Category: 2011

Grails upgrade – lessons learned

Grails is a great framework for rapidly developing web applications on top the proven java/servlet platform. Especially smaller, short-lived projects can be a real breeze with all the scaffolding, GORM and convention-over-configuration built into grails.

We happen to use it for a quite complex web application project for almost 3 years now. Half a year ago we upgrade from grails 1.0.x to 1.3.4. That makes 3 major versions in one upgrade step and produced obviously a lot of work and many small bugs. I do not want to put the blame on the grails guys here, because most of the stuff was mentioned in the release notes and it was a big step we decided to take when the decision came to continue the project for several years to come.

Our upgrade policy changed due to that experience and we try to stay a lot more current to be able to adapt our software to framework changes more incrementally. Some weeks ago we upgraded from 1.3.4 to 1.3.7 and this experience was not pleasant. Even though we skimmed through the changelogs and release notes and thought the update should be uncritical for us grails behaviour changed in two aspects which broke things for us:

  1. An API-change where GroovyPagesException was changed to GrailsTagException
  2. Behavioural change where no application context and injections are available in functional tests anymore

Item 1 was easy to fix but you need really good testing to spot it before it slips into production. Such subtle API changes should not happen in micro-version updates as that can easily break parts of the system whithout you knowing because of groovy/grails’ dynamic nature. No compiler saves you here.

Item 2 produced some amount of work for us because we build a quite extensive acceptance test suite using services and domain objects to setup the initial environment for each test. Luckily, there is the grails remote control plugin which you can use for things like that.

Lessons learned

  • You should have extensive automated test suites when developing a grails application over a longer period because things can break in unexpected ways without code changes on your side.
  • Try to plan upgrades some time ahead of releases and dedicate time to scanning the release notes and actually performing the upgrade. It may take you significantly longer than the smooth upgrade procedure itself suggests.

The grails team seems to be increasingly aware of backwards compatibility but they still have some way to go. We hope and expect to see fewer unexpected breakages to occur in the future.

The Grails performance switch: flush.mode=commit

Some default configuration options of Grails are not optimal for all projects.

— Disclaimer —
This optimization requires more manual work and is error prone but isn’t this with most (big) performance improvements?
For it to really work you have to structure your code accordingly and flush explicitly.
—
Recently in our performance measurements of a medium sized Grails project we noticed a strange behavior: every time we executed the same query the time it took increased. It started with 40ms and every time it took 1 ms more. The query was simple like Child.findAllByParent(parent)
The first thought: indexes! We looked at the database (a postgresql db) and we had indexes on the parent column.
Next: maybe the session cache got too large. But session.flush() and session.clear() did not solve that problem.
Another post suggested using a HQL query. Changing to

Child.executeQuery("select new Child(c.name, c.parent) from Child c where parent=:parent", [parent: parent])

had no effect.
Finally after countless more attempts we tried:

session.setFlushMode(FlushMode.COMMIT)

And not even the query executed in constant time it was also 10x faster?!
Hmmm…why?
The default flush mode in Grails is set to AUTO
Which means that before every query made the session is flushed. Every query regardless of the classes effected. The problem is known for hibernate but after 4! years it is still unresolved.
So my question here is: why did Grails chose AUTO as default?

Test Framework Classpath Forgery

A lesson learnt when using HttpUnit with all its dependencies. Xerces changed the system behaviour, but with the test classpath only.

Recently, I had an interesting problem using a testing framework with third-party dependencies. When writing integration tests with JUnit against a very small embedded web application (think of the web based management console for your printer as an example), I chose to use HttpUnit as an auxiliary framework to reduce and clarify the test code.

HttpUnit for testing web applications

If you need to test a classic request/response web application, HttpUnit serves its purpose very well. You can write test code concise and to the point. Downloading and integrating HttpUnit is straight-forward, you can immediately get it to work. Here is an example of a test that asserts that there is at least one link on the web application’s main page:

WebConversation web = new WebConversation();
WebResponse response = web.getResponse(fromServer(port));
WebLink[] allLinks = response.getLinks();
assertTrue("No links found on main webpage", ArrayUtil.hasContent(allLinks));

Test failures appear

After this test was written and included into the build, the continuous integration suddenly reported test failures – in the unit tests. I didn’t change any test there and had no need to change the production code, either. So what was causing the test to fail?

The failing unit test class was very old, ensuring the persistence of some data structure to XML and back. The test that actually failed took care of the XML parser behaviour when an empty XML file was read:

public void testReadingEmptyXML() throws IOException {
    try {
        new XMLQueryPersister(new StringReader(XMLQueryPersisterTest.EMPTY_XML), null).loadQueries();
        Assert.fail();
    } catch (ParseException e) {
        Assert.assertEquals("Error on line 1: Premature end of file.", e.getMessage());
    }
}

The assertion that checks the exception message failed, stating that the actual message was now “Error on line -1: Premature end of file.”

Hunting the bug

How can the inclusion of a new integration test have such an impact on the rest of the system? Thanks to continuous integration, the cause for the behaviour change could only lie in the most recent commit. A quick investigation revealed the culprit:

HttpUnit has a third-party dependency on the Xerces xml parser (or another equivalent org.xml.sax parser), see their FAQ for details. When I included the libraries, I accidentally changed the default xml parser for the whole system to Xerces in the version that HttpUnit delivers. This altered the handling of the “premature end of file” case to the new behaviour, causing the test to fail. As these libraries are only included in the classpath when tests are run, the change only happens in the test environment, not in production.

Test classpath versus production classpath

The real issue here isn’t the change in behaviour, this can be taken into account if you have a good test coverage. The issue is different classpaths for test and production environments. If you don’t want to deploy all your test scope libraries (thus making the production classpath similar to the test classpath), you should pay extra attention to what you include in your test classpath. It might alter your system, so that you don’t test the real behaviour anymore.

Resolving the issue

In my case, it was sufficient to remove the Xerces jar from the classpath again. A compliant org.xml.sax parser is already included in the Java core API. It’s the parser that already got used in production and should be used for the tests, too.

Update/Correction: After removing Xerces, HttpUnit stopped working correctly. The quick fix now is to include Xerces in the production classpath and deal with the behaviour changes. I will investigate this issue further and append the outcome as a comment to this blog entry. Update 2: Issue resolved, see comment section for the solution.

This taught me a lesson to always be aware of the dependencies, even if it’s “only” the test scope dependencies.

Summary

Including the Xerces xml parser as a dependency for a testing framework (HttpUnit) changed the behaviour of my system under test, albeit for the tests only. The issue was easily resolved by removing it again, but now I know that testing frameworks have side effects, too.

Old Code

Why bother buying Stephen King’s horror books, just take a look at your old code.

There is a saying that if you don’t be embarrassed by code that you wrote six month ago, you haven’t learned anything. Recently, I stumbled upon a C/C++ project that dates back to the very early days of my programming career – this was many * six months ago – and I can tell you, I was very embarrassed.

I had just “learned” C++ and object-orientation at that time and, of course, wanted to program that way. The result was terrible. The only small piece of object-orientation was the use of the keyword class. There were public fields all over the place,  no interfaces or abstractions of any kind, switches over type-ids, and so on.

Another highlight was the vast amount of literals scattered all over the code. For example, as it was a curses-based application, I had to read and display user input using curses methods like

int mvwgetch(WINDOW *win, int y, int x);

and

 int mvwaddch(WINDOW *win, int y, int x, const chtype ch);

And what did I do? I hard-coded y and x positions on every call of those methods. So it would often be the case that I changed, say, the y position in one part and … well, you guessed it already.

Naming of variables was also big. Boolean values would often be called “flag”, a name length of more than 4 was considered way too long.

But there was also progress. In later parts of the software I started to use “advanced” things like auto_ptrs, std::list, and std::map. Hooray!

The only positive thing about this project was that since I made every possible mistake one can imagine, I learned quite lot about programming. And I remember that at the end of the project, I was already very embarrassed about the whole thing…

So if you like reading horror stories, try digging up your old code 😉 And share if you like.

A tale of scrap metal code – Part III

The third and last part of a series about the analysis of a software product. This part tries to give some rules of thumb on how to avoid failure like in this project.

In the first part of this tale about an examined software project, I described the initial situation and high-level observations about the project. The second part dove into the actual source code and pointed out what’s wrong on this level. This part will summarize everything and give some hints on how to avoid creating scrap metal code.

About the project

If you want to know more about the project, read the first part of this tale. In short, the project looked like a normal Java software, but unfolded into a nightmare, lacking basic requirements like tests, dependency management or continuity.

A summary of what went wrong

In short, the project failed in every respect except being reasonable functional and delivering business value to the customers. I will repeat this sentence soon, but let’s recall the worst parts again. The project had no tests. The project modularization was made redundant by circular dependencies and hardwired paths. No dependency management was in place, neither through the means of a build tool nor by manual means (like jar versions). The code was bloated and overly complex. The application’s data model was a widely distributed network of arbitrary collections with implicit connections via lookup keys. No effort was spent to grasp exception handling or multithreading. The cleverness was rather invested into wildcard usage of java’s reflection API capabilities. And when the cleverness of the developer was challenged, he resorted to code comments instead of making the code more accessible.

How can this be avoided?

First, you need to know exactly what it is you want to avoid. Let me repeat that the project was sold to happily paying customers who gained profit using it. Many software projects fail to deliver this utmost vital aspect of virtually every project. The problem with this project isn’t apparent yet, because it has a presence (and a past). It’s just that it has no future. I want to give some hints how to develop software projects with a future while still delivering business value to the customer.

Avoid the no-future trap

http://www.istockphoto.com/stock-photo-5407438-percent-blocks.phpThe most important thing to make a project future-proof is to restrain yourself from taking shortcuts that pay off now and need to be paid back later. You might want to believe that you don’t need to pay back your technical debts (the official term for these shortcuts) or that they will magically disappear sometimes, but both scenarios are quite unlikely. If your project has any chance to keep being alive over a prolonged amount of time, the technical debts will charge interest.

Of course you can take shortcuts to meet tight deadlines or fit into a small budget. This is called prototyping and it pays off in terms of availability (“time to market”) and scope (“trial version”). Just remember that a prototype isn’t meant for production. You definitely need the extra time and/or budget to fix the intentional shortcomings in the code. You won’t feel the difference right now (hey, it works, what else should it do?), but it will return with compound interest in a few years. The project in this tale was dead after three years. The technical debt had added up beyond being repairable.

Analyzing technical debts

It’s always easy to say that you should “do it right” in the first place. What could the developer for project at hands have done differently to be better off now?

1. Invest in automated tests

When I asked why the project has no tests at all, the developer replied that “it surely would be better to have tests, yet there was no time to write them“. This statement implies that tests take more time to write than they save acting as a guideline and a safety net. And it is probably true for every developer just starting to write tests. You will feel uncomfortable, your tests will be cumbersome and everything will slow down. Until you gain knowledge and experience in writing tests. It is an investment. It will pay off in the future, not right now. If you don’t start now, there will be no future payout. And even better: now your investment, not your debt, will accumulate interest. You might get used to writing tests and start being guided by them. They will mercilessly tell you when your anticipated solution is overly complex. And they will stay around and guard your code long after you forgot about it. Tests are a precaution, not an afterthought.

2. Review and refactor your code

The project has a line count of 80,000 lines of ugly code. I’m fairly confident that it can be reduced to 20,000 lines of code without losing any functionality. The code is written with the lowest possible granularity, with higher concepts lurking everywhere, waiting to be found and exposed. Of course, you cannot write correct, concise and considerate code on your first attempt. This is why you should revisit old code in a recurring manner. If you followed advice number one and brought your tests in place, you can apply every refactoring of the book’s catalog and still be sure that you rather fixed this part instead of breaking it. Constantly reviewing and refactoring your code has the additional advantage of a code base that gets more proficient alongside yourself. There are no “dark regions” (the code to never be read or touched again, because it hurts) if you light them up every now and then. This will additionally slow you down when you start out, but put you on afterburner when you realize that you can rescue any code from rotting by applying the refactoring super-powers that you gained through pratice. It’s an investment again, aiming at midterm return of investment.

3. Refrain from clever solutions

The project of this tale had several aspects that the developer thought were “clever”. The only thing with “clever” is that it’s a swearword in software programming. Remember the clever introduction of wildcard runtime classloading to provide a “plugin mechanism”? Pure poison if you ever wanted your API to be stable and documented, just like a plugin interface should be. Magic numbers throughout your code? Of course you are smart enough to handle this little extra obfuscation. Except when you aren’t. You aren’t sure how exception handling works? Be clever and just “empty catch Exception” everywhere the compiler points you to. In this project, the developer knew this couldn’t be the right solution. Yet, he never reviewed the code when he one day knew how to handle exceptions in a meaningful manner. Let me rest my case by stating that if you write your code as clever as you can handle it, you won’t be able to read it soon, as reading code is harder than writing it.

Summary

Over the course of this tale, you learned a lot about a failed project. In this article, I tried to give you some advice (in the form of three basic rules) on how this failure could probably have been avoided. Of course, the advice isn’t complete. There is much more you could do to improve yourself and your project. Perhaps the best self-training program for developer skills is the Clean Code Developer Initiative (it’s mostly german text yet, so here is an english blog post about it), based upon the book “Clean Code” by Robert C. Martin (Uncle Bob).

Invest in the future of your project and stay clean.

SSL with POCO

A short introduction to using SSL support in POCO C++ libraries.

Admittedly, the topic of this post is very specific but I hope it will still be of some value for some people.The task for today is to setup SSL server and client with POCO framework classes. I will leave out the whole certificate managing issues and just assume that the right files are at hand.

The SSL related part of  the POCO libraries essentially wraps the OpenSSL library into a nice object-oriented interface. When you know OpenSSL, you can instantly relate to classes like Poco::Net::Context, or the …Handler classes (if you replace “handler” with “callback”).

“SSL” stands for Secure Socket Layer, so the first thing to discover is class Poco::Net::SecureServerSocket. As you would expect, this class is derived from Poco::Net::ServerSocket, extending it only with SSL related stuff. And sure enough, some constructors of Poco::Net::ServerSocket take a Poco::Net::ContextPtr as argument.

But why only some constructors? Since there is no setContext method, there must be some other mechanism in place by which SecureServerSockets get their SSL context.

Introducing Poco::Net::SSLManager. From the API docs:

SSLManager is a singleton for holding the default server/client Context and handling callbacks for certificate verification errors and private key passphrases.

Proper initialization of SSLManager is critical.

Aha! So all the constructors of SecureServerSocket that do not take Context pointers simply get it from the SSLManager singleton.

But how to initialize SSLManager?

1. The POCO Way:

If you developed your application with POCO from the ground up there probably exists a sub-class of Poco::Application, and all the configuration is handled by the built-in configuration classes.

With this in place, all you have to do is to add the proper ssl configuration elements:

openSSL.server.privateKeyFile = /path/to/key/file
openSSL.server.certificateFile = /path/to/certificate/file
openSSL.server.verificationMode = none
openSSL.server.verificationDepth = 9
openSSL.server.loadDefaultCAFile = false
openSSL.server.cypherList = ALL:!ADH:!LOW:!EXP:!MD5:@STRENGTH
openSSL.server.privateKeyPassphraseHandler.name = KeyFileHandler
openSSL.server.privateKeyPassphraseHandler.options.password = securePassword
openSSL.server.invalidCertificateHandler = AcceptCertificateHandler

2. Manually:

Depending on which side you are – client or server – you have to call SSLManager::initializeClient or  SSLManager::initializeServer. Both methods take three arguments:

  1. PrivateKeyPassphraseHandler pointer
  2. InvalidCertificateHandler pointer
  3. Context pointer

This is where it becomes a little bit tricky: If you try to instantiate a Context with a privateKey file in order to provide it as argument to the initialize… method, a PrivateKeyPassphraseHandler might be needed. This handler is fetched from the SSLManager singleton – which you are just about to initialize!.

This circular dependency between Context and SSLManager can be overcome e.g. if you call SSLManager::initializeServer first only with a PrivateKeyPassphraseHandler, a InvalidCertificateHandler and null Context pointer. Then instantiate the Context and call SSLManager::initializeServer again.

Now that SSL Manager is initialized we can use Secure… prefixed classes as we would used their non-SSL counterparts. As with SecureServerSocket, other Secure… classes are derieved from corresponding non-secure base classes.

Conclusion: Once you got around the initialization of SSLManager singelton, using SSL POCO classes is very easy and straight forward. Check it out!

Information Hiding in Source Code Comments

Use comments for additional details, not as an issue tracker.

Once in a while you come across a part of code needs to be fixed. You put in a comment saying something like ‘FIXME’ or ‘BROKEN’ or just another ‘TODO’. Sure you have no time, you need to ship. You don’t even know how much work it is to fix it or how many tests break after your modifications. A comment is safe and clear.
After a while the code around it changes. The software is shipped and used. The comment is long forgotten.
One day a customer calls and reports about an incident which sounds like a bug. After minutes or hours debugging and finally pinning the bug you rediscover a long lost information:

// FIXME: shouldn't we do this here?

So next time you find a code part which looks strange or seems to have a bug, create an issue and describe why you think it does not do what it should. And even better: write a test and fix it.

A tale of scrap metal code – Part II

The second part of a series about the analysis of a software product. This part investigates the source code and reveals some ugly practices therein.

In the first part of this tale about an examined software project, I described the initial situation and high-level observations about the project. This part will dive into the actual source code and hopefully reveal some insights. The third and last part will summarize everything and give some hints on how to avoid creating scrap metal code.

About the project

If you want to know more about the project, read the first part of this tale. In short, the project looked like a normal Java software, but unfolded into a nightmare, lacking basic requirements like tests, dependency management or continuity.

About the developer

The developer has a job title as a “senior developer”. He developed the whole project alone and wrote every line of code. From the code, you can tell his initial uncertainty, his learning progress, some adventurous experiments and throughout every file, a general uneasiness with the whole situation. The developer actively abandoned the project after three years of steady development. From what I’ve seen, I wouldn’t call him a “senior” developer at all.

About the code

The code didn’t look very repellent at first sight. But everywhere you looked, there was something to add on the “TODO list”. Let me show you our most prominent findings:

Unassigned constructors

The whole code was littered with constructor calls that don’t store the returned new object. What’s the point in constructing another instance of you throw it away in the next moment, without ever using it? After examining these constructors, it became apparent that they only exist to perform side effects each. The new object is registered with the global data model while it’s still under construction. It was the most dreadful application of the Monostate design pattern I’ve ever seen.

Global data model

Did I just mention the global data model? At the end of the investigation, we found that the whole application state lives in numerous public static arrays, collections or maps. These data fields are accessible from everywhere in the application and altered without any protection against concurrent modifications. These global variables were placed anywhere, without necessarily being semantically associated with their enclosing class. A data model in the sense of some objects being tied together to form an instance net with higher-level structures could not be found. Instead, different lookup structures like index-based arrays and key-based maps are associated by shared keys or obscure indices. The whole arrangement of the different data pieces is implicit, you have to parse the code for every usage. Mind you, these fields are globally accessible.

Manual loop unrolling

Some methods had several hundred lines of the exact same method call over and over again. This is what your compiler does when it unrolls your foreach loops. In this code, the compiler didn’t need to optimize. To add some myth, the n-th call usually had a slight deviation from the pattern without any explanation. Whenever something could easily be repeated, the developer pasted it all over the place. Just by winding up the direct repititions again, the code migth shrink by one quarter in length.

Least possible granularity

Just by skimming over the code, you’d discover plenty of opportunities to extract methods, raising the level of abstraction in the code. The developer chose to stick with the least possible granularity, making each non-trivial code a pain to read. The GUI-related classes, using Swing, were so bloated by trivialities that even a simple dialog with two text fields and one button was represented by a massive amount of code. Sadly, the code was clearly written by hand because of all the mistakes and pattern deviations. If the code had to deal with complex data types like dates, the developer always converted them to primitive data types like int, double or long and performed the necessary logic using basic math operators.

“This code is single-threaded, right?”

Despite being a Java Swing application, the code lacked any strategy to deal with multithreading other than ignoring the fact that at least two threads would access the code. We didn’t follow this investigation path down to its probably bitter end, but we wouldn’t be surprised if the GUI would freeze up occasionally.

“Exceptions don’t happen here”

If you would run a poll on the most popular exception handling strategy for this code, it would be the classic “local catch’n’ignore”. The developer dismissed the fact that exceptions might happen and just carried on. If he was forced to catch an exception, the catch block followed immediately and was empty in most cases. Of course, the only caught exception type was the Exception class itself.

“This might be null“

One recurring pattern of the developer was a constructor call, stored in a local variable and immediate null check. Look at this code sample:

try {
    SomeObject object = new SomeObject();
    if (object != null) {
        object.callMethod();
    }
    [...]
} catch (Exception e) {
}

There is no possibility (that I know of) of object being null directly after the constructor call. If an exception is thrown in the constructor, the next lines won’t be executed. This code pattern was so prevalent in the code that it couldn’t be an accidental leftover of previous code. The accompanying effect were random null checks for used variables and return values.

Destabilized dependencies

If there is one thing that’s capable of derailing every code reader, analysis tool and justified guess, it’s wildcard use of Java’s reflection capabilities. The code for this project incorporates several dozens calls to Class.forName(), basically opening up the application for any code you want to dynamically include. The class names result from obscure string manipulation magic or straight from configuration. It’s like the evil brother of dependency injection.

Himalaya indentation

Looking at the indentation depths of the code, this wasn’t the worst I’ve ever seen. But that doesn’t mean it was pleasant. Like in Uncle Bob’s infamous “A crap odyssey”, you could navigate some classes by whitespace landscape. “Scroll down to the fifth crest, the vast valley afterwards contains the detail you want to know”.

Magic numbers

The code was impregnated with obscure numbers (like 9, 17, 23) and even more bizzare textual constants (like “V_TI_LB_GUE_AB”) that just appeared out of nowhere several times. This got so bad that the original author included lengthy comment sections on top of the biggest methods to list the most prominent numbers alongside their meanings. Converting the numbers to named constants would probably dispel the unicorns, as we all know that unicorns solely live on magic numbers(*). Any other explanation escapes my mind.

(*) On a side note, I call overly complex methods with magic numbers “unicorn traps”, as the unicorns will be attracted by the numbers and then inevitably tangled up in the complexity as they try to make their way out of the mess.

Summary

This was the list of the most dreaded findings in the source code. Given enough time, you can fix all of them. But it will be a long and painful process for the developer and an expensive investment for the stakeholder.

To give you an overall impression of the code quality, here is a picture of the project’s CrapMap. The red rectangles represent code areas (methods) that need improvements (the bigger and brighter, the more work it will take). The green areas are the “okay” areas of the project. Do you see the dark red cluster just right the middle? These are nearly a hundred complex methods with subtle differences waiting to be refactored.

Prospectus

In part three, I’ll try to extract some hints from this project on how to avoid similar code bases. Stay tuned.

SSD? Don’t think! Just Buy!

SSDs makes everything blazingly fast – even Grails + IDEA development

My personal experience with SSDs began with an Intel X25M that I built into a Lenovo Thinkpad R61. It replaced a Seagate 160 GB 5400rpm which in combination with Windows Vista … well, let’s just say, it wasn’t that fast.

The SSD changed everything. It was not just faster, it was downright awesome! As if I had a completely new computer.

With that in mind I thought about my desktop PC. It’s a little more than 2 year old Windows XP box, Intel Core2Duo 2.7 GHz, 4GB RAM, with a not so slow Samsung HDD. I use it mainly for programming, which is most of the time Grails programming under IntelliJ IDEA.

And let me tell you, the Grails + IDEA combination can get dog slow at times. The start-up time of IDEA alone gives you time to skim over the first three pages of Hacker News and read the latest XKCD.

So the plan was to put an extra SSD into the Windows box and put only programming related stuff on it. This would save me the potential hassle of moving my whole system but would still give me development speed-up.

I had to be a little careful because the standard settings for IDEA’s so-called “system path” and “config path” is in the user’s home directory. (Btw, this settings can be changed in file “idea.properties” which resides in “IDEA_INSTALLATION_DIR\bin”, e.g.: c:\Progam Files\JetBrains\IntelliJ IDEA 9.0.4\bin)

I think you already guessed the result. Three words: fast, faster, SSD. It’s just amazing! IDEA start-up is so fast now, I barely have time for a quick look at the newest headlines on InfoQ.

The next step is of course to put the whole system on SSD but that will probably have to wait until we upgrade the whole company to Win7. Can’t wait… 🙂

Acceptance testing a grails app with selenium-rc

Having extensive acceptance tests is the basis of delivering high quality releases with very few regressions for long time projects. This is even more true when your environment uses dynamically typed languages and changing requirements. One of our Grails projects is running for several years now and continues to evolve and grow.  We are in dire need of more acceptance tests and especially their automated execution. Manual testing is not feasible and our coverage through unit and integration tests is not enough. We have a nice set of Selenium IDE acceptance tests already though. They were executed very infrequently which let some bugs slip through into production.

I want to describe our approach to automated and extensive acceptance testing below:

  1. We create the acceptance tests using capture & replay with selenium IDE. This is a fast way to exercise a new feature through a repeatable test.
  2. We think that maintaining the tests in code offers much more flexiblity and is easier to run in continuous integration (CI) than maintaining the selenium IDE html code. So we export the captured test to play nicely with the grails selenium-rc plugin. Kurt Harriger explains the setup and usage of the selenium-rc grails plugin. You need to make some changes to the exported code for everything to work nicely:
    1. Delete or change package declaration.
    2. Choose a grails functional test compatible file name like MyAcceptanceTests.groovy. We use the Junit4 export but Groovy export works also because the difference is only marginal and both must be adjusted in some places.
    3. Change the class name to match the file name without extension if they are not equal already.
    4. Change the exported test to extends from GroovyTestCase instead of *SeleneseTestCase.
    5. Add the @Mixin(SeleniumAware) annotation to the test class.
    6. Remove the setup and teardown methods.
    7. Replace verifyTrue() and friends with junit assertions.
  3. Each test has to setup it’s initial state. This leads to independent acceptance tests at the expense of some longer running time but is well worth the cost imho.
  4. The resulting selenium-rc tests can be run easily using grails test-app -functional and thus integration in the build process is pretty straighforward. We currently use ant to wrap the grails calls, but other ways may be more feasible depending on your infrastructure.

The end result is fast creation of acceptance tests and much flexibility setting up the test fixture and maintaining the tests. Using the grails plugin you gain easy execution of the tests on the developer machines and CI servers as well. With extensive automated acceptance tests the danger of regressions is greatly reduced. But be sure to not neglect unit and integration tests!