people know that “this demo that I’m building here is not based on any patterns so … you don’t want to … emulate this.” … Pete Brown says “as I’m building this keep in mind that this is demo code. Ideally, when you’re moving into an application, I show you a pattern you can use but right now, this code is really to demonstrate this feature and I don’t want to clutter it up with the conversation around the pattern _yet_. ” Eventually he gets to MVVM …

AS: I think caveats of “here there be dragons” are fantastic. I just haven’t seen them yet at a launch event. It seems like there is a fear of criticizing the product, that it’s not super easy to use .. if there’s a trend toward doing that [caveats], I’m 100% for that because that’s what people need to get in their heads, this is not a model of how you should do your real world development. I’m trying to show you features.

The other aspect of d&d demos that gets under my skin is “I didn’t write a single line of code”. What the hell is wrong with writing code? … Why should dragging something from the toolbox be a better experience than actually writing it in the editor?

GAD: I’ll give you an example. I don’t ever want to have to write a login dialog again in my life. I don’t want to have to write that code. So when we brought out ASP.NET v. 2 and VS 2005 and you could just drag a login control and if you had the membership service provider set up, you were good to go. That’s a big win to me and I don’t see any downside to that particularly given that if you have to change out your membership provider … you don’t have to change anything.

AS: … I agree that I don’t want to write plumbing code every time. But I don’t mind writing code to re-use a component

GAD: So, ultimately, if you have that code somewhere and you can drag it into the code, are you guilty of d&d development too?

AS (incredulous): Why would I ever do that?

GAD (frustrated): So do you object to controls?

AS: I’m leaving visual designers out of this. I’m talking about _non-visual_ components where you create a visual designer for no reason. Why would I drag that code out of somewhere? Why wouldn’t I configure my IoC container to inject that component and then just use it in my code? Intellisense is a wonderful thing. I don’t need some graphic on the screen; I don’t need to reach for the mouse to add these things into my application.

GAD: I think you’re getting into a philosophical disagreement …

AS: No, no, no! You are still wrong.

GAD: … Trying to make a “one size fits all” statement about how development should be done, I don’t see that working in our industry. There are many different ways that you can successfully build software …

AS: Let me leave it at this. I’ve never seen it work. I’ve never seen it work in a long term, maintainable sense. Only in the initial release. And, honestly, the initial release is not where the cost of any application lies. It is always down the line, in extending and maintaining the application. …

Kevin Dente: … Not only do I never see it working, I never use it in my work. And whenever I hear Microsoft say “We have resource constraints so we have to make judgment calls” well when tons of effort is poured into those things for which I think have no value and aren’t put into places which have a lot of value, I get very frustrated.

“And the Winner Is …”

My heart is with Andrew (GAD). I do a lot of demos myself both of Microsoft technology and of our product. I appreciate the challenge of demonstrating features such that they can be seen without scaffolding … scaffolding that proper patterns may introduce.  I love IoC and wouldn’t develop without it. I confess I can’t do a demo with it … today … and still reach my audience.

But my head says Alan (AS) is dead-on right.

This is not a “philosophical disagreement”. The drag-and-drop approach in which you populate your views with non-visual controls is always wrong outside of a demo. There is no “choice” here. The d&d proponents are “flat earthers”, pretending that this isn’t a settled fact.

How can I be so sure? Let me offer the same reason that others brought up. If these components had a legitimate place in my application, I would use them. If I would always yank them out, they have no good reason to be there in the first place. And if it’s not good enough for my code, why would it be good enough for my customer’s code?

Think about this. When you never see these controls in good code and you often see them in bad code, what must you conclude?

Note that AS and GAD do not disagree about “good” and “bad” code in this regard. It isn’t as if reasonable minds differ on this. GAD never says that coding with these components is ever good. He admits straight-up that their use incurs “technical debt”.

And by the way, GAD, these components don’t “have the potential to incur technical debt”. There is no “potential”. They are always technical debt. The only question is when you pay for it.

All of us understand that we incur “technical debt”. It’s an essential fact of development. But we don’t take on debt frivolously. And these d&d components are frivolous choices.

We know from our own experience that it is neither easier nor faster to use these components than to do it right the first time. Once you know the proper way, d&d saves you zero effort. I say this assuming that you employ some amount of decent structure the first time … that your app is not a total hack job from day one to first release. Would any of us do that?

No, the only place for d&d non-visual components is in demos. MS is properly faulted for failing to warn people. They should be faulted for wasting resources on these components and their visual designers … resources that could otherwise be deployed to improve the platform and the experience.

Today, my poster child for waste happens to be the RIA Services DomainDataSource which mimics the comparably atrocious ASP ObjectDataSource.

And guess what? Because Microsoft is doing it, we at IdeaBlade feel that we have to do it too. That’s right … we’re forced to write our own Silverlight ObjectDataSource just so we won’t lose customers to a Microsoft demo. What a frigging waste all around.

Our customers suffer doubly. They are encouraged to develop poorly and they are deprived of means and guidance to develop well. What a shame.

Screen 54, Where Are You?

The justly renowned Jeremy Miller just issued a call for real-world “screen activation” scenarios.

Jeremy is writing a book on UI design and development patterns. A number of us who think and pontificate about this a lot (including Glenn Block, Rob Eisenberg, John Papa, Shawn Wildermuth, and many, many others) are helping Jeremy by chipping in with our favored approaches and examples drawn from our own experiences and our customers’ experiences.

I’m pretty active in this area as you, dear reader, know well. I hope you’ll take a moment to (a) look at Jeremy’s agenda and (b) contribute your thoughts – either directly to his post or to mine.

I’ll see that he gets them. And I’ll highlight with my own commentary some that pique my interest. All of it, from all sources, will help us at IdeaBlade do a better job of providing you with supportive infrastructure and guidance.

--

Footnote: You are all too young to remember the early 60’s TV hit that inspires this title: “Car 54, Where Are You” with the late Fred Gwynne, also of “The Munsters” and “My Cousin Vinny” fame.

Entity Framework N-Tier Anti-Patterns and DevForce

Danny Simmons, dev manager for Microsoft’s Entity Framework, wrote an article published in the June MSDN Magazine called “Entity Framework: Anti-Patterns To Avoid In N-Tier Applications”. It’s a wonderful place to discover some of the pitfalls of developing your own n-Tier infrastructure. You will learn that such development won’t come cheap. My take-away: “don’t be foolish; don’t write this yourself.” Of course fools abound and you’re welcome to join them.

I arrive at the same conclusion from reading Julie Lerman’s essential book on Programming Entity Framework.

Danny could have shouted: “Don’t do that!”  It would just piss you off. Instead, he begins as if this were something you might do and he’s only going to guide you in doing it: “I will try to set a foundation on which you can build for success in this part of your applications”. Yeah, right.

I encourage you to read his article. Look at the six “anti-patterns” he identifies – the shortcuts that you are likely to take in your naive attempt. See for yourself not only the deep and dangerous traps … but also the sheer difficulty and complexity of avoiding those traps.

Go ahead, read it. I’ll be here when you get back.

So what do you do now? You have an application that crosses tiers. I know you do. If it’s a Silverlight application, you know you do.

My answer … wait for it … is DevForce. We solved this one seven years ago and have been enriching our product over many years. Try it and you will be relieved of all of Danny’s anti-patterns (and all of the gyrations described by Julie too). I’m not suggesting it addresses every n-Tier problem; nothing does. But you won’t succumb to these traps and you won’t waste your time – or your employer’s time – fighting your way through the cross-tier application jungle.

I cannot close without just a bit of criticism. I can’t give all of Danny’s anti-patterns equal weight and I would add some different ones of my own. In one case, I just think he’s wrong. 

Under the heading “Two Tiers Pretending to be Three”, he seems to say that it is imprudent to attempt cross-tier queries and updates.  He is literally saying that he thinks it is improper for the Entity Framework to do so but, in the context of the rest of the article, you are lead to believe that is imprudent for you to do so.

It is certainly difficult. But we still have to do it in some fashion or another. That’s why there is an ADO.NET Data Services and a RIA Services. These technologies, in their somewhat crippled way, facilitate client side queries and updates. DevForce does an even better job. My point is that they fill a need and they are capable of doing so without wrecking your application.

Danny distorts the original question. He mutates from “Why can't you make the Entity Framework serialize queries across tiers?” to “Why not just expose the database directly?” This is not what people are asking for. They want to be able to compose a query on the client – compose it in entity terms, not database terms – and send that query over the wire in the full expectation that they will get entities back.

Of course he is right to caution against “introducing a mid-tier that is simply a thin proxy for the database”? But this warning, like the others, is just more of the same advice: do a good job of writing the middle-tier – don’t hack it together – and it’s going to be hard. Point taken. Use DevForce … or RIA Services or CSLA if you prefer.

Danny tries to predicate his argument on Fowler’s “First Law of Distributed Object”. That won’t fly. Worse, Danny, like many others, misrepresents Fowler’s “law”. Let’s clear that up.

The short statement of the law is “Don’t distribute your objects”. What the heck does that mean? You’ll have to look at the chapter in Patterns of Enterprise Architecture (highly recommended) … or read the article he wrote in Dr. Dobbs that covers the same ground. You will see that it is a jeremiad against calling methods on a remote object.

Fine. I get it. Don’t call Employee.CalculateSalary() and have that be a cross-tier call. Fowler is inveighing against a design which would present the fiction of an object instance that is half on one tier and half on another. It is distributed in the sense that it has one foot on the server and another on the client. Don’t do that. I agree.

But we are interested in an entirely different notion of “distributed objects”. The notion I have in mind is more akin to “mobile objects”. 

When we query across tiers we intend to retrieve persistent state and reconstitute an object with that state on the client. That object can have behavior to go with the state; it should have behavior.

Make no mistake. We are executing locally on the client when we invoke CalculateSalary() on it; there is no ambiguity about this at all. Now it also may be (as it is in DevForce and RIA Services and CSLA) that the same type is available on the middle tier; that means we can reconstitute an instance with persistent state there – on the server – and invoke Calculate Salary() on it there – on the server. There is no ambiguity in this case either. The instance exists on a single tier and the method executes on the same tier as the instance. No instance straddles tiers … or even pretends to straddle tiers. No instance violates Fowler’s “law”.

This is what we expect from a cross-tier query. It is a perfectly reasonable expectation. Danny has announced that Entity Framework will not support that kind of query. That’s ok too. That’s a choice. Let’s not dress it up into some kind of principle.

My rant is over. Please read what Danny has to say … in this article and elsewhere. You will always be glad that you did.

DevForce Predicate Builder

One in the “DevForce is the Shiznit” series of boastful posts about our product in which I describe a cool feature that may even interest those who have yet to discover the wonders of DevForce.

The time comes when you want to construct a LINQ “Where” clause programmatically. It should be easy. It turns out to be more challenging … until you use the DevForce PredicateBuilder.

The DevForce PredicateBuilder shares a common purpose and name with the Albahari brothers’ PredicateBuilder described here and here. I would be remiss if I failed to note their good work and inspiration. I’ll cover important differences in our solutions towards the end of this post.

February 2011 Update:

Much has happened since this post was written. PredicateBuilder has been joined by a number of related components such as PredicateDescription and SortDescription. Learn more at the DevForce Resource Center (DRC).

Intersoft is also introducing their UXGridView as I write. UXGridView can be bound to their QueryDescriptor component in the ViewModel that brings QBE functionality to the grid. The QueryDescriptor is backed by a DevForce "Data Provider" that uses these features. Bill Gower wrote a tutorial about it.

Imagine a product search interface. The user can enter words in a “Name Search” text box. Your program should find and display every product that contains any of the words entered by the user. You don’t know how many words the user might enter. What do you do?

The solution would be easy if you knew the user would enter exactly one word.

Let’s illustrate with the Northwind database. Assume “Manager” is some apparatus for producing an IQueryable object tied to some persistence apparatus; when we call ToList on the query object, the apparatus uses the query to fetch data from the database.

var word = "Sir";
var q = Manager.Products
  .Where(p => p.ProductName.Contains(word));
var results = q.ToList();// returns 3 products

Of course you don’t know how many words the user will enter. You want to be prepared for more than one so you write this too-simple helper method that returns an array of words from the text entered in the text box:

private IEnumerable<String> GetWords(string phrase) {
  return phrase.Split(new[] {' '}, StringSplitOptions.RemoveEmptyEntries);
}

Now all you have to do is replace the “Where” clause with a sequence of OR clauses. You’ll want to construct it by iterating over the words. Go ahead and write it. I’ll wait.

Having trouble? I’ll give you the user’s input: “Sir Cajun Louisiana”. Did that help?

You will probably come up with the following:

var q = Manager.Products
  .Where(p =>
     p.ProductName.Contains("Sir")
     p.ProductName.Contains("Cajun")
     p.ProductName.Contains("Louisiana")
  );

var results = q.ToList(); // returns 6 products

This is ultimately what the lambda expression must look like.

Of course you cannot demand that the user enter exactly three words any more than you can insist she enter exactly one. You want to construct the lambda dynamically based on the actual number of words entered. Sadly there is no obvious way of constructing a lambda expression dynamically.

The DevForce PredicateBuilder can help you build predicates dynamically.

What’s a “predicate”?

A “predicate” is a function that examines some input and returns true or false.

The code fragment, “p => p.ProductName.Contains(“Sir”)” , is a predicate that examines a product and returns true if the product’s ProductName contains the “Sir” string.

The CLR type of the predicate in our example is:

Func<Product, bool>

Which we can generalize to:

Func<T, bool>

We almost have what we want. When the compiler sees an example of this kind of thing it immediately resolves it into an anonymous delegate. We don’t want the delegate. We need a representation that retains our intent and postpones the resolution into a delegate until the last possible moment. We need an expression tree that we can analyze and morph if necessary. We want a Predicate Expression

Expression<Func<Product, bool>>

This is exactly what the DevForce “Where” extension method demands.

public static IEntityQuery<T> Where<TSource>(
  this IEntityQuery<T> source1, Expression<Func<T,bool>> predicate)

The methods of the static PredicateBuilder class combine two or more Predicate Expressions into a single Predicate Expression that we can pass to this “Where” method.

Let’s stick with the example and see one of those methods in action. Let’s write a little method to produce an IEnumerable of Predicate Expressions, one expression for each of given word.

private IEnumerable<Expression<Func<Product, bool>>> ProductNameTests(
  IEnumerable<String> words) {
    foreach (var each in words) {
      var word = each;
      yield return p => p.ProductName.Contains(word);
    }
}

The result is an IEnumerable of Predicate Expressions about the Product entity. The body is an iterator that returns a Predicate Expression for each word. The expression is exactly the same as the first predicate we wrote when we knew only one word.

If we give it the three-word input in our example, we’ll get an IEnumerable of three Predicate Expressions, each looking for one of the words in the product’s ProductName.

We want to OR these Predicate Expressions together so we will use this static method of PredicateBuilder:

public static Expression<Func<T, bool>>Or<T>(
   params Expression<Func<T, bool>>[] expressions)

You see it takes an array (a params array to be precise) of Predicate Expressions. We will convert the output of our ProductNameTests into an array before giving it to this PredicateBuilder method. The final code looks like so:

var words = GetWords("Sir Cajun Louisiana");
var tests = ProductNameTests(words).ToArray();
if (0 == tests.Length) return;
var productNamePredicate = PredicateBuilder.Or(tests);
var q = Manager.Products.Where(productNamePredicate);
var results = q.ToList(); // returns 6 products

In plain language:

• Split the user’s search text into separate words


• Generate an array of Predicate Expressions that look for each word in the ProductName


• Skip the query if there are no clauses … because there are no words


• Ask “PredicateBuilder.Or” to combine the tests into a single Predicate Expression


• Run it to get results.

Other PredicateBuilder Methods

There are 6 interesting methods.

Method	Syntax by example
Or	p1.Or(p2)
Or	PredicateBuilder.Or(p1, p2, p3 .. pn)
And	p1.And(p2)
And	PredicateBuilder.And(p1, p2, p3 .. pn)
True	PredicateBuilder.True()
False	PredicateBuilder.False()

“p” = Predicate Expression.

All expressions must be of the same type (e.g., Product).

Examples:

Expression<Func<Product, bool>> p1, p2, p3, p4, bigP;

// Sample predicate expressions

p1 = p => p.ProductName.Contains("Sir");

p2 = p => p.ProductName.Contains("Cajun");

p3 = p => p.ProductName.Contains("Louisiana");

p4 = p => p.UnitPrice > 20;

bigP = p1.Or(p2); // Name contains "Sir" or "Cajun"

// Name contains any of the 3
bigP = p1.Or(p2).Or(p3);

bigP = PredicateBuilder.Or(p1, p2, p3); // Name contains any of the 3

bigP = PredicateBuilder.Or(tests); // OR together some tests

bigP = p1.And(p4); // "Sir" and price > 20

// Name contains "Cajun" and "Lousiana" and the price > 20

bigP = PredicateBuilder.And(p2, p3, p4);

bigP = PredicateBuilder.And(tests); // AND together some tests

// Name contains either “Sir” or “Louisiana” AND price > 20
bigP = p1.Or(p3).And(p4);

bigP = PredicateBuilder.True<Product>().And(p1);// same as p1

bigP = PredicateBuilder.False<Product>().Or(p1);// same as p1

// Not useful

bigP = PredicateBuilder.True<Product>().Or(p1);// always true

bigP = PredicateBuilder.False<Product>().And(p1);// always false

Observations

Notice that one each of the OR and the AND methods are Predicate Expression extension methods; they make it easier to compose predicates at number of Predicate Expressions known at design time.

Put a breakpoint on any of the “bigP” lines and ask the debugger to show you the result as a string. Here is the Immediate Window output for “bigP = p1.Or(p3).And(p4);”

{p => ((p.ProductName.Contains("Sir")   p.ProductName.Contains("Louisiana")) &&
  (p.UnitPrice > Convert(20)))}

The True and False methods return Predicate Expression constants that help you jumpstart your chaining of PredicateBuilder expressions. Two of the combinations are not useful.

Compared to Albahari Brothers’ Predicate Builder

The Albahari brothers covered similar ground with their PredicateBuilder described here. Why duplicate their work?

Actually, we are not. We would have been happy to use their PredicateBuilder (which is open source) … if it worked in Silverlight. But it doesn’t. Moreover, it relies on a peculiar trick that adds mystery with no apparent benefit.

Why doesn’t it work in Silverlight? Because their implementation depends upon private reflection … which is forbidden in Silverlight.

Why do they require private reflection? Because they postpone resolution of the modified Expression tree until the query is resolved. In order to “lazily resolve” the expression tree, they have to carry unresolved expressions around inside the modified trees. The only way to do this is if the inner, unresolved expressions are closures … and closures are private.

The DevForce Predicate Builder resolves combined expressions immediately. Write “p1.Or(p2)” and you immediately get back the new expression

 p => p.ProductName.Contains(“Sir”)  p.ProductName.Contains(“Cajun”)

With the Albahari Predicate Builder you’d get something sort of like:

 p => p1.Invoke(p)   p.ProductName.Contains(“Cajun”)

“p1” is the closure I’m talking about.

Which brings me to the other apparent strangeness here. What is “Invoke” doing in there? I don’t want to call “invoke” … ever.

Don’t worry. “Invoke” is never actually called. “Invoke” is a placeholder in the expression tree. The Albahari’s have hijacked the “Invoke” method and are using it as a marker that means “when you finally resolve this expression, replace the invoke with the mini-expression inside the closure ‘p1’.”

They have also cleverly hooked their own expression “pre-compiler” into the expression object so that, when something tries to use this LINQ expression, this “pre-compiler” gets to fix up all the “Invoke” markers before that something get its chance to evaluate the expression.

Bet you didn’t know you could do that. It’s a good example of the Chain of Responsibility pattern.

I don’t want to go any deeper than this. You can learn more by reading up on their LinqKit and you can see where they discovered how to do it from Tomas Petricek.

I am only going this far so I can explain that

1. They are postponing Expression tree resolution until the LINQ query is actually consumed
2. The trick to doing so is to embed a closure of the left expression in the new Expression tree and mark it with an “Invoke” method
3. The ultimate resolution of the LINQ query requires penetrating that closure
4. Closures are private so you can’t penetrate them in Silverlight
5. Therefore we can’t use their PredicateBuilder to dynamically construct LINQ in SIlverlight

Thankfully, the DevForce PredicateBuilder resolves these dynamically constructed LINQ expressions immediately so there is never an embedded closure. Nor does it need to hijack the “Invoke” method as a marker.

Importantly, there is no loss of expressiveness, performance, or capability by immediate resolution. I have no idea why they introduced this complication. The closures and “Invokes” add unnecessary mystery in my opinion … without apparent benefit.

To be clear, you still can use their PredicateBuilder to dynamically construct LINQ queries in .NET code. Their PredicateBuilder works fine with DevForce LINQ queries in regular .NET … as long as you remember to use AsExpandable (as you would for a dynamically composed Entity Framework LINQ query).

I prefer to use the DevForce version in both .NET and Silverlight myself.

Simplify the Prism Event Aggregator Protocol

I read with fascination Jeremy’s  “Braindump on the Event Aggregator Pattern” and recommend it to you, gentle reader. You’ll find concise coverage of the intent and the issues that confront anyone who would build his/her own EA.

Half way in Jeremy launches a critique of EA as implemented in Prism. It reminded me of my initial reaction to Prism's EA which was "man, this seems clunky."  Ouch!

One forgets in time and just accepts all the extra motion as "just the way it is". After awhile you don't even see it anymore. Then someone – a Jeremy – comes along and wags his finger at it.

Now I like Prism's reliance on strongly typed EA events. Much better than the string “Event Topics” of CAB. But, he is right. I shouldn't have to publish an event by writing:

  // Publish an event with no payload (have to fake it); the event type is the message
  _eventAggregator.GetEvent<CacheClearEvent>.Publish(null);

  // Publish an event with a strongly typed payload
  _eventAggregator.GetEvent<EntityNotificationEvent>().Publish(eventPayload);

Aside: I don't mind defining strongly-typed CacheClearEvent and EntityNotificationEvent. But Prism forces me to define an EventNotificationEventPayload class to support the second event.  This smelled wrong ... but I persevered.

Nor should I have to subscribe by writing:

  _eventAggregator.GetEvent<CacheClearEvent>().Subscribe(Clear);
  _eventAggregator.GetEvent<EntityNotificationEvent>().Subscribe(SetCustomer);

The issue here is that it takes two steps to publish and subscribe. First I have to get the event object from the EA and then I tell it what I want to do. This bugs Jeremy. And now it bugs me.

My initial reaction was "let's just clean this up with some extension methods."

I set aside, for the moment, his desire to eliminate the subscription line altogether; one step at a time, OK?

Ah ... but as soon as I started working on those extensions, I realized why the Prism team hadn't done this themselves. I saw how the team got hung up.

The Prism EA designers made a fundamental decision that the strongly-typed event must have a separate, strongly-typed payload. You see this in the signature of the base class for all such events, CompositePresentationEvent<TPayload>. If you want to define a pub/sub event in Prism you must inherit from this class.

This leads you to the following two extension method signatures:

  public static void Subscribe<TEventType, TPayload>(
    this IEventAggregator aggregator, Action<TPayload> action)
      where TEventType : CompositePresentationEvent<TPayload>

  public static void Publish<TEventType,TPayload>(
    this IEventAggregator aggregator, TPayload payload)
      where TEventType : CompositePresentationEvent<TPayload>

These seem ok until you try to use them. You end up with this:

  _eventAggregator.Publish<CacheClearEvent, object>(null);
  _eventAggregator.Publish<EntityNotificationEvent, EntityNotificationEventPayload>(eventPayload);

  _eventAggregator.Subscribe<CacheClearEvent, object>(null);
  _eventAggregator.Subscribe<EntityNotificationEvent, EntityNotificationEventPayload>(SetCustomer);

I’m not sure I’ve gained much clarity. Those type parameters are plain ugly.

While I can add more extension methods to smooth the way for the events that take no payload (e.g., CacheClearEvent), I'm stuck with this syntax for the more interesting events that take a payload. Maybe you can finesse them away; I can’t find a way.

This lead me to ask "what if an event that needed a payload was itself the payload?" I realized I could bring this off with the existing Prism EA ... if I adopted a rather strange convention.

Here is my new CacheClearMessage for example.

  public class CacheClearMessage : CompositePresentationEvent<CacheClearMessage> { }

Notice how it inherits from CompositePresentationEvent<TPayload> as required. But it cleverly references itself as the payload.

My new EntityNotificationMessage looks like this:

  public class EntityNotificationMessage
    : CompositePresentationEvent<EntityNotificationMessage> {

    public EntityNotificationMessage() {}
    public EntityNotificationMessage(Type entityType, object entityId) {
      EntityType = entityType;
      EntityId = entityId;
    }
    public object EntityId { get; private set; }
    public Type EntityType { get; private set; }
    // more stuff
  }

Notice that it contains its own payload which happens to be info about the subject of the entity notification. I no longer need my EntityNotificationEventPayload class which I delete from my project (yipee!)

Notice I can instantiate the message without a payload too. Prism requires a parameterless ctor in order to register the event; you wouldn't actually use this one.

Now I add two extension methods that look like this:

  public static void Publish<TMessageType>(
    this IEventAggregator aggregator, TMessageType message)
      where TMessageType : CompositePresentationEvent<TMessageType>

  public static void Subscribe<TMessageType>(
    this IEventAggregator aggregator, Action<TMessageType> action)
      where TMessageType : CompositePresentationEvent<TMessageType>

My client code looks much cleaner now:

  _eventAggregator.Publish<CacheClearMessage>();
  _eventAggregator.Publish(entityNotificationMessage);
 
  _eventAggregator.Subscribe<CacheClearMessage>(Clear);
  _eventAggregator.Subscribe<EntityNotificationMessage>(SetCustomer);

I get type inference in only one usage (one of the Publish calls); maybe the explicitness is not so bad. At least there is only one type parameter.

--

I'm not as freaked out by the subscriptions but I get Jeremy's point. I should be able to identify the Clear and SetCustomer methods as methods to be subscribed to. I shouldn't have to explicitly import the EventAggregator and wire the methods to it.

I'm not sure what the best way to get around that is just yet (and he doesn't seem to have settled on an answer either). So I'll just stop right here for now.

Tonight, as I post, I'm feeling that this was a good refinement. I see no benefit in forcing separation of the event class and the payload class. Maybe the Prism designers will educate me. Maybe I'll wake up tomorrow with a hangover and wish I'd left well enough alone.

Your thoughts are welcome.

Tom DeMarco Recants

Developers under thirty may not know the name, Tom DeMarco, but if you ever drew a paycheck from a large organization, you’ve felt his influence. When your boss said “You can’t control what you can’t measure”, he was channeling Mr. DeMarco.

I can think of no single individual who has had a more pervasive and decisive say in how we manage software development. Corporate CIOs, IT directors, and senior architects have listened to him prescribe “best practices” for more than 30 years. So when he says “I was wrong all along” … it’s like hearing Robert McNamara confessing his tragic mistakes. You feel you knew it all along … but it hits hard to hear him say it.

In “Software Engineering: An Idea Whose Time Has Come and Gone?”, Mr. DeMarco confronts his life’s legacy -- his insistence that precise planning and intense monitoring are essential to project success -- and condemns it.

Let’s not go overboard. He isn’t against measurement (and neither am I).  If you can measure it … and your measure bears a well-understood relationship to the good or evil you want to assess … and the cost of measurement is reasonable … then measure it.

But putting control and measurement first utterly obscures what really matters: the potential value of the project and the forces that can determine the timing and success of the project. Indeed, planning and measurement can … and often do … actively impede success.

Re-blogging is an etiquette violation. I’m doing it anyway because I fear readers unfamiliar with Mr. DeMarco will miss the many gems in his short and sweet mea culpa. Some juicy quotes to entice you.

“Strict control is something that matters a lot on relatively useless projects and much less on useful projects”

“The more you focus on control, the more likely you’re working on a project that’s striving to deliver something of relatively minor value.”

“First we need to select projects where precise control won’t matter so much. Then we need to reduce our expectations for exactly how much we’re going to be able to control them, no matter how assiduously we apply ourselves to control.”

“Consistency and predictability are still desirable, but they haven’t ever been the most important things. The more important goal is transformation.”

“You say to your team leads, for example, “I have a finish date in mind, and I’m not even going to share it with you. When I come in one day and tell you the project will end in one week, you have to be ready to package up and deliver what you’ve got as the final product.””

Before we get all smug about agile and the disaster that is “Waterfall Design” … I want you to think a solid few minutes about whether you are susceptible to putting engineering practices in front of something more important.

His critique seems to target only an overweening attention to planning and control. That just happens to be his personal road to perdition. He aims at a wider target.

Look at the title again. He asks is “Software Engineering” an idea whose time has gone? If you find that TDD or DDD or BDD … are what you talk about first, are you making the same category of mistake? How can these practices – whatever their merits - be more important than the value of the project and whether it is delivering on its promise? If, for example, you believe that there isn’t great software without unit tests … have you fallen into the same trap?

Imagine the courage it takes to come to such conclusions about your life’s work? I salute Mr. DeMarco. And I hasten to add that his writing was always more nuanced and more useful than his own harsh self-criticism suggests. He remains well worth reading … more so in the light he shines from this vantage point.

Aside: I learned about this article from a Kent Beck tweet. You’ll find the endlessly fascinating Kent Beck blog here.