As an example, suppose you're converting an existing Ant build that doesn't follow the standard Maven project structure.  Maybe your project puts its source and test code right under src and test.

You start the conversion by creating a bare bones POM:

<project>
  <modelVersion>4.0.0</modelVersion>
  <groupId>com.mycompany</groupId>
  <artifactId>someproject</artifactId>
  <packaging>jar</packaging>
  <version>1.0-SNAPSHOT</version>
  <dependencies>
    <dependency>
      <groupId>junit</groupId>
      <artifactId>junit</artifactId>
      <version>4.7</version>
      <scope>test</scope>
    </dependency>
  </dependencies>
</project>

Then, you think: "Okay.  My project structure doesn't follow the defaults.  I'll want to change this eventually, but I want to see if this POM is okay.  I don't like to go long periods of time without seeing something working.  How do I tell Maven to change where it should look for source and test code?"

By running mvn help:effective-pom, you can find this quickly:

<project xmlns="http://maven.apache.org/POM/4.0.0"
  ...
  <build>
    <sourceDirectory>.../someproject/src/main/java</sourceDirectory>
    <testSourceDirectory>.../someproject/src/test/java</testSourceDirectory>
  ...

"Oh yeah.  There are the settings.  All I need to do is change this in my POM"

<project>
  <modelVersion>4.0.0</modelVersion>
  <groupId>com.mycompany</groupId>
  <artifactId>someproject</artifactId>
  <packaging>jar</packaging>
  <version>1.0-SNAPSHOT</version>
  <dependencies>
    <dependency>
      <groupId>junit</groupId>
      <artifactId>junit</artifactId>
      <version>4.7</version>
      <scope>test</scope>
    </dependency>
  </dependencies>
  <build>
    <sourceDirectory>src</sourceDirectory>
    <testSourceDirectory>test</testSourceDirectory>
  </build>  
</project>

The help:effective-pom goal is a trick I use to quickly look up a project setting as well as to see all the settings together for a given project or module.

public class Person {
    private final String firstName;
    private final String lastName;
    private final int age;

    public Person(String firstName,
                  String lastName,
                  int age) {
        this.firstName = firstName;
        this.lastName = lastName;
        this.age = age;
    }
    
    public String getFirstName() {
        return firstName;
    }

    public String getLastName() {
        return lastName;
    }

    public int getAge() {
        return age;
    }
}

Sure, it didn't take me too long to write the Java code because of the handy dandy source code generation features in my IDE. But the real problem is the maintenance. I or someone on my team will have to read this cluttered code many more times than the one time I wrote it.

Take a look at the equivalent Scala code:

class Person(val firstName: String,
             val lastName: String,
             val age: Int)

Which version would you rather maintain?

Some of our industry's thought leaders have already made contributions.  If you'd like to contribute and possibly see your name listed alongside them, see How to Become a Contributor.

I immediately thought of Crap4j; another tool that combines a set of metrics for identifying the riskiest areas of the code base to maintain.  Crap4j implements the CRAP metric, which combines cyclomatic complexity and test coverage, but at a method level.  If a method is both complex and not very well tested, then it's risky to change.

This all led me to a new, more ultimate set of metrics that could be combined to home in on the riskiest areas of a code base:

1. Code coverage
2. Cyclomatic complexity
3. Code execution frequency in the real world

Complex code, executed very often, with low test coverage.

For practical purposes, I like sticking with the granularity of a method.  I can use tools like Cobertura to find the test coverage and JavaNCSS to find the cyclomatic complexity.  (Isn't cyclomatic complexity best applied to the method level anyway?)

That just leaves me with the which-methods-execute-the-most-in-production problem.  This is hard because I can run the other two as part of a continuous build, but I won't be able to identify the hot methods until I get to production and measure true usage.  So the static and dynamic metrics will always be out of sync at some level even if I could get estimated usage through continuous functional and higher level testing.

So what can I do?  I want this to run as part of a continuous build to get feedback as soon as possible that a method is getting a little risky (or with a legacy code base, is already risky).  So, I'll fall back on afferent coupling for practicality.  But afferent coupling is typically measured at a package level.  The finest granularity that I'm aware of with current tools is measuring at a class level with ckjm.  That's a good starting point for identifying highly used code.

So here's my plan.  Use the CRAP metric to find the methods and then factor in the afferent coupling of the those methods' classes to give a prioritized list of methods to go clean up.  I'll see how this goes and consider factoring in method execution frequency from higher level testing runs.

My contribution was about starting with a Walking Skeleton and building it out.  I was lucky enough to learn this early in my career while working alongside a gentleman named Bernie Thuman.  Bernie applied this technique as he led a team developing a 3-tiered distributed enterprise application.

Anyway, I'm glad to be a part of this project.  The book is filled with pearls of wisdom and is a must read for any software architect or really any professional looking to develop better software.

I want to talk about essence vs. ceremony in unit testing.

If I give you this:

OrderTest
    testCalculatePrice
        1 20.00
        2  5.00
          30.00

Can you tell what this is about? This is a specification of the calculatePrice() method on an Order object, where there are two LineItem objects and the expected price is $30.00. Did you figure that out without the explanation? This is essence.

Instead, what you often see is something like this:

public class OrderTest { 

    @Test
    public void testCalculatePrice() {
        Order order = new Order();
        Product product1 = new Product(20.00);
        LineItem lineItem1 = new LineItem(1, product1);
        order.add(lineItem1);
        Product product2 = new Product(5.00);
        LineItem lineItem2 = new LineItem(2, product2);
        order.add(lineItem2);
        assertEquals(30.00, order.calculatePrice());
    }
}

Here, you're blinded with irrelevant details that obscure what this test is all about. The essence of the specification is buried in details. What you want is clarity at this level.

You can factor out the obscurity with something like this:

public class OrderTest {
    private Order order = new Order();

    @Test
    public void testCalculatePrice() {
        lineItem(1, 20.00);
        lineItem(2, 5.00);
        expectPrice(30.00);
    }
}

That's gets us about as close as we can in Java.

A consequence of this approach is that you have more methods overall, but I've always felt that clarity trumps in specifications. Others have agreed.

What's interesting with these helper methods is that I can change how I implement them and not change the higher level, essence methods at all. For example, in lineItem() I can inject real production objects (as I did in the original example with LineItem and Product), or I can inject a mock or stub LineItem.

So my suggestion is when you are specifying these high level methods, attempt to only show the essence of the test and factor out the ceremony.

Ted Neward: "Your peer group is more important than any tool or book we can recommend."  Great advice.

David Bock: "The existence of the system changes the requirements of the system."  Dave was talking about how seeing the system run changes the customer's mind of what he really wants.

David Bock: "Never believe someone who tells you he's 90% done."  Ever notice how that last 10% takes a long time?

I look forward to next time.

The reality is that properties files are the simplest and easiest way to configure an application, particularly for non-developers.  What I'd like to see is a PropertyPlaceholderConfigurer that can configure an application per environment.  This is what SPR-1876 is all about.  However, it suggests there be a file per environment.  As an alternative, I suggest a single properties file.  A single file containing all the properties would be the simplest to maintain.  Think of it.  I've got a new property.  Do I want to add it to three files or one?  If I want to see all the possible values of a property, do I want to open and hunt through three files or one?

Grails supports the idea of grouping all the properties together, but it uses multiple configuration files (Config.groovy and DataSource.groovy) depending on the property.

Okay, so here's what I propose the properties file look like:

# this is the default value for all environments
my.prop=defaultValue
# but in the production environment, it is different
[production].my.prop=productionValue
 
# here's another property with no default
# (adapted example from the Grails user guide)
[development].dataSource.url=jdbc:hsqldb:mem:devDB
[test].dataSource.url=jdbc:hsqldb:mem:testDb
[production].dataSource.url=jdbc:hsqldb:file:prodDb;shutdown=true
 

The context definition simply uses the property name without the environment and the custom per environment PropertyPlaceholderConfigurer provides the appropriate value:

<bean id="dataSource"
      class="org.springframework.jdbc.datasource...">
  <property name="url" value="${dataSource.url}" />
</bean>

So what are the downsides to this approach?  I suppose you can argue that a single properties file could be cluttered with details you don't care about.  For example, a production administrator may not care to see all the development and test values.

The other downside I can think of is forgetting to override a property.  That is, nothing reminds you that you have to override the DataSource URL for production.  It's content with using a default value.  The separate file approach would handle this because the property would be undefined.

So if I were reading this I'd say, "Spring is open source dude.  And those Spring guys are busy.  Go implement it and attach it to SPR-1876."  And if I get some time...

Apprentice:  Journeyman, I'm having trouble writing the fixtures for my JUnit tests.  They take too long.  I spend about 10 times longer writing tests than the production code.

Journeyman:  Really?  Unit test expert Gerard Meszaros says that tests should take only 10% to 20% of development time.

Apprentice:  Well, I'm experiencing the exact opposite.

Journeyman:  Let me take a look.

Apprentice:  Sure.

Author Note:  This is a trivial example for this blog entry.  I've seen some really obscure test setup methods.  The example is based on the sequence diagram chapter in Martin Fowler's UML Distilled book.

public class OrderTest {
 
     private Order order;
 
     @Before
     public void setUp() throws Exception {
         Product product1 = new Product(11.00);
         OrderLine line1 = new OrderLine(2, product1);
         
         Product product2 = new Product(22.00);
         OrderLine line2 = new OrderLine(3, product2);
 
         OrderLine[] lines = new OrderLine[] {line1, line2};
         order = new Order(Arrays.asList(lines));
     }
     
     
     @Test
     public void priceShouldBe88() {
         assertEquals(88.00, order.getPrice());
     }
 }
  

Journeyman:  Okay, can I see the Order.getPrice() method?

    public double getPrice() {
         double price = 0.00;
         
         for (OrderLine orderLine : orderLines) {
             Product product = orderLine.getProduct();
             double productPrice = product.getPrice();
             int quantity = orderLine.getQuantity();
             price += productPrice * quantity;
         }
         
         return price;
     }
 

Journeyman:  Ah, I see the issue.  Have you ever heard of the Law of Demeter (LoD)?

Apprentice:  Huh?

Journeyman:  The LoD has a few different names and related principles, but basically it means that an object should only deal with and only with its collaborators.

Apprentice:  Huh?

Journeyman: You see how Order iterates through its OrderLine objects?

Apprentice:  Yeah.

Journeyman:  Well, that's fine.  But then you grab the Product and then grab its price.  You've violated the LoD.  Order should only know about OrderLines.  You've got more of a procedural design here, where you reach down into all the objects in the graph, grabbing all the data you need, and then do your thing.  A more object oriented design would be to distribute the work.  Do a little bit of work and delegate the rest to collaborating objects.

Apprentice:  Oh, I think I'm beginning to see.

Journeyman:  You should go read up on Craig Larman's GRASP principles.  He calls the LoD Don't Talk to Strangers.

Apprentice:  Okay, will do.

Journeyman:  Have you been following test driven development (TDD)?

Apprentice:  Uh..., well..., not really.  But I do write the tests afterwards.

Journeyman:  Complex test fixtures jump right out at you when you're doing TDD.  You realize something is wrong right away.  In fact, many times your testing style changes into more of an interaction based style.  But I'm getting ahead of myself.  Let's refactor this together.

Apprentice:  Okay, cool, pair programming.

Journeyman:  Right.  Remember, Order should only deal with OrderLines.

Now, the setUp() method looks like:

     @Before
     public void setUp() throws Exception {
         OrderLine line1 = stubOrderLineGetPriceToReturn(22.00);
         OrderLine line2 = stubOrderLineGetPriceToReturn(66.00);
 
         OrderLine[] lines = new OrderLine[] {line1, line2};
         order = new Order(Arrays.asList(lines));
     }
 

And the Order.getPrice() method looks like:

    public double getPrice() {
         double price = 0.00;
         
         for (OrderLine orderLine : orderLines) {
             price += orderLine.getPrice();
         }
         
         return price;
     }
 

Apprentice:  Okay, now I really see.  Order and its test are simpler, but don't we have the same amount of work anyway?  We had to write OrderLine.getPrice() and test that.

Journeyman:  Yes.  The work is distributed.  But I prefer lots of simpler objects with simpler test specifications than more complex objects and tests.

Apprentice:  What do you mean, more objects?  We didn't create any new objects.

Journeyman:  I meant generally speaking, not necessarily in this case.  Like I said, do a little work and pass the buck.  That might mean creating Data Clumps, aggregate objects for collections, objects that represent abstract data types, and so on, but again, I'm getting ahead of myself.

Apprentice:  Should I always follow the LoD?

Journeyman:  Well, there are of course consequences.  Classes tend to have larger APIs because sometimes they need to wrap the functionality of collaborating objects.  And you need to step through more objects to understand an algorithm as a whole because it's distributed.  But most of the time, I follow LoD because dependencies are reduced.

Apprentice:  All right, I'll give it a shot.  Thanks.

It's inevitable. As more and more features are added to an application, the code base grows. The build has more to do: more compiling, more tests and metrics to run, more reports to generate. We've done numerous things to keep the build time down. In this entry, I'd like to describe one of them: running standard JUnit tests concurrently.

A colleague of mine came up with this idea. He first did a preliminary search to see if anything already existed.  Pretty much everything he found was intrusive, e.g., requiring extending a specialized TestCase.  He then attempted to combine Ant's Parallel and JUnit tasks. That is, he could simply nest N number of JUnit tasks inside a Parallel task. The problem here was coming up with a good way to divide the tests up into equal parts beforehand. My colleague quickly scrapped this idea in favor of a custom Ant task that wrapped the JUnit task.

The thought was to leverage the JUnit task as much as possible, but fronting it with the ability to run tests in parallel. The conceptual design looks like this:

There are N number of TestExecutors. A TestExecutor runs in a single JVM. Thus, there are N JVMs. Say we have a dedicated, 4 processor integration machine running builds. We may choose to configure our custom task with 5 TestExecutors (we'll add one assuming we're not 100% compute bound). Note: if you're familiar with the JUnit task, you may be wondering how its fork/forkmode attributes fit in. The answer is that they are eliminated in favor of this jvm-count attribute.

There's a single TestProvider. His job is to gather all the tests and hand them out one at a time to whichever TestExecutor is ready. A simple protocol exists to let a TestExecutor tell the TestProvider that he's done and ready for another test.  This should maximize parallelism.

The biggest negative with this approach is that since all tests run in a fixed number of JVMs, there's some loss of isolation here. Class level/global state set from a previous test could affect later tests. We're willing to trade this for increased performance.

The other possible downside is the use of a non-standard Ant task.

Using this approach we've successfully reduced our build time to usable levels.

As I described in a previous blog entry, I've had problems in the past with people breaking the build. To help, I've used techniques like the rotating Build Nazi and the put a dollar in the broken build jar, but these are all negative focused. How about something that rewards developers that don't break the build? How about rewarding developers for following the best practice of breaking their work into smaller chunks and checking in early and often?

So I'm thinking of a game where a developer gets, say 1 point for getting his name on a successful build. The number of files checked in is ignored. We want to discourage big check-ins, so you get more points for smaller grained check-ins since your name will show up on more successful builds. And on the other side, you get points taken away when you break the build. Now, we want to keep things simple, so we could probably stop right here, but I'm thinking that some failures are worse than others. So maybe we have something like:

It's a game, so we need something for the winner. Bragging rights is too lame, so maybe lunch on the team, or some kind of trophy kept on the winner's desk to remind everybody of his champion status.

Now, there are some negatives. Perhaps not everybody would want to play. Particularly, notorious build breakers wouldn't want everybody (specifically management) to see their poor results. In that case, I suppose we could only publicly display the leaders, or top half, but that wouldn't be as fun.

People could easily cheat too. Maybe, write a cron job that every hour checks out a file, changes a comment and checks back in. We'd have to look out for that kind of thing.

What about the analysis time required to keep score? I could easily see how a post processing ant task could be developed to update the points for developers on a successful build. But for a failure, I think you'd need human analysis. That's a negative because it requires time, so the job could be rotated. On the plus side, what I've noticed is that analyzing why the build failed brings awareness to issues. Issues like some people needing training, or a test requiring change because it's non-deterministic, or a hole in the pre-check-in process used to ensure a successful build.

To keep the game fresh and the developers motivated, we'd have to reset the points periodically. Iteration boundaries seem appropriate here.

Well, maybe I'll give it a shot and see what happens...

I always give them the example of a circle.  Here’s some Groovy code:

class Circle {
    def radius
    
    def getDiameter() {
        2 * radius
    }

    def getArea() {
        Math.PI * Math.pow(radius, 2)
    }
    
    def getCircumference() {
        2 * radius * Math.PI
    }
}

radius is a field, also known as an instance variable, also known as a member variable in C++.  It’s an implementation detail of the Circle, i.e., it’s a private variable that gets stored as part of the Circle object.

Properties are more public aspects of an object.  In Java, following JavaBeans conventions for getters and setters allows you to expose properties.  Those properties don’t have to be backed up by fields, but in many cases are.

Coming back to the Groovy Circle, what are its properties?  We’ll you get radius for free as a read-write property.  But because we’ve got some getters that compute other values based on radius, we also have diameter, area, and circumference properties.  These happen to be read-only properties since we didn’t provide setters, but we could have (which would update the radius field accordingly).  You can dump all of a Circle object’s properties in Groovy like this:

println myCircle.properties

The point is Circle could be implemented in different ways: by using a diameter field instead of radius, for instance, but it would have the same properties.

Now, let’s talk about attributes.  I typically think of UML when I hear attributes.  Here’s UML for Circle:

Notice that I’ve distinguished the derived diameter, area, and circumference attributes with a “/”.  This is standard UML.  radius is a regular, non-derived attribute.  We can distinguish fields from other computed properties like this.  However, because we’re modeling, and thinking in a higher level of abstraction, I like to think of attributes as more synonymous with properties.  I’d rather not make the assumption that every attribute is implemented with a field.

A similar discussion of these terms can be found here.

Apprentice:  Hey, can I ask you something?
Journeyman:  Sure.
Apprentice:  I was looking at a big constructor on one of our domain objects and I saw some groupings and was wondering whether it would be good to create objects for those groupings.  Is there some kind of pattern for that?
Journeyman:  Well first, how many parameters are there?
Apprentice:  72.
Journeyman:  72!?  Wow.  Talk about a code smell .  PMD's ExcessiveParameterList rule will flip out on this one.  I guess the author mapped those parameters straight from the out-of-our-control XML schema, whose document instances get unmarshalled in order to create the object.  We don’t have to follow suit and create such a flat domain model that matches the schema.
Apprentice:  What do you mean?
Journeyman:  Well, what I really mean is that I'd much prefer distributing the logic among some richer objects.  That is, create more little objects with little methods that do a little of the work.  Not one big object that does it all.  There would be an impedance mismatch between XML and the domain model, but that’s typical because of the technology differences.  I suppose the only benefit of the big constructor approach is the simpler mapping... maybe, but the negatives are much too great.
Apprentice:  What are the negatives?
Journeyman:  Well, for one, trying to list all 72 parameters in the correct order would be a pain.  And trying to understand the object as a whole with all those fields would be difficult.  Here [grabbing his Refactoring book].  You were initially asking about the “Introduce Parameter Object” refactoring.  Read that refactoring.  It can probably give you better details.
Apprentice:  Martin Fowler.  Does he know what he’s talking about?
Journeyman:  Oh yeah.  He’s one of the Masters.  Let’s take a look at the class.  [Brings up the class in the IDE]

  public ImportantDomainObject(

    ...

    int x, int y,

    ...

    double min, double max,

    ...)

Well, I can see a few data clumps already.  You see the <code>x</code> and <code>y</code>?
Apprentice:  Yeah.
Journeyman:  That’s probably a <code>Point</code>.  Also, that <code>min</code> and <code>max</code>, that looks like a <code>Range</code>.  What I would do is search for data clumps like these and create objects for them.  Then, see where they’re used in <code>ImportantDomainObject</code> and start moving behavior into the new objects.  Distribute the logic.  We can create a richer domain model this way.  You’ll probably find clumps of these clumps and can follow the same process.
Apprentice:  Okay.  This sounds great.  Thanks.  I’m on it.
Journeyman:  Thank you for recognizing this and taking the initiative to improve the code base.

• Weaken the preconditions of the base class, not strengthen them (contravariance).
• Strengthen the postconditions of the base class, not weaken them (covariance).

There was some confusion as to why this was the case. So I came up with an analogy to explain. If you think in terms of the client requesting services of an object specifying the preconditions and postconditions, it all makes sense.

So the analogy? Have you ever had to request a help desk ticket? Maybe you need administrative rights or something. And you hope that a specific someone is assigned to that ticket because she's really good and can get the job done quickly and accurately. Let's use that as an example.

The general IT support group contract is:

• Precondition: Submit a help desk ticket filling in all the forms on the web site.
• Postcondition: We'll get back to you within two hours.

Now, suppose Suzy Support works in the IT support group. She's really good and efficient. You can think of her as a subclass of the IT support group base class. In the past, whenever I've had a problem and I just happened, purely by chance, to pass her in the hallway and mention it to her, she could solve it in minutes. From my perspective, as a client of the IT support group, that's acceptable.

Suzy's contract is:

• Precondition: Just tell me the problem. You don't have to fill out a help desk ticket for me. This is weaker than the IT support group's precondition because I have less work to do.
• Postcondition: I'll solve it in minutes. This is a stronger postcondition. I get better response time.

Make sense?

By the way, for all the IT support group managers out there, this is just an analogy. I know bypassing the help desk ticketing system messes up metrics.

Ruby and particularly Ruby on Rails has been hot for awhile now. Ruby in the enterprise is becoming more accepted. As Martin Fowler pointed out nearly a year ago in this blog entry , some top minds in our field including Bruce Tate, Justin Gehtland, and Stuart Halloway have moved Beyond Java . Even one of my old mentors, Mike Gaffney , has repeatedly told me to bail on Java. But I agree with Scott Davis as he points out in this podcast , Java is not dead.

The growing acceptance of Ruby is a strong reason to consider JRuby. Being able to run Ruby on Rails apps on the JVM is very appealing. But remember the context of the decision here. We're talking about a group of developers whose only development language has been Java, and whose projects for the foreseeable future will run on the Java platform. Groovy wins in this context.

There are lots of reasons to get excited about Groovy. It too runs on the ever-improving JVM. In fact, it was the first scripting language officially approved by Sun to run on the JVM. It can be compiled to byte code or run in script form. The interoperability is fantastic. Developers can still use Java, its tools and familiar libraries such as Spring and Hibernate , where appropriate, and mix in Groovy, where that is more appropriate.

But what really makes this decision easy is the similarity in language syntax. You can take a snippet of Java code, drop it into a Groovy file, and it will nearly work. Then, you can take that snippet and start making use of Groovy language improvements such as closures to make the code more readable and intention revealing .

So the decision is easy here. Take the seamless step forward and introduce Groovy.

1. AtLeastOneConstructor: Why? Code is smaller without an empty, no-args contructor.
2. AvoidInstantiatingObjectsInLoops: This one just seems to generate too many false positives. In addition, for short-lived objects, garbage collection is essentially free.
3. CallSuperInConstructor: super() is called implicitly and requiring it just adds more lines of code.
4. JUnitAssertionsShouldIncludeMessage: Most of the methods in Assert already generate good messages. I do include a message for those that don't (assertTrue(), assertFalse(), fail()).
5. LocalVariableCouldBeFinal: final for variables is usually overkill and it adds line length.
6. LongVariable: Clarity rules!
7. MethodArgumentCouldBeFinal: Same as LocalVariableCouldBeFinal.
8. OnlyOneReturn: I think it's clearer to exit early. It reduces the amount of things to think about below the exit.
9. PositionLiteralsFirstInComparisons: This really isn't that bad, but myString.equals("x") just reads better than "x".equals(myString).
10. ShortVariable: There are just too many cases where it's acceptable to have short variables in small methods .

 These two just missed the cut:

• ShortMethodName: I try to make my names expressive. I've never seen this rule fire.
• SignatureDeclareThrowsException: This is an okay rule, but for JUnit tests, who cares?

I disagree with this stance, but I think Harold has a really good idea there, which I'll get to in a moment. I think breaking the build does include passing all the unit tests and should be part of the continuous integration process. I've always liked the idea of the tests serving as a safety net when I add new functionality or refactor existing code. The tests not only specify the desired behavior, but act as a regression suite.

So what's the idea? I first learned of the concept of test categorization from Cedric Beust's TestNG framework and the writings of disco king Andrew Glover . Essentially, the concept is to break tests into groups such as unit, component, and system and run them at different intervals. Unit tests are run many times a day as code is developed, component tests less frequently, and system tests the least frequently of all. You can categorize tests in several different ways such as by file naming conventions, placing them in different directories, or using Java annotations.

Harold brings up a new category - the todo category. (I was originally thinking the fixme category, but I thought todo could serve more purposes.) The todo category contains all tests that serve as a reminder of work to do. They obviously fail or there would be nothing to do. When it's finally time, the test moves out of the todo category and into the unit, component, or system category as appropriate. Although you could have a todo category for each of the original, regression categories (todoUnit, todoComponent, and todoSystem), I think that would be overkill. Notice that I slipped the word regression into "original categories" above. I'm advocating that the original categories should pass at all times - they should break the build if they fail (well, at a minimum the unit category). Once tests get into the regression categories, they never go back to todo.

The obvious downside of this approach and of test categorization in general is the added complexity. It's more work to maintain categories and in this case, move tests into different categories than a simple one category fits all approach. Also, there's more work (at least initially) to setup the running of those categories. As always, I recommend weighing the options and deciding what's best for your particular situation. See Glover's writings for more benefits of test categorization.

Let's quickly look at an example, the idea stolen from one of David Chelimsky's blog entries about specifying the behavior of a stack. We're focused here on the specification of an empty stack:

// notice the class name specifies the context
public class EmptyStack {
    private Stack stack = null;
    
    @Before
    public void setUp() {
        // set up the context
        stack = new Stack();
    }
    
    // notice the name focuses on the context
    @Test
    public void shouldBeEmpty() {
        assertTrue("not empty", stack.isEmpty());
    }
    
    @Test(expected=EmptyStackException.class)
    public void shouldComplainOnPeek() {
        stack.peek();
    }
    
    
    // more specification focused methods
}

The point here is that you can get many of the benefits of BDD (a focus on specification rather than testing) using the familiar JUnit framework.

Now, if you're a hardcore BDD'er, then you might complain that you still have to use a test-centric vocabulary. You still need those Test annotations and method calls like assertEquals rather than Dave Astels' preferred shouldEquals calls.

Also, from the legacy side of the fence, you lose the convention of method names starting with testSomething and class names ending with Test. It's sometimes hard to let go of that if you've been writing tests for a long time and it's super clear to spot those test methods and test classes if you're following a naming convention. Furthermore, Ruby on Rails has taught us that convention is a good thing. The test method naming convention doesn't really bother me so much. The @Test annotation makes things clear enough and shouldSomething really gets us focused on specification. However, I haven't let go of ending the class name with Test. Maybe it's more because Ant can pick those tests up easier if there's a standard naming convention.

The last negative I can think of is that of consistency. Having a mix of old style test centric tests and BDD style tests could bother some.

So, let's be honest. What am I really doing? Well, I am incorporating more and more BDD into my work. However, I still shy away from creating new classes. The BDD style really lends itself to many test classes (contexts) per class under test. In the case of Stack, as David Chelimsky points out, you'd also need AlmostEmptyStack, AlmostFullStack, and FullStack classes to fully specify the behavior. I just can't commit myself to writing all those. But I am focusing more on the set up of a context and the specification methods of that context. I just may cheat a little and combine contexts into a single class. So perhaps in the Stack example, I'd combine the empty and almost empty stack contexts under one test class. You know, I'd set up an empty Stack in setUp and for the almost empty context, I'd do a little more set up (push something) in the appropriate shouldSomething methods to make it almost empty.

I also revert back to legacy style testing sometimes. This is usually when I'm modifying an existing, old style test. It's just simpler and faster, and typically the context isn't set up too well for the BDD style.

So my advice is to use the right tool for the job. Use BDD style when that makes things clearer and you want to focus on specification. Use test centric style when that's easier. Try to do a better job of focusing more on specification and less on verification, especially when you're adding new behavior.

So if I had a method like this:

    private void someHelperMethod() {
    } 

I'd expose it like this:

    /** Exposed for testing purposes only. */
    void someHelperMethod() {
    } 

I'd laugh at people who used the <code>PrivateAccessor</code> of the JUnit-addons project or directly used reflection to get the method and do a <code>method.setAccessible(true)</code>. I'd think, "What's the big deal? Just change the access of the method. Then, you'll make things easier for your IDE when you want to do a method rename refactoring or just find references to this method."

I've recently hopped the fence and I'm now campaigning for the other side.

Now, before I get into the why, I should point out that red flags should have gone up after you read the first question of this entry. Testing private methods? Why would you do that? Believe me, as I've mentioned here , I think unit tests should focus on black box testing as much as is practically possible and therefore, you shouldn't be testing private methods. I'm also familiar with the argument, "You could perhaps move those private methods to a helper class and therefore, they wouldn't be private." But I still believe there are cases where it makes sense to (have and) test private methods. And in this blog, I'm talking about these exceptional cases.

So what are some exceptional cases? Think of a high level method specifying the steps of an algorithm or a method on a mediator/controller-like class collaborating with multiple objects. You did break that method up into smaller, easier to understand, more communicative methods as described here , didn't you? So which is easier: setting up the object under test and its collaborating objects (the fixture) and calling the private helper method directly; or setting up the fixture and calling the public, higher level method that does some processing before calling the private helper method? In the latter case, you surely had to do more set up to navigate down to the same entry point in the private method. If you didn't or it wasn't much more work, then I'm talking about a different context.

All right, so why the switch? Well, first off, the real intent is to make them private. Private leaves no doubt that the method isn't meant to be used outside the class.

Along with that is encapsulation. As multiprocessor systems become more mainstream, exploiting concurrency by multi-threading adds complexity to software development. Encapsulating as much as possible can only simplify reasoning about thread safety.

One final reason is that making things private helps static analysis tools like FindBugs and PMD better identify dead code that can be eliminated. "Hey, this method is private and it's not being called!" Keeping the code base lean and clean is a key to maintainability.

The biggest downside, as I've hinted at above, is doing things like renaming/changing the signature of the method, or just finding references to it. The IDE won't make the appropriate changes to the test or find the reference in the test. But you'll find this pretty quickly when you run the test and it fails. Hint: Maybe you should have changed the test first anyway.

So join me on this side and remember that testing privates is better meant for special cases.

So what can really irritate the rest of team? How about checking in a bunch of changes that break the build and then leaving for the day? This is bad enough for small teams, but consider bigger, distributed teams developing complex applications. You could have team members who start their day after you've left. If the team follows the rule of not checking in on a broken build, what's the team supposed to do? Well, to continue progress, somebody else has to step in and fix the build. This can really slow a team down.

A good practice is to wait for a successful build after your last check in before leaving. After all, you know the most about the changes that would have broken the build. This practice implies some prerequisites: a continuous integration system that runs often and builds that run quickly.

Now that I've ranted enough, what can you do to reduce the chances of breaking a build in the first place? Listed below are steps to follow. It starts at the point where you're ready to check in. Essentially, you mimic the continuous integration system on your development workstation.

1. Update your view so you have the latest versions of all the files.
2. If you use configuration management software that allows hijacking and reserving files (like Rational ClearCase), now is the time to "unhijack" and reserve existing files with changes. By existing files, I mean files the CM software knows about (not new). You only need to reserve files if your team follows that practice. If nobody on the team reserves files, then skip this part, but it's important to reserve at this point if your team follows this practice to ensure all of your changes can get checked in.
3. If necessary, merge any existing files that have been updated and checked in by others since you checked out. You want your files to represent what will be checked in. Now, your files will contain your changes merged with the most recently checked in versions.
4. Make note of any new files that need to be added to source control and any deleted files that must be removed from source control. Forgetting to do this can cause a build failure.
5. Run the application to make sure your changes function and integrate well.
6. Do a clean build, compile and run all the unit tests. Consider a single ant target that does this for you.
7. Check in. Don't forget those new and deleted files!

Do I follow all of these steps on every check in? No way. That wouldn't be practical. You need to consider the size and impact of your changes. The bigger they are, the more rigorous you should be. If I made just a small change, I would probably just make sure the code compiled and unit tests for that change passed.

Well, that's it. Hope you enjoyed.

Ideally, what should happen is the following:

1. Notification of a build failure is distributed by e-mail to the team. This is part of the publishing functionality of CI systems like CruiseControl.
2. Particularly, whoever checked in since the last successful build analyzes the results.
3. The person who broke the build mans up and replies to all, “This is me. I’m on it.”
4. Nobody checks in until the build is fixed unless the check in is for the purpose of fixing the build.
5. Somebody fixes the build, obtaining any help necessary.
6. (Optional) Once the build is believed to be fixed, the person from step 3 sends an e-mail saying the build should be fixed.

Let’s talk more about step 3. I know it could be embarrassing to break the build. The step is not to call anyone out or lay blame. The point is to communicate to the team, who may be large and distributed, that somebody is taking responsibility for fixing the build. Several others may be analyzing and trying to fix the problem and this communication is to reduce that wasted effort. I have been on teams where the build has been broken for hours and nobody knows if anybody is doing anything about it.

Sometimes nobody is doing anything about it. If this is a recurring problem, I recommend the use of a Build Nazi (in honor of the Seinfeld Soup Nazi). The job of the Build Nazi is to make sure somebody is responsible for fixing a broken build. The job is not to fix the build (unless help is needed). The Build Nazi job is not a fun one and therefore should be rotated. It is also a controversial role in an agile environment where the practice of collective code ownership is being followed. Ideally, the team is following the steps above and a Build Nazi is totally unnecessary. I’ve found I’ve had to resort to the role for short term periods in times of chaos when the build is breaking more than it is succeeding. Once things get back on track, the Build Nazi role typically becomes dormant.

The last step I’d like to elaborate on is step 4. One of the worst things to do is check in a bunch of changes that don’t have anything to do with fixing the build. Checking in causes the CI system to start another build that will result in another failure. Piling on check ins during a broken build prolongs the broken state of the build and therefore the feedback cycle that is all important in agile development. Another risk is that the fix necessary to correct the original problem must be applied to the check ins that occurred after the first failure, further prolonging the broken state.

I plan on finishing this series next time. Stay tuned.

The first obsolete practice is reserving (locking) a file by default when checking out the file. (I’m using ClearCase terms here). This is like pessimistic locking in the database world. The thought is, “I’m going to reserve this so that no one else will affect me. I’ll make my changes, then check in. If anybody else wants to modify the same file, they’ll have to wait for me to finish, then deal with merging in their changes.” That doesn’t sound like collective code ownership to me. Sharing the code base means sometimes having to deal with merge issues.

Reserving files prohibits others who are ready to check in. In the worst case, suppose a person with reserved files leaves for the day (or for vacation). Others are likely working different hours, so what do they do? They try to get in touch with you to see what’s up. They can’t get a hold of you, but they are ready to check in and move on. So after some time not hearing from you, do they unreserve the files? In some stricter environments, what if they don’t have rights? What if they can’t get in touch with the select few that have rights to unreserve files? (By the way, if you do unreserve a file, make sure you tell the person who reserved it that you did that.) The point is reserving files slows down the team.

There are of course exceptions. An important change or a change that has a big impact to the system might merit an early reserve on a check out. In this blog, I’m talking about reserving files by default.

The biggest positive of reserving early is finding out that somebody else is working on the same file. For example, I try to check out a file and reserve it right away and find that I can’t because Joe’s got the file reserved. Depending on the circumstances, I may want to know exactly what Joe’s doing. I’m not familiar with another way of getting this notification as easily without getting spammed by unwanted notifications. In other words, I might want to know, upon check out of a file, whom else has the file checked out.  For me, this advantage doesn't outweigh the disadvantages mentioned above.

Next time I’ll talk about an annoyance or two that occurs when a build breaks.