Slashdot Mirror
Pragmatic JUnit Testing
What's the Approach? You can go a long way in the unit testing world with this book. It's a great starter guide for learning about unit testing; the first chapter has several small sections which answer many of the questions a novice might have about the subject, including some of the most common objections to unit testing. Then the book launches into a detailed user guide to the JUnit package, including some useful but lesser-known topics like writing your own custom asserts.
After you've got the basics down, the book spends several chapters drilling into the concepts, considerations, and best practices for writing good tests. This section of the book is quite practical, and is where the true worth of the book comes out. Any developer who reads this section will come away better for it even if they never consistently write unit tests. It's amazing how many testing issues turn out to be development issues at their core; this section of the book points some of them out. Simply being aware of these issues can help you to avoid a number of common pitfalls.
Finally, the book discusses common unit testing problems and their solutions. This section is a lot of fun because of the anecdotes that Andy and Dave include; unfortunately some of them I recognized from my own experiences as a developer. (What, me do that? No, uh, of course not!)
Who's it For? This book is primarily for developers (and testers who are intimately involved in the development process), since unit testing per se is primarily a developer task. Depending on the organization of the software shop, by the time code gets to a formal testing or QA group it's probably too late to apply unit testing methods effectively. (Note I said unit testing methods here.) The book is ideal for the unit testing novice, and there's a lot of good information here for more advanced developers to ponder.The book is targeted at those who want to learn "how to think about tests, how testing affects design, and how to handle certain team-wide issues you may be having" (from the Preface). In particular the book is an excellent source of information for those developers who have recently moved into the Java area and are looking for a good testing framework.
No matter what your role in the development process, this book can help give you a fresh perspective into unit testing and related issues. It prods you to think about issues that perhaps you hadn't considered before.
Pros As you can probably tell, I like the Pragmatic Programmer writing style. It's informal and friendly without being chatty or wordy, and the main text moves along well. More technical discussions are reserved for sidebars, and there are "Joe Asks..." sections throughout the book that anticipate questions that Joe Programmer might ask about the current topic.The examples in the book are excellent. Many books err on the side of code snippets that don't have enough context to be understandable, or multi-page code extravaganzas that bury the details of the topic at hand under mountains of cruft. I found the examples here to give just the right amount of detail to understand what's going on. And the source for all the examples is available from the author's web site.
Cons What this book is not is a reference manual for JUnit. The book doesn't include a javadoc listing for the JUnit package, nor does it have a section that digs into JUnit's internals. This is not necessarily a Bad Thing because that information is available from the JUnit web site, but it does bear mentioning.If there was anything I found to complain about it was the use of acronyms in the concepts section of the book. I'll admit right here that I'm no lover of acronyms, so I felt that using things like Right-BICEP, CONNECT, and A-TRIP as mnemonics for the concepts was a bit over the top. (Read the book to see what I mean.) However, I have to admit that they help me remember the concepts discussed, so I guess they do their job (grumble).
Why Should I Read this Book?If you hate testing, this book will very probably change your mind, and you'll have fun in the process (the stories are great!). And even if you still hate testing after you've finished the book, you will have enhanced your development skills enough to be able to delude yourself that testing isn't necessary for you (well, maybe not that much). At the very least, this book will make you aware of the various issues that proper testing can expose and maybe help you try to avoid them. By shining a light into these areas, it can still make you a more effective developer... maybe effective enough to realize how much more solid your work could be with this type of testing coverage.
If you love testing, this book will remind you why you began the love affair. Your love will be strengthened and deepened, and the tests you write will undoubtedly get better, too.
For Java developers specifically, this book will get you started writing JUnit tests painlessly and effectively and will take you a long way down the road to unit test proficiency. It will give developers in any language quite a bit to take "back to the terminal" regarding design and process as well as testing. An excellent and much needed book.
Like volume 1 of the Pragmatic Starter Kit series ( Slashdot Review), this book is available from the Pragmatic Programmer website. The printed version is $29.95 and there's a PDF version available for $20.00. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.
and I prefer this one. Just my opinion though.
When anger rises, think of the consequences.
Confucius (551 BC - 479 BC)
It might be wiser to trust someone slightly more removed from the creators and creation of this book with respect to its utility.
Jeez, I like being both pragmatic and extreme when I use JUnit.
You can defy gravity... for a short time
I guess the question is, are the concepts in the book presented generically enough such that it would port (for a sufficiently loose definition of port) to another framework such as CppUnit?
John
"JUnit, yeah, we'll take your ass out. JUnit, yeah, that's what we're about!"
...was a good one for learning about CVS.
Especially helpful was a small appendix which listed various recipes; i.e., the series of 3 or 4 steps necessary to merge a bugfix from one branch to another or whatever. I always forget the order of those cvs -j options...
The Army reading list
But I applaud your attempt.
Err, JUnit and Extreme Programming are really pretty orthogonal. XP, which is a programming methodology, preaches the idea of writing unit tests before writing code. JUnit is a unit testing framework for Java. Therefore, JUnit can be useful tool in the practice of Extreme Programming, as it is designed for the express purpose of simplifying the creation of unit tests.
My understanding of JUnit is that it's a test framework that simplifies running all the tests you've developed in conjunction with a project. So, if you use the JUnit foundation to build your unit tests, theory says that before turning your code over to the QA teams you'd run JUnit over the whole ball of wax to make sure you aren't sending out something stupidly broken.
But your question: "is JUnit linked with Extreme Programming" is a lot like asking if C++ is linked with Borland or Microsoft. If you're into extreme programming in Java and testing with JUnit, then yes, they're linked. And if you're into extreme programming in Java, then you're probably already using JUnit.
John
Here's an example of one of the PMD JUnit tests that checks the command line option processing:and another that makes sure that an Exception is thrown in another case:When I make a code change, I just rerun all the tests to make sure I didn't break anything. If someone reports a bug, I add a test so that bug won't reappear. Sweet.
The Army reading list
I'm not sure if this is a troll or not... but, what the heck.
> Extra code == extra bloat
Writing tests helps student understand their code and track down bugs. If you don't want students to submit tests, tell them not to.
> Too much documentation
What does JavaDoc - which the compiler discards - have to do with unit tests?
> Unit tests are written with no extra
> skill than the code they test
Don't throw the baby out with the bathwater. If a student is writing tests, at least he's trying to improve his code quality. You're a grad student; show him how to write better code _and_ better tests.
The Army reading list
Junit tests are meant to be fairly simple and if used correctly (e.g. run after every change) can really help to reduce bugs and thus save you time.
where the whole unit testing idea falls down is when you have things you can't unit test like jsps and servlets, these require much more complex testing methods.
Odd, when I took an obligatory basic Java course, the grading was quite the opposite. They WANTED javadoc, they WANTED good comments, and they WANTED us to make sure the code worked right.
They went so far as to give us a few test cases (and if we didn't pass those test cases, we got a zero).
In fact, most of the grading process was to have their automatic system compare the output of our solution with the reference solution...
That said, the TAs were known to be lazy and I don't think anyone ever actually read our comments. I found this out when my solution differed slightly from what they expected (you had to do something different than they expected to direct the output to a file, namely java SomeProg > output, which was WELL commented...) and they couldn't figure it out... (hell, they didn't understand what that DID when I talked with them...)
So I can believe that you're a TA, but beyond that...
You are a fool. Yes, it does add to the bloat of code that you have to read through (poor baby!) but considering the unit test isn't ever shipped as the final product it adds nothing to the finished product besides proof that the program might work.
I cannot ever begin to express how amusing it is to hear a guy say that he wants professional coders and then complain that unit testing gives him more documentation to read. The horror!
That reminds me, I need to call up M$ and tell them that I want a patched version of XP that doesn't include any help documents. Real coders don't need documents.
If you have this TA, change your class now.
Next week, Steve Ballmer reviews "The Road Ahead"
Um, this is a joke, right? You teach at Harvard!!??? Now I know why we're losing all our jobs to offshoring. If this the best 'teaching' our educational system produce, we're really in trouble.
Seriously, it sound like your objections are just based on laziness - you don't want to look at extra code, documentation, etc. Too much work for you. Oh, that's another reason we're losing to offshore folks - laziness.
I guess we're illogical too. What does documentation have to do with unit testing? Can't you do one without the other???
Sorry, I'm just flabbergasted.
Stephen Molitor steve_molitor@yahoo.com
I've tried my hardest to get the profs to teach i386 assembly and then C so that students understand how a computer works, but to no avail.
Teaching i386 then assembly would be a sure-fire way to mentally scar and/or fail 99% of your students. You simply can't presume that all students have the level of competency in computing that learning i386 requires when they first start a computing degree.
This competency can be picked up, and I'm not arguing that both assembly (to a certain degree) and a lower-level language like C are necessities for any good computing graduate, however a nicer ride at the beginning of the road is also a necessity. Giving someone an i386 reference manual and telling them to code won't produce more professional coders, it will instead drive away bright students who perhaps just don't have the necessary computing skills yet.
my UID is lower than yours
Apparently you beat me in the IQ department too.
anywho, your students are going to end up as PHBs & the rest of us would appreciate it if you didn't turn them off to testing before unleashing them onto the working world. Thanks.
Technology Consulting & Free Downloads
Every day I am reminded why the universities consistently fail to produce programmers who can work in the real world (assuming this wasn't just a troll).
In the real world, code isn't handed in, marked, and that's that. Code is checked in, becomes part of a large, complicated system, and sometimes goes unvisited until years later, when entirely different personel have to take a gander at it and figure out what the heck it does, and why it was designed the way it was.
The more comments and documentation, the better.
In my current job, I'm in the middle of trying to untangle a badly-designed subsystem so it can be redesigned for new applications. There's no documentation, no unit tests. Only very high level descriptions of the I/O. The business logic is buried deep in the code, costing the company weeks of my time sifting through code. Some documentation on the outside would have cost a little bit more then, but saved a whole lot now.
Furthermore, if your interfaces between subsystems are well-designed, JUnit should be all you need to warrant that YOUR code will interact with the rest of the project according to the specification. And if there's a problem, it's the specification's fault, not yours. Granted, JUnit isn't really used class-by-class, but it is very useful a step above that, when it comes to integrating your code into a large project.
The problem in the academic world is there is no concept for how hugely huge projects get in the real world. No one is taught proper methodologies and good software engineering. Every assignment is a microcosm, and the short, quick, bang-out-the-code-and-ask-questions-later technique is what gets promoted in the process.
I'm sure there are schools that aren't like this, and I'm making sweeping generalizations, but this is what I have seen.
You know that instead of having to read all of this code you could just write a unit test yourself and determine pass/fail according to that. Then you wouldn't have to actually read (poor thing) the stuff that is given to him to read...
Of course that would require the thought process of a real programmer. Seems the Ivy Leagues aren't all they were cracked up to be. I won't be hiring any Harvard graduates for quite some time...
Well, I teach at Harvard also... And I'm glad I haven't run into you! Seriously, such arrogance has no place in a teaching situation.
1. Extra bloat my eye! What you're saying here is that you're a lazy bum and can't read through extra class files. If you want to do something constructive then how about instructing on how to lay out a project intelligantly?
2. Too much documentation? Again you're just a lazy TA and have no place grading. Regardless of the merits of HTML as a documentation medium, documentation is usually much more useful in grading than the actual code: it shows the thought process behind the code. Sure, I'm interested in that the code works as advertised, but I'm also interested in how you got there. It makes my job much easier to keep my students from falling into common pitfalls.
3. I do agree here, but not as far as the basic premise that unit testing is bad. Bad tests are useless, while a good test can be rather helpful, especially in a regression situation. Teach your students how to write good test and their code will be better. Punish them for writing bad tests and they'll never test.
Obviously you're not too proud of your convictions, or you wouldn't be an AC. I feel sorry for any professors you work with. I know mine are happy to work with me...
RM
}#q NO CARRIER
public void testPow() {
assertEquals( "Unexpected Math.Pow result", Math.pow(2), 4 );
}
Yeah, I can create a verbose, unreadable and unrealistic example in Perl, too. It doesn't make the language bad, rather it reflects badly on the programmer.
My business: Farstrider Studios.
You're totally clueless, and don't want anyone to know? ;-)
I'm a teaching assistant (TA) at Harvard University, and I mark a lot of Java code as part of a job that supports my PhD studies. Code submitted to me with unit tests face an automatic 5 mark penalty, before i even begin marking. There are several reasons for this.
You are a total moron...of course not many go to Hahvahd for CS studies. I wonder what your counterparts at real engineering schools like MIT or CalTech have to say about your viewpoint... Actually my best guess (and hope) is that you're a troll trying to land fish such as myself... ;-)
Extra code == extra bloat. The unit test code is submitted as an extra class file, which means there is one more thing that I have to read over before I can judge their code on its merits.
Oh, so the good programming practices of your students make your life more difficult - so you penalize them? You should be fired.
I find that unit tests are a pain to compile, normally found deep in a directory structure where they make little sense.
Of course the unit tests must be at the proper place in the directory heirarchy, or they wouldn't compile (Java namespace implementation). So, I doubt you know what you're talking about.
Too much documentation. People who submit unit tests generally document the interfaces using javadoc, which just increases the amount of reading I have to do before I can assign a mark. HTML as a documentation format is stupid. There is nothing wrong with submitting code with comments at the start of a function or method and leaving it with that. If the code needs documentation because it is doing something non-obvious, it is time to rethink the code.
Er, Javadocs aren't written in HTML...HTML is generated from Javadocs. As to your other remarks regarding code documentation, obviously you've never done any real-world programming. That is often a problem with PhD. candidates.
Unit tests are written with no extra skill than the code they test. Unit tests induce a level of confidence in the programmer that their code works as a black box. This is a bad idea. Programmers should understand why each and everyone of their modules works the way it does, not assume that everything is well because it passes some rudimentary test that was created with the same skill level as the rest of their code.
Code should work like a black box. That is the exact idea of encapsulation, a primary tenet of OO programming. You are quite obviously a troll with nothing to contribute to this discussion.
As you can see, I'm against unit testing. I've tried my hardest to get the profs to teach i386 assembly and then C so that students understand how a computer works, but to no avail. If we are to regain our jobs from the Indians, we must be more professional coders.
Ah yes, the final prong on the hook. How many of those Indian coders do you think do "i386 assembly" and "C"? Just around 0%...and don't fool yourself into thinking that either of those disciplines will do much to make entry-level coders better programmers. The important concepts are algorithmic, numeric and complexity oriented. Programming language isn't so important...and the more expressive and less bug-prone the better. Java is a pretty good choice, and has the added benefit of being widely used in the "real world" - a place you're obviously not too familiar with yourself.
Academia. Land of TA's who want to do as little work as possible to earn that money that supports their studies.
Your 1,2,3 are all faulty if applied to the working world. Which, by the way, is where real life is.
Your last sentence has a blob of irony in it. Can you spot it?
Come on, people, relax. The above comment (#8433108) is obviously a Troll. ...
It is very well crafted, but it does show all the characteritics of a Troll. Honest tone, but inflamatory arguments. Personal testimony (supposedly), but anonymous post. 'Insightful' opinions, but lack of touch with the actual matter etc
My friend wrote it! And I proofread it! And I'm submitting it to you! Strangely, I really liked it! Really!!
I'm posting anonymously for obvious reasons.
I'm a teaching assistant (TA) at Beaver College, and I mark a lot of Java code as part of a job that supports my PhD studies. Code submitted to me that works faces an automatic 5 mark penalty, before i even begin marking. There are several reasons for this.
1. Working code == extra effort. Code submitted that actually runs means that I have to read over before I can give the code an arbitrary, uninformed grade. I find that working code makes me have to figure out how to compile it, which is terrible since I have no clue what I'm doing.
3. (I can't count.) Working code is written by people who are more skilled than I am. Working code induces a level of confidence in the programmer and they think they are better than me. This is a bad idea. Programmers should understand why each and everyone of their modules shouldn't work, and not assume that everything is well because it actually does what it was documented to do.
As you can see, I'm against working code. I've tried my hardest to get the profs to teach abacuses and then flip lots of little switches so that students understand how hard it is to make a computer work, but to no avail. If we are to regain our jobs from the Indians, we must be more arrogant a-hole coders.
I've found it very difficult to write good tests on applications that were web-based, and highly dependent on the data. Does this book have good solutions for these problems?
Responding to the post above, there are a number of 'hidden benifits' to writing Unit tests for Real Programmers (and not Teacher's Assistants).
1) It's fun (sad but true). I write my test cases first, and that gives me a chance to challenge myself. I'll try and write tests that are very comprehensive or hard to pass (but only if they adhere to the specification). I then hide the tests while I implement the code to see how much I remembered.
2) It helps me seperate the notion of System code from Client code. This is pretty important because I'm often both the system architect and the client coder, and it's often hard to keep the two distinct when you're doing both.
By writing the tests first, you are forced to think of the (unwritten) code as a client of the system. You ask yourself, 'How should this module work? What are the most simple and logical interfaces?' You then write code to those interfaces, and end up with clean architecture. You then implement the code to pass the tests.
While that doesn't sound revolutionary, I find that without tests I have a tendency to let the logic dictate the interface. For example, I find myself thinking, 'this method would be really efficient if rather than receiving an abstract widget, it took 4 int args' (or something like that).
There are some other benefits to Unit testing (code stability, confidence, etc), but those two were surprising 'discoveries' when I started unit testing.
Kent Beck's book Test-Driven Development does a nice job of presenting where unit testing fits into Extreme Programming. It comes off as a bit of a cookbook and some of the examples are in Smalltalk, but even so I consider it a valuable addition to my bookshelf.
The net will not be what we demand, but what we make it. Build it well.
Don't worry you young students... the PHB's don't read your code, and your coworkers will hate you for writing it after you're gone.
Do unit testing. Do simple coding. Do it for you. It's good for your soul. It'll save your career.
Case in point - I was working on ASP pages that get user ID/password information from a db before validating to the next layer. I refactored the entry page so that I could test all the pieces parts since we have different clients who log in through the same page and get taken to other places. Anyhow, I typed in our demo account and password and got in - had about four tests... so then I tried four different variations on the password, camel case, and an extra space at the end thinking it would pass. Wrong. The new tests all bombed and let the person in. Why? Because SQL is set by default to be case-insensitive, and the code only checked to see if there was a matching username/password combination in the table. One line of code checking the actual password coming back fixed a bug we had in the system since inception. And who knows what else is in there.
If it doesn't fit on a screen, then break it up or throw it away. It's not extreme - it's the only way to make sense of a large system. If you write code at that level, you only need Javadoc to tell why you wrote the story the way you did...
You'll be much happier. Ignore the TA from Harvard and make yourself better. You don't need to be extreme to do so.
J.
This space for rent.
Unit testing does not require eXtreme Programming. On the other hand, eXtreme programming does require unit testing. We've studied XP's practices, and they seem to be roughly broken into two groups: programming for ever-changing requirements, and quality assurances (refactoring, unit testing) to make sure the code doesn't fall apart.
As many anti-XP pundits will point out, unit testing is worth applying to any project that you can apply it to, and there are precious few that can't.
I do a lot of framework-level programming, and I must say that stability, as well as my confidence in the code, has increased substantially from putting unit testing into the mix. Some folks advocate TDD (Test-Driven Development) in the canonical sense (test first), but I find that the completely up-front test creation works in practice only when you have rock-solid easy-to-translate-into-source-code checklisted requirements. I compromise, and build the tests after the interfaces, rough algorithm and rough support objects have been put together.
Here are a few things I've learned over my unit testing experiences:
I haven't figured out a good means to unit test concurrent code, but you can often stack the deck for fast-running processes by adding in delays in key spots to extend the exposure to weak spots.
Unit testing isn't for everything. It coughs and sputters a bit when you don't have full control over the environment (databases, sockets, network configuration, user interaction). You can shoehorn it in, but it's usually better left for the likes of regression testing and functional integrated testing.
For a quick start to functional testing, we've started test-driving TestComplete at our office; it's proven itself to be a capable tool (our lead QA man has put up a sign on the computer entitled, "I am Q-Tron; I Test Software While Mere Mortals Sleep" - it has been testing out communications while he sleeps :)
Binary geeks can count to 1,023 on their fingers