Those of us who have been working in strongly-typed OO languages for a very long time – since, say, the GOF Design Patterns book was first published – learned not only OO concepts and the syntactic idiosyncrasies of our language, but also design patterns – commonly-accepted ways of solving recurring problems in our applications.

If you’ve been with Java over the years – say, from Java 3 to 4 to 5, you’ve seen the language evolve to include features formerly relegated to the domain of the design pattern. Java’s for-each loop is one such example – once that became a language feature, it was suddenly less important to know about the more generalized Iterator pattern.

Enter Ruby (and dynamic languages in general). It turns out that many of the design patterns in that grand old Gang of Four book were situational – they apply generally to strongly-typed languages. If there is one “metapattern” that describes most of the original design patterns, it is this: “Use composition rather than inheritance in order to introduce runtime flexibility into your designs.” Quite simple: object composition – relationships between objects – can be changed at runtime in a strongly-typed language like Java. Inheritance – the only other way of introducing externally-defined capabilities to a class – cannot.

In Ruby, the very nature of the language makes much of that line of thinking obsolete. Through its dynamic nature, Ruby classes can be modified programmatically, at runtime. This statement often scares the daylights out of a Java developer when it hits them for the first time. Upon closer inspection, though, this language feature can be used to great effect. Behavior and associated information can be ascribed to specific instances of classes based on the situation in which they find themselves. Imagine a Person object who happens to walk into a busy bank and gets into the back of a queue. In a Ruby system describing this Person, we could blend in methods like start_waiting and elapsed_wait_time? – and those methods could then be discarded later when the Person finishes its transaction and leaves the bank.

How might we have done this in Java? After all, situationally-appropriate behavior is not unique to Ruby – it’s a generalized concept that could be expressed in any language. In Java, where class behavior is fixed at compile time, we use a specialized form of composition. The Gang of Four called it the Decorator pattern. So we have a situation where an inherent language feature of one language (Ruby) completely obliterates the need for a design pattern that helped legions of OO problem-solvers in earlier languages.

It turns out that the evolution away from patterns toward language features is something that has been noted by many others. Suggested reading for those interested:

• http://designpatternsinruby.com/section01/article.html
• http://lukeredpath.co.uk/blog/decorator-pattern-with-ruby-in-8-lines.html

Just as a more or less holistic picture of this “new world” was beginning to form in my head, Adam Wiggins of Heroku published a very nice, quick-read “web booklet” – The Twelve-Factor App. It’s not deep, but does a great job of circumscribing the set of concerns one should have in mind when building an app that is destined to be deployed in the cloud.

If you are relatively new to cloud-based deployment of web applications, it may not be obvious that deployment to a cloud-based server should affect your application in so many ways (thread management, stateless processes, absolute separation of configuration data from source code, etc.). Build, deploy and manage an app or two, and you’ll learn some of these lessons the hard way.

If you do open source development (like Ruby on Rails), learning the hard way has probably become a way of life. Do yourself a favor and remind yourself how nice it is when someone takes the time to put up a few guideposts to mark your way, to keep you on the trail and out of the weeds (or over the cliff’s edge). Check out this informative site. You can get through the content in fifteen minutes. Thirty if you take time to let the points sink in a bit. You’ll save yourself a lot more than that if you use these concepts when you build your next hosted application.

Most of these features – closures, ultra-convenient collection-oriented operators like the spread-dot (*.), etc., are well-documented. Some, however, are a bit more arcane, and not realizing they exist can actually cost you a lot of time if you stumble into them inadvertently.

The feature I have in mind at the moment is a very convenient but somewhat surprising bit of collection behavior in Groovy. I’ve been working with the language for quite a while, and have dutifully pored over books and language reference documents to become familiar with all of the groovy goodness at my disposal, but somehow I missed this one until someone on my project team found it – quite by accident.

Here’s the magic find. In Groovy, if you have a list of objects, you can access, on the list itself, properties that belong to objects within the list. The result, somewhat intuitively (but somehow quizzical at the same time), is a list of values of that property across all the members in the list. Here’s an example to illustrate what I mean – just copy this and run it as a Groovy script.

class Player {
def name
def teams

public String toString() {
"${name} - played for ${teams}"
}
}

def players =
[
new Player(name: "Greg Maddux", teams: ["Cubs", "Braves"]),
new Player(name: "Ron Santo", teams: ["Cubs", "White Sox"]),
new Player(name: "Billy Williams", teams: ["Cubs", "Athletics"]),
new Player(name: "Bruce Sutter", teams: ["Cubs", "Cardinals"]),
new Player(name: "Lou Brock", teams: ["Cubs", "Cardinals"]),
new Player(name: "Gary Gaetti", teams: ["Twins", "Cubs"]),
new Player(name: "Ron Cey", teams: ["Dodgers", "Cubs"]),
new Player(name: "Ryne Sandberg", teams: ["Cubs"]),
new Player(name: "Mark Grace", teams: ["Cubs", "Diamondbacks"]),
new Player(name: "Goose Gossage", teams: ["Athletics", "Yankees", "Cubs"]),
new Player(name: "Sammy Sosa", teams: ["Rangers", "White Sox", "Cubs"]),
new Player(name: "Larry Bowa", teams: ["Phillies", "Cubs"]),
new Player(name: "Rick Monday", teams: ["Dodgers", "Cubs"]),
new Player(name: "Dave Kingman", teams: ["Rangers", "Mets", "Cubs"]),
new Player(name: "Joe Girardi", teams: ["Cubs", "Cardinals", "Yankees"])
]

def playerNames = players.name
println "Here, amazingly enough, is a list containing the names of all those players:\n $(playerNames)"

Once you know this behavior exists, it’s pretty easy to find situations where it’s an insanely convenient feature. If you don’t know it exists, you can get some pretty confusing results. In our case, we were running a GORM query to retrieve a domain object using an alternate ID (expecting a single result), and incorrectly used the list() method (we should have used get(), which returns the domain object, whereas list() always returns a list).

In our code, after querying for the domain object, we accessed properties of the object, and were puzzled when instead of getting Strings and Integers and Dates, we were getting single-member lists of Strings and Integers and Dates. Each list contained the value we expected, inexplicably embedded in a List instance. It took a little head-scratching before we figured out that the root cause was that our query was returning a list containing our domain object, and our code that accessed the properties, which would have failed miserably in Java if attempting to treat a List as though it were a domain object, was happily drilling through the wall of that list and getting at its contents.

Syntactic sugar? Definitely. Violation of the principle of least surprise? Maybe. Will I use it, now that I know it’s a tool at my disposal? Absolutely.

Stay Groovy, my friends…

But such peace of mind comes at a cost. I’ve heard estimates from various teams I’ve talked to stating that the cost of writing unit tests for adequate code coverage (usually 50% to 80% on Java projects) can range from equivalent to the production code itself, to five times that. The problem with these estimates, of course, is subjectivity. For example, if you have found a bug using a test and work to fix the test, are you mentally counting that time as test-writing or code-writing?

Estimating factors aside, the plain fact is that testing does amplify the cost of software development – at least in the short term. Which means that a software engineer who really cares about his customer’s bottom line should always be on the lookout for ways to avoid the increased cost of testing.

This line of thinking arose the other day on the project I’m currently working on. It’s a Groovy/Grails web application with some public, customer-facing components and some internal-use components. We’re using Grails scaffolding as much as is practical for the relatively infrequently-used internal-use features to save time and money. In most cases, we have generated and tweaked the UI components (GSP pages) to smooth over some of the uglier bits of scaffolded pages, but we have tried not to generate any of the controller code, instead relying on Grails’ generate-at-runtime approach by including def scaffold = true in these controllers.

Why not generate the source? Because generated code is deceptively expensive! The minute a feature springs into existence in your code base, it needs to be tested – even if that code was generated by a framework like Grails. And Grails controller closures – insert and update, as the case in point here, do a significant amount work for you. Data binding. Optimistic locking exception handling. Routing of validation errors. All things that “just work” in scaffolded code, but which might be broken by someone once the code actually becomes part of your code base.

So when the need arose to customize part of the logic in insert and update closures (to do some custom interpretation of a series of check boxes on a form that couldn’t be made to conform to the default binding mechanism), we began looking for ways to override binding logic without generating and modifying the closures themselves.

As usual, Grails has a solution at the ready – controller interceptors. All we had to do was declare an interceptor that fires before insert and update closures, and in that closure, inspect our list of check boxes and set a params value that would then be assigned to our domain object through the normal binding process.

By using an interceptor, we completely avoided the need to generate (and assume custodianship of) the insert and update closures in this controller – which means we didn’t have to spend valuable time writing unit tests to cover these Grails-supplied features that could be broken once the code actually exists.

I guess I can summarize this little victory this way: Having tested code for an application feature is far better than having untested code. Having no code at all is better than either.

Stay Groovy, my friends.

See, technology is my day job.  When I’m not working (or walking the dog, or schlepping kids around town), I usually have some sort of musical instrument in my hands. or in front of me. Guitar, mandolin, bass, keyboards, drums and trumpet, for now.  Hope to add to the collection soon – I think I need a bouzouki.  Or maybe a dobro.  But I digress…

When I have a block of free time and a little peace and quite in the house, I like to break out my recording gear and see what I can throw together.  And, occasionally, share it with a few friends to see what they think.  Or even send a track to a friend who’s a real drummer to lay down a drum track for me to mix in later.

In the past, sharing music online has been inconvenient at best.  Oh, there are music sharing sites like SoundClick, or social sites like MySpace that cater to musicians.  But they all have some pretty serious limitations (like having to actually log into their service to find and listen to tracks).

SoundCloud is different.  It’s a seriously open, cloud-based music storage and publishing service.  I just signed up an hour ago, and within minutes I was uploading my first track.  (I love, by the way, that you specify the file to upload first, and it automatically starts uploading while you fill in the track info form.)

But the really cool part is the openness of the service.  There’s no need for anyone to go to their site to hear my music.  I can email it to them, or drop a widget into Facebook or a blog entry (as I’ve done here), and voila.  Music.  Where I want it.  No strings attached.

And now, if you’ll excuse me, I have a recording project to dive into…

There is one thing, however, that has bothered me since I started using Hibernate several years ago.  A deep bother.  Not a “this is a nuisance, but I can live with it” bother, but more of a “this is so fundamental an issue that there just has to be a better solution to it than is being offered by Hibernate’s authors” bother.

The problem:  bridging the gap between object identity and database identity.  In most modern database schemas, tables are defined with a numeric primary key column, which is populated through some unique key assignment mechanism (Identity columns in SQL Server or MySQL, sequences in Oracle, etc).

This works great at the database layer.  But problems arise in the JVM (or CLR for you NHibernate folks).  In order for objects representing database entities (those managed by Hibernate) to be handled properly in collections (including Hibernate cache), objects that have the same database identity need to be recognized as equals as well.  In Java, this means overriding equals() and hashCode() such that any two objects that represent the same database entity will be equal, and return the same hash code.

The “obvious” solution, at first blush, would be to base equals() and hashCode() on entity ID values.  This works great for entities created by Hibernate, which represent existing rows in a database, all with unique ID’s.  The problem has to do with new entites that you create programmatically, which haven’t yet been saved.  Those entities will typically have a null ID value – and if you base equality on ID values, any two unsaved instances will be considered equal.

There are actually two problems when dealing with unsaved entity objects.  The first, as stated above, is that they don’t yet have a unique identity – so multiple new instances will collide in collections like Maps and Sets.  The other problem is that the identity of these objects will be set at some point, which violates part of the contract of hashCode() – the part that says its value must not mutate over time.

To avoid all of thes problems, Hibernate’s authors have long recommended a workaround – use “natural keys” instead of identity values for equals() and hashCode(). The problem with this is that many entities simply don’t have a natural key – that is, a set of one or more attributes whose values uniquely identify the entity.  And those that do often fail the hashCode() qualification that the value should not change over time.

There are other possible solutions I’ve seen or come up with myself, but all had at least one glaring weakness.  It seemed that there was no silver bullet for dealing with database-generated identity values in Hibernate.

And therein lies the silver bullet.  Don’t use the database (or Hibernate) to generate identity values.

I stumbled across a wonderful article on O’Reilly’s OnJava site that quite clearly describes this problem (in more detail than I do here), and offers what I believe to be the only real solution.  Check it out at http://www.onjava.com/pub/a/onjava/2006/09/13/dont-let-hibernate-steal-your-identity.html.

And now, as I look out the window at the late February snowstorm, I think maybe it’s time for me to hibernate…