91

There's a debate going on in our team at the moment as to whether modifying code design to allow unit testing is a code smell, or to what extent it can be done without being a code smell. This has come about because we're only just starting to put practices in place that are present in just about every other software dev company.

Specifically, we will have a Web API service that will be very thin. Its main responsibility will be marshalling web requests/responses and calling an underlying API that contains the business logic.

One example is that we plan on creating a factory that will return an authentication method type. We have no need for it to inherit an interface as we don't anticipate it ever being anything other than the concrete type it will be. However, to unit test the Web API service we will need to mock this factory.

This essentially means we either design the Web API controller class to accept DI (through its constructor or setter), which means we're designing part of the controller just to allow DI and implementing an interface we don't otherwise need, or we use a third party framework like Ninject to avoid having to design the controller in this way, but we'll still have to create an interface.

Some on the team seem reluctant to design code just for the sake of testing. It seems to me that there has to be some compromise if you hope to unit test, but I'm unsure how allay their concerns.

Just to be clear, this is a brand new project, so it's not really about modifying code to enable unit testing; it's about designing the code we're going to write to be unit testable.

  • 33
    Let me repeat this: you colleagues want unit tests for new code, but they refuse to write the code in a way it is unit testable, though there is no risk in breaking anything existing? If that's true, you should accept @KilianFoth's answer and ask him to highlight the first sentence in his answer in bold! Your colleagues apparently have a very big misunderstanding about what their job is. – Doc Brown Jan 29 at 12:36
  • 20
    @Lee: Who says that decoupling is always a good idea? Have you ever seen a codebase in which everything is passed as an interface created from an interface factory using some configuration interface? I have; it was written in Java, and it was a complete, unmaintainable, buggy mess. Extreme decoupling is code obfuscation. – Christian Hackl Jan 29 at 13:51
  • 8
    Michael Feathers' Working Effectively with Legacy Code deals very nicely with this issue, and should give you a good idea about the advantages of testing even in a new code base. – l0b0 Jan 30 at 1:37
  • 8
    @l0b0 That's pretty much the bible for this. On stackexchange it wouldn't be an answer to the question, but in RL I would tell OP to get this book an read it (at least partly). OP, get Working Effectively with Legacy Code and read it, at least partly (or tell your boss to get it). It addresses questions like these. Especially if you didn't do testing and now you're getting into it - you might have 20 years experience, but you will now do things you don't have experience with. It's so much easier to read about them than to painstakingly learn all of that by trial and error. – R. Schmitz Jan 30 at 10:50
  • 4
    Thanks for the recommendation of Michael Feathers' book, I will definitely pick up a copy. – Lee Jan 30 at 11:44

15 Answers 15

203

Reluctance to modify code for the sake of testing shows that a developer hasn't understood the role of tests, and by implication, their own role in the organization.

The software business revolves around delivering a code base that creates business value. We have found, through long and bitter experience, that we cannot create such code bases of nontrivial size without testing. Therefore, test suites are an integral part of the business.

Many coders pay lip service to this principle but subconsciously never accept it. It is easy to understand why this is; the awareness that our own mental capability is not infinite, and is in fact, surprisingly limited when confronted with the enormous complexity of a modern code base, is unwelcome and easily suppressed or rationalized away. The fact that test code is not delivered to the customer makes it easy to believe that it is a second-class citizen and non-essential compared to the "essential" business code. And the idea of adding testing code to the business code seems doubly offensive to many.

The trouble with justifying this practice has to do with the fact that the entire picture of how value is created in a software business is often only understood by higher-ups in the company hierarchy, but these people don't have the detailed technical understanding of the coding workflow that is required to understand why testing can't be gotten rid of. Therefore they are too often pacified by practitioners who assure them that testing may be a good idea in general, but "we are elite programmers who don't need crutches like that", or that "we don't have time for that right now", etc. etc. The fact that business success is a numbers game and that avoiding technical debt, assuring quality etc. shows its value only in the longer term means that they are often quite sincere in that belief.

Long story short: making code testable is an essential part of the development process, no different than in other fields (many microchips are designed with a substantial proportion of elements only for testing purposes), but it's very easy to overlook the very good reasons for that. Don't fall into that trap.

  • 39
    I would argue it depends of the kind of change. There's a difference between making code easier to test, and introducing test-specific hooks that should NEVER be used in production. I am personally wary of the latter, because Murphy... – Matthieu M. Jan 29 at 13:16
  • 60
    Unit tests often break encapsulation and make the code under test more complex than would otherwise be required (e.g. by introducing additional interface types or adding flags). As always in software engineering, every good practice and every good rule has its share of liabilities. Blindly producing a lot of unit tests can have a detrimental effect on business value, not to mention that writing and maintaining the tests already costs time and effort. In my experience, integration tests have a much greater ROI and tend to improve software architectures with fewer compromises. – Christian Hackl Jan 29 at 13:17
  • 20
    @Lee Sure but you need to consider whether having a specific type of tests warrants the increase in code complexity. My personal experience is that unit tests are a great tool up until the point where they require fundamental design changes to accommodate mocking. That’s where I switch to a different type of tests. Writing unit tests at the expense of making the architecture substantially more complex, for the sole purpose of having unit tests, is navel-gazing. – Konrad Rudolph Jan 29 at 15:17
  • 21
    @ChristianHackl why would a unit test break encapsulation? I've found that for the code I've worked on, if there is a perceived need to add extra functionality to enable testing, the actual problem is that the function you are to test needs refactoring, so all the functionality is at the same level of abstraction (it's differences in abstraction level that usually create this "need" for extra code), with lower level code moved to their own (testable) functions. – Baldrickk Jan 29 at 16:29
  • 29
    @ChristianHackl Unit tests should never break encapsulation, if you are trying to access private, protected or local variables from a unit test, you are doing it wrong. If you are testing functionality foo, you are only testing if it actually worked, not if local variable x is the square root of the input y in the third iteration of the second loop. If some functionality is private then so be it, you're going to be testing it transitively anyway. if it's really big and private? That's a design flaw, but probably not even possible outside of C and C++ with header implementation seperation. – opa Jan 29 at 17:18
75

It's not as simple as you might think. Let's break it down.

  • Writing unit tests is definitely a good thing.

BUT!

  • Any change to your code can introduce a bug. So changing the code without a good business reason is not a good idea.

  • Your 'very thin' webapi doesn't seem like the greatest case for unit testing.

  • Changing code and tests at the same time is a bad thing.

I would suggest the following approach:

  1. Write integration tests. This should not require any code changes. It will give you your basic test cases and enable you to check that any further code changes you make don't introduce any bugs.

  2. Make sure new code is testable and has unit and integration tests.

  3. Make sure your CI chain runs tests after builds and deployments.

When you have those things set up, only then start thinking about refactoring legacy projects for testability.

Hopefully everyone will have learnt lessons from the process and have a good idea of where testing is most needed, how you want to structure it and the value it brings to the business.

EDIT: Since I wrote this answer, the OP has clarified the question to show that they are talking about new code, not modifications to existing code. I perhaps naively thought the "Is unit testing good?" argument was settled some years ago.

It's hard to imagine what code changes would be required by unit tests but not be general good practice you would want in any case. It would probably be wise to examine the actual objections, possibly it's the style of unit testing which is being objected to.

  • 12
    This is a much better answer than the accepted one. The imbalance in votes is dismaying. – Konrad Rudolph Jan 29 at 20:24
  • 4
    @Lee A unit test should test a unit of functionality, which may or may not correspond to a class. A unit of functionality should be tested at its interface (which may be the API in this case). Testing may highlight design smells and the need to apply some different/more levelisation. Build your systems from small composable pieces, they will be easier to reason about and test. – Wes Toleman Jan 30 at 3:44
  • 2
    @KonradRudolph: I guess missed the point where the OP added that this question was about designing new code, not changing existing one. So there is nothing to break, which makes most of this answer not applicable. – Doc Brown Jan 30 at 12:10
  • 1
    I strongly disagree with the statement that writing unit tests is always a good thing. Unit tests are good only in some cases. It's silly to use unit tests to test frontend (UI) code, they are made to test business logic. Also, it's good to write unit tests to replace missing compilation checks (e.g. in Javascript). Most frontend-only code should write end-to-end test exclusively, not unit tests. – Sulthan Jan 31 at 7:32
  • 1
    Designs can definitely suffer from "Test induced damage". Usually testability improves design: You notice when writing tests that something can't be fetched but must be passed in, making for clearer interfaces and so on. But occasionally you'll stumble across something that requires an uncomfortable design only for testing. An example could be a test-only constructor required in your new code due to existing third party code that uses a singleton for example. When that happens: take a step back and make an integration test only, instead of damaging your own design in the name of testability. – Anders Forsgren Feb 1 at 12:15
18

Designing code to be inherently testable is not a code smell; on the contrary, it is the sign of a good design. There are several well-known and widely-used design patterns based on this (e.g., Model-View-Presenter) that offer easy (easier) testing as a big advantage.

So, if you need to write an interface for your concrete class in order to more easily test it, that is a good thing. If you already have the concrete class, most IDEs can extract an interface from it, making the effort required minimal. It is a little bit more work to keep the two in sync, but an interface shouldn't change much anyway, and the benefits from testing may outweigh that extra effort.

On the other hand, as @MatthieuM. mentioned in a comment, if you're adding specific entry points into your code that shouldn't ever be used in production, solely for the sake of testing, that might be a problem.

  • That problem can be solved through static code analysis - flag the methods (e.g. must be named _ForTest) and check the codebase for calls from non-test code. – Riking Feb 4 at 2:49
13

It is IMHO very simple to understand that for creating unit tests, the code to be tested must have at least certain properties. For example, if the code does not consist of individual units which can be tested in isolation, the word "unit testing" does not even start to make sense. If the code does not have these properties, it must be changed first, that is pretty obvious.

Said that, in theory, one can try to write some testable code unit first, applying all the SOLID principles, and then try to write a test for it afterwards, without further modifying the original code. Unfortunately, writing code which is really unit testable is not always dead simple, so it is quite likely there will be some changes necessary which one will only detect when trying to create the tests. This is true for code even when was written with the idea of unit testing in mind, and it is definitely more true for code which was written where "unit testability" was not at the agenda at the beginning.

There is a well-known approach which tries to solve the problem by writing the unit tests first - it is called Test Driven Development (TDD), and it can surely help to make code more unit testable right from the start.

Of course, the reluctance of changing code afterwards for making it testable arises often in a situation where the code was manually tested first and/or works fine in prodcution, so changing it could actually introduce new bugs, that's true. The best approach to mitigate this is to create a regression test suite first (which can often be implemented with only very minimal changes to the code base), as well as other accompanying measures like code reviews, or new manual test sessions. That should you give enough confidence to make sure redesigning some internals does not break anything important.

  • Interesting that you mention TDD. We're attempting to bring in BDD/TDD, which also has met some resistance - namely what "the minimum code to pass" really means. – Lee Jan 29 at 12:08
  • 2
    @Lee: bringing changes into an organization always causes some resistance, and it always needs some time to adapt new things, that's not new wisdom. This is a people problem. – Doc Brown Jan 29 at 12:12
  • Absolutely. I just wish we'd been given more time! – Lee Jan 29 at 12:13
  • It is frequently a matter of showing people that doing it this way will save them time (and hopefully quickly too). Why do something that will not benefit you? – Thorbjørn Ravn Andersen Feb 2 at 22:44
  • @ThorbjørnRavnAndersen: The team may likewise show to the OP that their approach will save time. Who knows? But I wonder if we aren't actually facing problems of a less technical nature here; the OP keeps coming here to tell us what his team does wrong (in his opinion), as if he was trying to find allies for his cause. It might be more beneficial to actually discuss the project together with the team, not with strangers on Stack Exchange. – Christian Hackl Feb 5 at 13:26
11

I take issue with the (unsubstantiated) assertion you make:

to unit test the Web API service we will need to mock this factory

That's not necessarily true. There are lots of ways to write tests, and there are ways to write unit tests that don't involve mocks. More importantly, there are other kinds of tests, such as functional or integration tests. Many times it is possible to find a "test seam" at an "interface" that is not an OOP programming language interface.

Some questions to help you find an alternative test seam, which might be more natural:

  • Will I ever want to write a thin Web API over a different API?
  • Can I reduce code duplication between the Web API and the underlying API? Can one be generated in terms of the other?
  • Can I treat the whole Web API and underlying API as a single "black box" unit and meaningfully make assertions about how the whole thing behaves?
  • If the Web API had to be replaced with a new implementation in the future, how would we go about doing that?
  • If the Web API was replaced with a new implementation in the future, would clients of the Web API be able to notice? If so, how?

Another unsubstantiated assertion you make is about DI:

we either design the Web API controller class to accept DI (through its constructor or setter), which means we're designing part of the controller just to allow DI and implementing an interface we don't otherwise need, or we use a third party framework like Ninject to avoid having to design the controller in this way, but we'll still have to create an interface.

Dependency injection does not necessarily mean creating a new interface. For example, in the cause of an authentication token: can you simply create a real authentication token programmatically? Then the test can create such tokens and inject them. Does the process for validating a token depend on a cryptographic secret of some kind? I hope you haven't hardcoded a secret -- I would expect you can read it from storage somehow, and in that case you can simply use a different (well-known) secret in your test cases.

This is not to say that you should never create a new interface. But don't get fixated on there only being one way to write a test, or one way to fake a behavior. If you think outside the box you can usually find a solution that will require a minimum of contortions of your code and yet still give you the effect you want.

  • Point taken about the assertions regarding interfaces, but even if we didn't use them we'd still have to inject objects some how, this is the concern from the rest of the team. i.e. some on the team would be happy with a parameterless ctr instantiating the concrete implementation and leaving it at that. In fact one member floated the idea of using reflection to inject mocks so we don't have to design code to accept them. Which is a reeky code smell imo – Lee Jan 31 at 14:31
9

You are in luck as this is a new project. I've found that Test Driven Design works very well for writing good code (which is why we do it in the first place).

By figuring out up front how to invoke a given piece of code with realistic input data, and then get realistic output data which you can check is as intended, you do the API design very early in the process and have a good chance of getting a useful design because you are not hindered by existing code that has to be rewritten to accomodate. Also it is easier to understand by your peers so you can have good discussions again early in the process.

Note that "useful" in the above sentence means not only that the resulting methods are easy to invoke, but also that you tend to get clean interfaces that are easy to rig up in integration tests, and to write mockups for.

Consider it. Especially with peer review. In my experience the investment of time and effort will very quickly be returned.

  • We have a problem with TDD as well, namely, what constitutes "minimum code to pass". I demonstrated to the team this process and they took exception to not just writing what we have already designed - which I can understand. The "minimum" doesn't seem to be defined. If we write a test and have clear plans and designs, why not write that to pass the test? – Lee Jan 30 at 8:20
  • @Lee "minimum code to pass"... well, this might sound a bit dumb, but it's literally what it says. E.g. if you have a test UserCanChangeTheirPassword, then in the test you call the (not yet existing) function to change the password and then you assert that the password is indeed changed. Then you write the function, until you can run the test and it does neither throw exceptions nor have a wrong assert. If at that point you have a reason to add any code, then that reason goes into another test, e.g. UserCantChangePasswordToEmptyString. – R. Schmitz Jan 30 at 11:10
  • @Lee Ultimately, your tests will end up being the documentation of what your code does, except it's documentation that checks if it is fulfilled itself, instead of just being ink on paper. Also compare with this question - A method CalculateFactorial that just returns 120 and the test passes. That is the minimum. It's also obviously not what was intended, but that just means you need another test to express what was intended. – R. Schmitz Jan 30 at 11:21
  • 1
    @Lee Small steps. The bare minimum may be more than you think when the code rises above trivial. Also the design you do when implementing the whole thing at once may again be less optimal because you make assumptions on how it should be made without having written the tests yet that demonstrate it. Again remember, the code should fail at first. – Thorbjørn Ravn Andersen Feb 1 at 0:22
  • 1
    Also, regression tests are very important. Are they in scope for the team? – Thorbjørn Ravn Andersen Feb 1 at 9:55
8

If you need to modify to the code, that is the code smell.

From personal experience, if my code is difficult to write tests for, it's bad code. It's not bad code because it doesn't run or work as designed, it's bad because I can't quickly understand why it is working. If I encounter a bug, I know it's going to be a long painful job to fix it. The code is also difficult / impossible to reuse.

Good (Clean) code breaks down tasks into smaller sections that are easily understood at a glance (or at least a good look). Testing these smaller sections is easy. I can also write tests that only test a chunk of the codebase with similar ease if I'm fairly confident about the subsections (reuse also helps here as it has already been tested).

Keep the code easy to test, easy to refactor, and easy to reuse from the start and you won't be killing yourself whenever you need to make changes.

I'm typing this while completely rebuilding a project that should have been a throwaway prototype into cleaner code. It's much better to get it right from the start and refactor bad code as soon as possible rather than staring at a screen for hours on end being afraid to touch anything for fear of breaking something that partially works.

  • 3
    "Throwaway prototype" - every single project ever starts life as one of those ... best to think of things as never being that. typing this as I'm .. guess what? ... refactoring a throwaway prototype that turned out not to be ;) – Algy Taylor Jan 30 at 11:49
  • 4
    If you want to be sure that a throw away prototype will be thrown away, write it in a prototype language that will never be allowed in production. Clojure and Python are good choices. – Thorbjørn Ravn Andersen Jan 30 at 13:37
  • 2
    @ThorbjørnRavnAndersen That made me chuckle. Was that meant to be a dig at those languages? :) – Lee Jan 31 at 15:04
  • @Lee. No, just examples of languages that might not be acceptable for production - typically because nobody in the organization can maintain them because they are unfamiliar with them and their learning curves are steep. If those, are acceptable choose another that isn't. – Thorbjørn Ravn Andersen Feb 1 at 6:10
4

I would argue that writing code that cannot be unit tested is a code smell. In general, if your code cannot be unit tested, then it is not modular, which makes it difficult to understand, maintain, or enhance. Maybe if the code is glue code that really only makes sense in terms of integration testing you can substitute integration testing for unit testing, but even then when the integration fails you are going to have to isolate the problem and unit testing is a great way to do it.

You say

We plan on creating a factory that will return an authentication method type. We have no need for it to inherit an interface as we don't anticipate it ever being anything other than the concrete type it will be. However, to unit test the Web API service we will need to mock this factory.

I do not really follow this. The reason to have a factory that creates something is to allow you to change factories or change what the factory creates easily, so other parts of the code do not need to change. If your authentication method is never going to change, then the factory is useless code bloat. However, if you want to have a different authentication method in test than in production, having a factory that returns a different authentication method in test than in production is a great solution.

You do not need DI or Mocks for this. You just need your factory to support the different authentication types and for it to be configurable somehow, such as from a configuration file or environment variable.

2

In every engineering discipline I can think of, there is only one way to achieve decent or higher levels of quality:

To account for inspection/testing in the design.

This holds true in construction, chip design, software development, and manufacturing. Now, this doesn't mean that testing is the pillar that the every design needs to be built around, not at all. But with every design decision, the designers must be clear about the impacts on testing costs and make a conscious decisions about the trade off.

In some cases, manual or automated (e.g. Selenium) testing will be more convenient than Unit Tests, while also providing acceptable test coverage on their own. In rare cases throwing something out there that's almost entirely untested can also be acceptable. But these have to be conscious case by case decisions. Calling a design that accounts for testing a "code smell" indicates a serious lack of experience.

1

I've found that unit testing (and other types of automated testing) have a tendency to reduce code smells, and I can't think of a single example where they introduce code smells. Unit tests usually force you to write better code. If you can't use a method easily under test, why should it be any easier in your code?

Well written unit tests show you how the code is intended to be used. They are a form of executable documentation. I've see hideously written, overly long unit tests that simply couldn't be understood. Don't write those! If you need to write long tests to set up your classes, your classes need refactoring.

Unit tests will highlight where some of your code smells are. I would advise reading Michael C. Feathers' Working Effectively with Legacy Code. Even though your project is new, if it doesn't already have any (or many) unit tests, you might need some non-obvious techniques to get your code to test nicely.

  • 3
    You can be tempted to introduce lots of indirection layers in order to be able to test, and then never use them as expected. – Thorbjørn Ravn Andersen Feb 1 at 6:12
1

In a nutshell:

Testable code is (usually) maintainable code - or rather, code that is hard to test is usually hard to maintain. Designing code that is not testable is akin to designing a machine that is not repairable - pity the poor shmuck who will be assigned to repair it eventually (it might be you).

One example is that we plan on creating a factory that will return an authentication method type. We have no need for it to inherit an interface as we don't anticipate it ever being anything other than the concrete type it will be.

You know that you will need five different types of authentication method types in three years time, now that you said that, right? Requirements change, and while you should avoid overengineering your design, having a testable design means that your design has (just) enough seams to be altered without (too much) pain - and that the module tests will provide you with automated means to see that your changes don't break anything.

1

Designing around dependency injection isn't a code smell - it's best practice. Using DI isn't just for testability. Building your components around DI aids modularity and reusability, more easily allows for major components to be swapped out (such as a database interface layer). While it adds a degree of complexity, done right it allows for better separation of layers and isolation of functionality which makes the complexity easier to manage and navigate. This makes it easier to properly validate the behavior of each component, reducing bugs, and can also make it easier to track down bugs.

  • 1
    "done right" is a problem. I have to maintain two projects where DI was done wrong (although aiming for doing it "right"). This makes the code plain horrible and much worse than the legacy projects without DI and unit testing. Getting DI right isn't easy. – Jan Feb 4 at 12:21
  • @Jan that's interesting. How did did they do it wrong? – Lee Feb 10 at 16:07
  • 1
    @Lee One project is a service that needs a fast start time but is horribly slow on start because all the class initialization is done upfront by the DI framework (Castle Windsor in C#). Another problem that I see in these projects is mixing up DI with the creation of objects with "new", sidestepping the DI. That makes testing hard again and led to some nasty race conditions. – Jan Feb 11 at 9:09
1

This essentially means we either design the Web API controller class to accept DI (through its constructor or setter), which means we're designing part of the controller just to allow DI and implementing an interface we don't otherwise need, or we use a third party framework like Ninject to avoid having to design the controller in this way, but we'll still have to create an interface.

Let's look at the difference between a testable:

public class MyController : Controller
{
    private readonly IMyDependency _thing;

    public MyController(IMyDependency thing)
    {
        _thing = thing;
    }
}

and non-testable controller:

public class MyController : Controller
{
}

The former option has literally 5 extra lines of code, two of which can be autogenerated by Visual Studio. Once you've setup your dependency injection framework to substitute a concrete type for IMyDependency at runtime - which for any decent DI framework, is another single line of code - everything Just Works, except now you can mock and thus test your controller to your heart's content.

6 extra lines of code to allow testability... and your colleagues are arguing that's "too much work"? That argument doesn't fly with me, and it shouldn't fly with you.

And you don't have to create and implement an interface for testing: Moq, for example, allows you to simulate the behaviour of a concrete type for unit-testing purposes. Of course, that won't be of much use to you if you can't inject those types into the classes you're testing.

Dependency injection is one of those things that once you understand it, you wonder "how did I work without this?". It's simple, it's effective, and it just Makes Sense. Please, don't allow your colleagues' lack of understanding of new things to get in the way of making your project testable.

  • 1
    What you are so quick to dismiss as "lack of understanding of new things" may turn out to be a good understanding of old things. Dependency injection is certainly not new. The idea, and probably the earliest implementations, are decades old. And yes, I believe your answer is an example of code becoming more complicated due to unit testing, and possibly an example of unit tests breaking encapsulation (because who says the class has a public constructor in the first place?). I've often removed dependency injection from codebases I had inherited from someone else, because of the tradeoffs. – Christian Hackl Feb 4 at 8:50
  • Controllers always have a public constructor, implicit or not, because MVC requires it. "Complicated" - maybe, if you don't understand how constructors work. Encapsulation - yes in some cases, but the DI vs encapsulation debate is an ongoing, highly subjective one that won't help here, and in particular for most applications, DI is going to serve you better than encapsulation IMO. – Ian Kemp Feb 4 at 9:37
  • Regarding public constructors: indeed, that's a particularity of the framework being used. I was thinking about the more general case of an ordinary class which is not instantiated by a framework. Why do you believe that seeing additional method parameters as added complexity equals a lack of understanding about how constructors work? However, I appreciate that you acknowledge the existence of a tradeoff between DI and encapsulation. – Christian Hackl Feb 4 at 12:07
0

When I write unit tests, I begin to think of what could go wrong inside my code. It helps me to improve code design and apply the single responsibility principle (SRP). Also, when I come back to modify the same code a few months later, it helps me to confirm that existing functionality isn't broken.

There is a trend to use pure functions as much as you can (serverless apps). Unit testing helps me to isolate state and write pure functions.

Specifically, we will have a Web API service that will be very thin. Its main responsibility will be marshalling web requests/responses and calling an underlying API that contains the business logic.

Write unit tests for the underlying API first and if you have sufficient development time, you need to write tests for the thin Web API service as well.

TL;DR, unit testing helps improve quality of code and helps make future changes to code risk free. It also improves readability of the code. Use tests instead of comments to make your point.

0

The bottom line, and what should be your argument with the reluctant lot, is that there is no conflict. The big mistake seems to have been that someone coined the idea to "design for testing" to people that hate testing. They should have just shut their mouths or word it differently, like "let's take the time to do this right".

The idea that "you have to implement an interface" to make something testable is wrong. The interface is already implemented, it just isn't declared in the class declaration yet. It is a matter of recognizing existing public methods, copy their signatures to an interface and declaring that interface in the class's declaration. No programming, no changes to existing logic.

Apparently some people have a different idea about this. I suggest you try to fix this first.

protected by gnat Feb 1 at 21:51

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.