We Keep Coding

Unit testing a service, which works as a mediator between two other services

I have a service, which works as a mediator between two other services. It basically validates the inputs, then passes them to those two service sequentially (by trying to keep transactional integrity), and then, if everything goes well, saves a result to the database.
My problem now is to test this service in isolation. Of course, I can provide stubs to satisfy the dependencies. I can also test the validation of inputs, whether appropriate data is saved in the DB in a normal case, as well as whether transactional integrity is kept if any of the dependencies throws an exception.
Yet, this is only half of what the service really does. My dilemma is if I should try to prove whether the other two dependency services actually processed the data appropriately as well? The scope of my service is quite broad, so I guess it is better to also know if the dependency services also did their job well. Yet, this gets out of the scope unit testing, and moves into integration testing, right?
I am kind of confused here.

If you're asking about unit-testing, then the way to do it is to test the class in isolation using mocks or stubs.
BUT, if you feel that just doing that is not enough, you can write some component tests, where you use the all the real classes you want to test, and use a stub (or inmemory) database and mock some of the dependencies that you consider not important for what you are trying to test.
In the past, I've tested small clusters of classes that had a high interaction between them in this way (and sometimes skipping unit-tests for those classes, as the component tests covered all the scenarios). Obviously, the problem with doing this is that the number of scenarios grows almost exponentially the more classes you're testing. Maybe you can test the bridge and the 2 real classes that use that bridge.

You should do both.
For unit testing, definitely use mock-objects for dependencies, preferrably using a tool like EasyMock. As a sidenote, if you feel that the functionality of your mediator service is too broad for unit testing, you may want to consider breaking it down into smaller pieces.
Of course, you additionally should do integration testing as well, using real dependencies, to make sure your services work together as intended.

Struts2 action - test it or not?

I'm currently working on one project, that uses Struts2 framework. We use separate component for DB accessing, which is well tested. At the same time, project, that we work on has a lot of Actions, that are not tested. In most of the actions we use at least one DB-service call. So on one hand these actions are pretty simple. I'm not sure - should unit tests be written for that or not?
I think that good practice is write unit tests always, but these actions are so simple and I'm under big pressure from management side right now. So, is it critical or not - to leave Struts2 actions without unit tests?

Here are the three main reasons for writing unit tests.
It helps you know that your code works now.
It helps defend your code against regression errors, when functionality is added or changed in the future.
If you write unit tests before you write your code, it focuses your design process in a really good way - read up on TDD to learn more about the benefits of doing this.
So ask yourself whether any of these three reasons for writing unit tests apply here. If the answer is "no" for all three questions, then consider the cost of writing the unit tests, and of keeping them in your code base. Compare this cost with the possible benefit. Make an intelligent decision about whether you should be writing unit tests, and be prepared to defend that decision to your manager.
But don't carry a preconceived notion that "unit tests are always good, for every class". And don't carry the opposite notion - that "unit tests are always unnecessary". Neither is true.

I'm in the same camp as Dhanji Prasanna who worked on Guice and Google Wave. Its not about 100% coverage, its about writing valuable tests that provide the right feedback for the right components in a way that aids development and protects against code regression.
For one of my Struts2 apps, we had very very complex data validation requirements. Thousands. We used the struts2-junit-plugin to test the action classes within their integrated context with Spring 3 IoC and Struts2 validation and a custom mechanism for populating mock requests with lots of different data scenarios. These tests were/are invaluable both during development and as a maintenance tool.
But for some of our simpler actions, I don't see much value coming back compared to the time spent to write them. But then, they don't take too long to write if they are very simple, either.
I have also seen cases where the 100% coverage notion led to 100% of classes having thoughtless, worthless tests being written for them. For my money, I vote for identifying the areas where the tests will provide the most value up front and focusing on doing those very well.

Must write Unit Test for functions could be problematic, but in anyway maybe in a future will be validations in the Actions that would be good can test that.
The time spend for test the actions must be a little bit and I would recommend do it, every tier in your app must have some functionally if not is a unnecessary and must review the architecture.

Java/J2EE standard practices and design choices

I have a couple of design/architectural questions that always come up in our shop. I said "our", as opposed to "me" personally. Some of the decisions were made and made when J2EE was first introduced so there are some bad design choices and some good.
In a web environment, how do you work with filters. When should you use J2EE filters and when shouldn't you? Is it possible to have many filters, especially if you have too much logic in them. For example, there is a lot of logic in our authentication process. If you are this user, go to this site and if not go to another one. It is difficult to debug because one URL path could end up rendering different target pages.
Property resource bundle files for replacement values in JSP files: It seems that the consensus in the Java community is to use bundle files that contain labels and titles for a jsp parsing. I can see the benefit if you are doing development with many different languages and switching the label values based on locale. But what if you aren't working with multiple languages? Should every piece of static text in a JSP file or other template file really have to be put into a property file. Once again, we run into issues with debugging where text may not show up due to misspelling with property value keys or corrupt property files. Also, we have a process where graphic designers will send us html templates and then we convert them to jsp. It seems it more confusing to then remove the static text, add a key, add the key/value in a property file, etc.
E.g. A labels.properties file may contain the Username: label. That gets replaced by some key and rendered to the user.
Unit Testing for all J2EE development - we don't encourage unit testing. Some people do but I have never worked at shop that uses extensive unit testing. Once place did and then when crunch time hit, we stopped doing unit testing and then after a while the unit tests were useless and wouldn't ever compile. Most of the development I have done has been with servers, web application development, database connectivity. I see where unit testing can be cumbersome because you need an environment to unit test against. I think unit test manifestos encourage developers not to actually connect to external sources. But it seems like a major portion of the testing should be connecting to a database and running all of the code, not just a particular unit. So that is my question, for all types of development (like you see in CRUD oriented J2EE development) should we write unit tests in all cases? And if we don't write unit tests, what other developer testing mechanisms could we use?
Edited: Here are some good resources on some of these topics.
http://www.ibm.com/developerworks/java/library/j-diag1105.html

Redirection is a simpler way to handle different pages depending on role. The filter could be used simply for authentication, to get the User object and any associated Roles into the session.
As James Black said, if you had a central controller you could obviate the need to put this logic in the filters. To do this you'd map the central controller to all urls (or all non-static urls). Then the filter passes a User and Roles to the central controller which decides where to send the user. If the user tries to access a URL he doesn't have permission for, this controller can decide what to do about it.
Most major MVC web frameworks follow this pattern, so just check them out for a better understanding of this.
I agree with James here, too - you don't have to move everything there but it can make things simpler in the future. Personally, I think you often have to trade this one off in order to work efficiently with designers. I've often put the infrastructure and logic in to make it work but then littered my templates with static text while working with designers. Finally, went back and pulled all the static text out into the external files. Sure enough, found some spelling mistakes that way!
Testing - this is the big one. In my experience, a highly disciplined test-first approach can eliminate 90% of the stress in developing these apps. But unit tests are not quite enough.
I use three kinds of tests, as indicated by the Agile community:
acceptance/functional tests - customer defines these with each requirement and we don't ship til they all pass (look at FitNesse, Selenium, Mercury)
integration tests - ensure that the logic is correct and that issues don't come up across tiers or with realistic data (look at Cactus, DBUnit, Canoo WebTest)
unit tests - both defines the usage and expectations of a class and provides assurance that breaking changes will be caught quickly (look at JUnit, TestNG)
So you see that unit testing is really for the benefit of the developers... if there are five of us working on the project, not writing unit tests leads one of two things:
an explosion of necessary communication as developers try and figure out how to use (or how somebody broke) each other's classes
no communication and increased risk due to "silos" - areas where only one developer touches the code and in which the company is entirely reliant on that developer
Even if it's just me, it's too easy to forget why I put that little piece of special case logic in the class six months ago. Then I break my own code and have to figure out how... it's a big waste of time and does nothing to reduce my stress level! Also, if you force yourself to think through (and type) the test for each significant function in your class, and figure out how to isolate any external resources so you can pass in a mock version, your design improves immeasurably. So I tend to work test-first regardless.
Arguably the most useful, but least often done, is automated acceptance testing. This is what ensures that the developers have understood what the customer was asking for. Sometimes this is left to QA, and I think that's fine, but the ideal situation is one in which these are an integral part of the development process.
The way this works is: for each requirement the test plan is turned into a script which is added to the test suite. Then you watch it fail. Then you write code to make it pass. Thus, if a coder is working on changes and is ready to check in, they have to do a clean build and run all the acceptance tests. If any fail, fix before you can check in.
"Continuous integration" is simply the process of automating this step - when anyone checks code in, a separate server checks out the code and runs all the tests. If any are broken it spams the last developer to check in until they are fixed.
I once consulted with a team that had a single tester. This guy was working through the test plans manually, all day long. When a change took place, however minor, he would have to start over. I built them a spreadsheet indicating that there were over 16 million possible paths through just a single screen, and they ponied up the $10k for Mercury Test Director in a hurry! Now he makes spreadsheets and automates the test plans that use them, so they have pretty thorough regression testing without ever-increasing QA time demands.
Once you've begun automating tests at every layer of your app (especially if you work test-first) a remarkable thing happens. Worry disappears!
So, no, it's not necessary. But if you find yourself worrying about technical debt, about the big deployment this weekend, or about whether you're going to break things while trying to quickly change to meet the suddenly-urgent customer requirements, you may want to more deeply investigate test-first development.

Filters are useful to help move logic such as is the user authenticated, to properly handle this, since you don't want this logic in every page.
Since you don't have a central controller it sounds like your filters are serving this function, which is fine, but, as you mentioned, it does make debugging harder.
This is where unit tests can come in handy, as you can test different situations, with each filter individually, then with all the filters in a chain, outside of your container, to ensure it works properly.
Unit testing does require discipline, but, if the rule is that nothing goes to QA without a unit test then it may help, and there are many tools to help generate tests so you just have to write the test. Before you debug, write or update the unit test, and show that the unit test is failing, so the problem is duplicated.
This will ensure that that error won't return, and that you fixed it, and you have updated a unit test.
For resource bundles. If you are certain you will never support another language, then as you refactor you can remove the need for the bundles, but, I think it is easier to make spelling/grammar corrections if the text is actually in one place.

Filters in general are expected to perform smaller units of functionality and filter-chaining would be used to apply the filters as needed. In your case, maybe a refactoring can help to move out some of the logic to additional filters and the redirecting logic can be somewhat centralized through a controller to be easier to debug and understand.
Resource bundles are necessary to maintain flexibility, but if you know absolutely that the site is going to be used in a single locale, then you might skip it. Maybe you can move some of the work in maintaining the bundles to the designers i.e let them have access to the resource bundles, so that you get the HTML with the keys in place.
Unit testing is much easier to implement at the beginning of a project as opposed to building it into a existing product. For existing software, you may still implement unit tests for the new features. However, it requires a certain amount of insistence from team leads and the team needs to buy into the necessity of having unit tests. Code review for unit tests helps and a decision on what parts of the code need to be absolutely covered can help developers. Tools/plugins like Coverlipse can indicate the unit testing coverage, but they tend to look at every possible code path, some of which may be trivial.
At one of my earlier projects, unit tests were just compulsory and unit tests would be automatically kicked off after each check-in. However, this was not Test-driven development, as the tests were mostly written after the small chunks of code were written. TDD can result in developers writing code to just work with the unit tests and as a result, developers can lose the big picture of the component they are developing.

In a web environment, how do you work with filters. When should you use J2EE filters and when shouldn't you?
Filters are meant to steer/modify/intercept the actual requests/responses/sessions. For example: setting the request encoding, determining the logged-in user, wrapping/replacing the request or response, determining which servlet it should forward the request to, and so on.
To control the actual user input (parameters) and output (the results and the destination) and to execute actual business logic, you should use a servlet.
Property resource bundle files for replacement values in JSP files.
If you don't do i18n, just don't use them. But if you ever grow and the customer/users want i18n, then you'll be happy that you're already prepared. And not only that, it also simplifies the use of a CMS to edit the content by just using the java.util.Properties API.
Unit Testing for all J2EE development
JUnit can take care about it. You can also consider to "officially" do user tests only. Create several use cases and test it.

End to End testing of Webservices

first time poster and TDD adopter. :-) I'll be a bit verbose so please bear with me.
I've recently started developing SOAP based web services using the Apache CXF framework, Spring and Commons Chain for implementing business flow. The problem I'm facing here is with testing the web services -- testing as in Unit testing and functional testing.
My first attempt at Unit testing was a complete failure. To keep the unit tests flexible, I used a Spring XML file to keep my test data in. Also, instead of creating instances of "components" to be tested, I retrieved them from my Spring Application context. The XML files which harbored data quickly got out of hand; creating object graphs in XML turned out to be a nightmare. Since the "components" to be tested were picked from the Spring Application Context, each test run loaded all the components involved in my application, the DAO objects used etc. Also, as opposed to the concept of unit test cases being centralized or concentrated on testing only the component, my unit tests started hitting databases, communicating with mail servers etc. Bad, really bad.
I knew what I had done wrong and started to think of ways to rectify it. Following an advice from one of the posts on this board, I looked up Mockito, the Java mocking framework so that I could do away with using real DAO classes and mail servers and just mock the functionality.
With unit tests a bit under control, this brings me to my second problem; the dependence on data. The web services which I have been developing have very little logic but heavy reliance on data. As an example, consider one of my components:
public class PaymentScheduleRetrievalComponent implements Command {
public boolean execute(Context ctx) {
Policy policy = (Policy)ctx.get("POLICY");
List<PaymentSchedule> list = billingDAO.getPaymentStatementForPolicy(policy);
ctx.put("PAYMENT_SCHEDULE_LIST", list);
return false;
}
}
A majority of my components follow the same route -- pick a domain object from the context, hit the DAO [we are using iBatis as the SQL mapper here] and retrieve the result.
So, now the questions:
- How are DAO classes tested esp when a single insertion or updation might leave the database in a "unstable" state [in cases where let's say 3 insertions into different tables actually form a single transaction]?
- What is the de-facto standard for functional testing web services which move around a lot of data i.e. mindless insertions/retrievals from the data store?
Your personal experiences/comments would be greatly appreciated. Please let me know in case I've missed out some details on my part in explaining the problem at hand.
-sasuke

I would stay well away from the "Context as global hashmap" 'pattern' if I were you.
Looks like you are testing your persistence mapping...
You might want to take a look at: testing persistent objects without spring

I would recommend an in-memory database for running your unit tests against, such as HSQL. You can use this to create your schema on the fly (for example if you are using Hibernate, you can use your XML mappings files), then insert/update/delete as required before destroying the database at the end of your unit test. At no time will your test interfere with your actual database.
For you second problem (end-to-end testing of web services), I have successfully unit tested CXF-based services in the past. The trick is to publish your web service using a light-weight web server at the beginning of your test (Jetty is ideal), then use CXF to point a client to your web service endpoint, run your calls, then finally shut down the Jetty instance hosting your web service once your unit test has completed.
To achive this, you can use the JaxWsServerFactoryBean (server-side) and JaxWsProxyFactoryBean (client-side) classes provided with CXF, see this page for sample code:
http://cwiki.apache.org/CXF20DOC/a-simple-jax-ws-service.html#AsimpleJAX-WSservice-Publishingyourservice
I would also give a big thumbs up to SOAP UI for doing functional testing of your web service. JMeter is also extremely useful for stress testing web services, which is particularity important for those services doing database lookups.

First of all: Is there a reason you have to retrieve the subject under test (SUT) from the Spring Application context? For efficient unit testing you should be able to create the SUT without the context. It sounds like you have some hidden dependencies somewhere. That might be the root of some of your headache.
How are DAO classes tested esp when a
single insertion or updation might
leave the database in a "unstable"
state [in cases where let's say 3
insertions into different tables
actually form a single transaction]?
It seems you are worried about the database's constistency after you have running the tests. If possible use a own database for testing, where you don't need care about it. If you have such a sandbox database you can delete data as you wish. In this case I would do the following:
Flag all your fake data with some common identifier, like putting a special prefix to a field.
Before running the test drop a delete statement, which deletes the flagged data. If there is none, then nothing bad happens.
Run your single DAO test. After that repeat step 2. for the next test.
What is the de-facto standard for
functional testing web services which
move around a lot of data i.e.
mindless insertions/retrievals from
the data store?
I am not aware of any. From the question your are asking I can infer that you have on one side the web service and on the other side the database. Split up the responsibilities. Have separate test suites for each side. One side just testing database access (as described above). On the other side just testing web service requests and responses. In this case it pays of the stub/fake/mock the layer talking to the network. Or consider https://wsunit.dev.java.net/.
If the program is only shoving data in and out I think that there is not much behavior. If this is the case, then the hardest work is to unit test the database side and the web service side. The point is you can do unit testing without the need for "realistic" data. For functional testing you will need handrolled data, which is close to reality. This might be cumbersome, but if you already unit tested the database and web service parts intensively, this should reduce the need for "realistic" test cases considerably.

First of all, make thing clear.
In an ideal world the lifecycle of the software your are building is something like this:
- sy makes a report with the customer, so you got an user story with examples about how the application should work
- you generalize the user story, so you got rules, which you call as use cases
- you start to write a piece of functional (end to end) test, and it fails...
- after that your build the ui and mock out the services, so you got a green functional test and a specification about how your services should work...
- your job is to keep the functional test green, and implement the services step by step writing integration tests, and mocking out dependencies with the same approach until you reach the level of unit tests
- after that you do the next iteration with the use cases, write the next piece of functional test, and so on until the end of the project
- after that you make acceptance tests with the customer who accepts the product and pays a lot
So what did we learn from this:
There are many types of tests (don't confuse them with each other)
functional tests - for testing the use cases (mock out nothing)
integration tests - for testing application, component, module, class interactions (mock out the irrelevant components)
unit tests - for testing a single class in isolation from its environment (mock out everything)
user acceptance tests - customer makes sure, that she accepts the product (manual functional tests, or presentation made from automatic functional tests in working)
You don't need to test everything by functional tests and integration tests, because it is impossible. Test only the relevant part by functional and integration tests and test everything by unit tests! Familiarize yourself with the testing pyramid.
Use TDD, it makes life easier!
How are DAO classes tested esp when a single insertion or updation might leave the database in a "unstable" state [in cases where let's
say 3 insertions into different tables actually form a single
transaction]?
You don't have to test your database transactions. Assume, that they are working well, because database developers have already tested them, and I am sure you don't want to write concurrency tests... Db is an external component, so you don't have to test it yourself. You can write a data access layer to adapt the data storage to your system, and write integration tests only for those adapters. In case of database migration these tests will work by the adapters of the new database as well, because your write them to implement a specific interface... By any other tests (except functional tests) you can mock out your data access layer. Do the same with every other external component as well, write adapters and mock them out. Put these kind of integration tests to a different test suite than the other tests, because they are slow because of database access, filesystem access, etc...
What is the de-facto standard for functional testing web services which move around a lot of data i.e. mindless insertions/retrievals
from the data store?
Your can mock out your data store with an in memory db which implements the same storage adapters until you implemented everything else except the database. After that your implement the data access layer for the database and test it with your functional tests as well. It will be slow, but it has to run only once, for example by every new release... If you need functional tests by developing, you can mock it out with an in memory solution again... An alternative approach to run only the affected functional tests by developing, or modify the settings of the test db to make things faster, and so on... I am sure there are many test optimization solutions...

I must say I don't really understand
your exact problem. Is the problem
that your database is left in an
altered state after you've run the
test?
Yes, there are actually two issues here. First one being the problem with the database left in an inconsistent state after running the test cases. The second one being that I'm looking for an elegant solution in terms of end-to-end testing of web services.
For efficient unit testing you should
be able to create the SUT without the
context. It sounds like you have some
hidden dependencies somewhere. That
might be the root of some of your
headache.
That indeed was the root cause of my headaches which I am now about to do away with with the help of a mocking framework.
It seems you are worried about the
database's constistency after you have
running the tests. If possible use a
own database for testing, where you
don't need care about it. If you have
such a sandbox database you can delete
data as you wish.
This is indeed one of the solutions to the problem I mentioned in my previous post but this might not work in all the cases esp when integrating with a legacy system in which the database/data isn't in your control and in cases when some DAO methods require a certain data to be already present in a given set of tables. Should I look into database unit testing frameworks like DBUnit?
In this case it pays of the
stub/fake/mock the layer talking to
the network. Or consider
https://wsunit.dev.java.net/.
Ah, looks interesting. I've also heard of tools like SOAPUI and the likes which can be used for functional testing. Has anyone here had any success with such tools?
Thanks for all the answers and apologies for the ambiguous explanation; English isn't my first language.
-sasuke

How can I improve my junit tests

Right my junit tests look like a long story:
I create 4 users
I delete 1 user
I try to login with the deleted user and make sure it fails
I login with one of the 3 remaining user and verify I can login
I send a message from one user to the other and verify that it appears in the outbox of the sender and in the inbox of the receiver.
I delete the message
...
...
Advantages:
The tests are quite effective (are very good at detecting bugs) and are very stable, becuase they only use the API, if I refactor the code then the tests are refactored too. As I don't use "dirty tricks" such as saving and reloading the db in a given state, my tests are oblivious to schema changes and implementation changes.
Disadvantages:
The tests are getting difficult to maintain, any change in a test affects other tests. The tests run 8-9 min which is great for continuous integration but is a bit frustrating for developers. Tests cannot be run isolated, the best you can do is to stop after the test you are interested in has run - but you absolutely must run all the tests that come before.
How would you go about improving my tests?

First, understand the tests you have are integration tests (probably access external systems and hit a wide range of classes). Unit tests should be a lot more specific, which is a challenge on an already built system. The main issue achieving that is usually the way the code is structured:
i.e. class tightly coupled to external systems (or to other classes that are). To be able to do so you need to build the classes in such a way that you can actually avoid hitting external systems during the unit tests.
Update 1: Read the following, and consider that the resulting design will allow you to actually test the encryption logic without hitting files/databases - http://www.lostechies.com/blogs/gabrielschenker/archive/2009/01/30/the-dependency-inversion-principle.aspx (not in java, but ilustrates the issue very well) ... also note that you can do a really focused integration tests for the readers/writers, instead of having to test it all together.
I suggest:
Gradually include real unit tests on your system. You can do this when doing changes and developing new features, refactoring appropriately.
When doing the previous, include focused integration tests where appropriate. Make sure you are able to run the unit tests separated from the integration tests.
Consider your tests are close to testing the system as a whole, thus are different from automated acceptance tests only in that they operate on the border of the API. Given this think about factors related to the importance of the API for the product (like if it will be used externally), and whether you have good coverage with automated acceptance tests. This can help you understand what is the value of having these on your system, and also why they naturally take so long. Take a decision on whether you will be testing the system as a whole on the interface level, or both the interface+api level.
Update 2: Based on other answers, I want to clear something regarding doing TDD. Lets say you have to check whether some given logic sends an email, logs the info on a file, saves data on the database, and calls a web service (not all at once I know, but you start adding tests for each of those). On each test you don't want to hit the external systems, what you really want to test is if the logic will make the calls to those systems that you are expecting it to do. So when you write a test that checks that an email is sent when you create an user, what you test is if the logic calls the dependency that does that. Notice that you can write these tests and the related logic, without actually having to implement the code that sends the email (and then having to access the external system to know what was sent ...). This will help you focus on the task at hand and help you get a decoupled system. It will also make it simple to test what is being sent to those systems.

unit tests should - ideally - be independent, and able to run in any order. So, I would suggest that you:
break up your tests to be independent
consider using an in-memory database as the backend for your tests
consider wrapping each test or suite in a transaction that is rolled back at the end
profile the unit tests to see where the time is going, and concentrate on that
if it takes 8 minutes to create a few users and send a few messages, the performance problem may not be in the tests, rather this may be a symptom of performance problems with the system itself - only your profiler knows for sure!
[caveat: i do NOT consider these kinds of tests to be 'integration tests', though i may be in the minority; i consider these kinds of tests to be unit tests of features, a la TDD]

Now you are testing many things in one method (a violation of One Assertion Per Test). This is a bad thing, because when any of those things changes, the whole test fails. This leads it to not being immediately obvious why a test failed and what needs to be fixed. Also when you intentionally change the behaviour of the system, you need to change more tests to correspond the changed behaviour (i.e. the tests are fragile).
To know what kind of tests are good, it helps to read more on BDD: http://dannorth.net/introducing-bdd http://techblog.daveastels.com/2005/07/05/a-new-look-at-test-driven-development/ http://jonkruger.com/blog/2008/07/25/why-behavior-driven-development-is-good/
To improve the test that you mentioned, I would split it into the following three test classes with these context and test method names:
Creating user accounts
Before a user is created
the user does not exist
When a user is created
the user exists
When a user is deleted
the user does not exist anymore
Logging in
When a user exists
the user can login with the right password
the user can not login with a wrong password
When a user does not exist
the user can not login
Sending messages
When a user sends a message
the message appears in the sender's outbox
the message appears in the reciever's inbox
the message does not appear in any other message boxes
When a message is deleted
the message does not anymore exist
You also need to improve the speed of the tests. You should have a unit test suite with good coverage, which can run in a couple of seconds. If it takes longer than 10-20 seconds to run the tests, then you will hesitate to run them after every change, and you lose some of quick feedback that running the tests gives you. (If it talks to the database, it's not a unit test, but a system or integration test, which have their uses, but are not fast enough to be executed continually.) You need to break the dependencies of the classes under test by mocking or stubbing them. Also from your description it appears that your tests are not isolated, but instead the tests depend on the side-effects caused by previous tests - this is a no-no. Good tests are FIRST.

Reduce dependencies between tests. This can be done by using Mocks. Martin Fowler speaks about it in Mocks aren't stubs, especially why mocking reduces dependencies between tests.

You can use JExample, an extension of JUnit that allows test methods to have return values that are reused by other tests. JExample tests run with the normal JUnit plugin in Eclipse, and also work side by side with normal JUnit tests. Thus migration should be no problem. JExample is used as follows
#RunWith(JExample.class)
public class MyTest {
#Test
public Object a() {
return new Object();
}
#Test
#Given("#a")
public Object b(Object object) {
// do something with object
return object;
}
#Test
#Given("#b")
public void c(Object object) {
// do some more things with object
}
}
Disclaimer, I am among the JExample developers.

If you use TestNG you can annotate tests in a variety of ways. For example, you can annotate your tests above as long-running. Then you can configure your automated-build/continuous integration server to run these, but the standard "interactive" developer build would not (unless they explicitly choose to).
This approach depends on developers checking into your continuous build on a regular basis, so that the tests do get run!
Some tests will inevitably take a long time to run. The comments in this thread re. performance are all valid. However if your tests do take a long time, the pragmatic solution is to run them but not let their time-consuming nature impact the developers to the point that they avoid running them.
Note: you can do something similar with JUnit by (say) naming tests in different fashions and getting your continuous build to run a particular subset of test classes.

By testing stories like you describe, you have very brittle tests. If only one tiny bit of functionality is changing, your whole test might be messed up. Then you will likely to change all tests, which are affected by that change.
In fact the tests you are describing are more like functional tests or component tests than unit tests. So you are using a unit testing framework (junit) for non-unit tests. In my point of view there is nothing wrong to use a unit testing framework to do non-unit tests, if (and only if) you are aware of it.
So there are following options:
Choose another testing framework which supports a "story telling"-style of testing much better, like other user already have suggested. You have to evaluate and find a suitable testing framework.
Make your tests more “unit test”-like. Therefore you will need to break up your tests and maybe change your current production code. Why? Because unit testing aims on testing small units of code (unit testing purists suggest only one class at once). By doing this your unit tests become more independent. If you change the behavior of one class, you just need to change a relatively small amount of unit test code. This makes your unit test more robust. During that process you might see that your current code does not support unit testing very well -- mostly because of dependencies between classes. This is the reason that you will also need to modify your production code.
If you are in a project and running out of time, both options might not help you any further. Then you will have to live with those tests, but you can try to ease your pain:
Remove code duplication in your tests: Like in production code eliminate code duplication and put the code into helper methods or helper classes. If something changes, you might only need to change the helper method or class. This way you will converge to the next suggestion.
Add another layer of indirection to your tests: Produce helper methods and helper classes which operate on a higher level of abstraction. They should act as API for your tests. These helpers are calling you production code. Your story tests should only call those helpers. If something changes, you need to change only one place in your API and don't need to touch all your tests.
Example signatures for your API:
createUserAndDelete(string[] usersForCreation, string[] userForDeletion);
logonWithUser(string user);
sendAndCheckMessageBoxes(string fromUser, string toUser);
For general unit testing I suggest to have a look into XUnit Test Patterns from Gerard Meszaros.
For breaking dependencies in your production tests have a look into Working Effectively with Legacy Code from Michael Feathers

In addition to the above, pick up a good book on TDD (I can recommend "TDD and Acceptance TDD for Java Developers"). Even though it will approach from a TDD point of view there is alot of helpful information about writing the right kind of unit tests.
Find someone who has alot of knowledge in the area and use them to figure out how you can improve your tests.
Join a mailing list to ask questions and just read the traffic coming through. The JUnit list at yahoo (something like groups.yahoo.com/junit). Some of the movers and shakers in the JUnit world are on that list and actively participate.
Get a list of the golden rules of unit tests and stick them on your (and others) cubicle wall, something like:
Thou shalt never access an external system
Thou shalt only test the code under test
Thou shalt only test one thing at once
etc.

Since everyone else is talking about structure I'll pick different points. This sounds like a good opportunity to profile the code to find bottleknecks and to run it through code coverage to see if you are missing anything (given the time it takes to run it the results could be interesting).
I personally use the Netbeans profiler, but there are ones in other IDEs and stand alone ones as well.
For code coverage I use Cobertura, but EMMA works too (EMMA had an annoyance that Cobertura didn't have... I forget what it was and it may not be an issue anymore). Those two are free, there are paid ones as well that are nice.