Welcome to the jOOQ Tuesdays series. In this series, we’ll publish an article on the third Tuesday every other month where we interview someone we find exciting in our industry from a jOOQ perspective. This includes people who work with SQL, Java, Open Source, and a variety of other related topics.
Mario, a long time ago, I have already stumbled upon your name when looking up the author of Lambdaj – a library that went to the extreme to bring lambdas to Java 5 or earlier. How does it work? And what’s the most peculiar hack you implemented to make it work?
When I started developing Lambdaj in 2007 I thought to it just as a proof-of-concept to check how far I could push Java 5. I never expected that it could become something that somebody else other than myself may actually want to use. In reality, given the limited, or I should say non-existing, capabilities of Java 5 as a functional language, Lambdaj was entirely a big hack. Despite this, people started using and somewhat loving it, and this made me (and possibly somebody else) realize that Java developers, or at least part of them, were tired of the pure imperative paradigm imposed by the language and ready to experiment with something more functional.
The main feature of Lambdaj, and what made its DSL quite nice to use, was the possibility to reference the method of a class in a static and type safe way and pass it to another method. In this way you could for example sort a list of persons by their age doing something like:
As anticipated what happened under the hood was a big hack: the on() method created a proxy of the Person class so you could safely call the getAge() method on it. The proxy didn’t do anything useful other than registering the method call. However it had to return something of the same type of the value returned by the actual method to avoid a ClassCastException. To this purpose it had a mechanism to generate a reasonably unique instance of that type, an int in my example. Before returning that value it also associated it, using a WeakHashMap, to the invoked method. In this way the sort() method was actually invoked with a list and the value generated by my proxy. It then retrieved from the map the Java method associated with that value and invoked it on all the items of the list performing the operation, a sorting in this case, that it was supposed to execute.
That’s crazy :) I’m sure you’re happy that a lot of Lambdaj features are now deprecated. You’re now touring the world with your functional programming talks. What makes you so excited about this topic?
The whole Lambdaj project is now deprecated and abandoned. The new functional features introduced with Java 8 just made it obsolete. Nevertheless it not only had the merit to make developers become curious and interested about functional programming, but also to experiment with new patterns and ideas that in the end also influenced the Java 8 syntax. Take for instance how you can sort a Stream of persons by age using a method reference
It looks evident how the method references have been at least inspired by the Lambdaj‘s on() method.
There is a number of things that I love of functional programming:
- The readability: a snippet of code written in functional style looks like a story while too often the equivalent code in imperative style resembles a puzzle.
- The declarative nature: in functional programming is enough to declare the result that you want to achieve rather than specifying the steps to obtain it. You only care about the what without getting lost in the details of the how.
- The possibility of treating data and behaviors uniformly: functional programming allows you to pass to a method both data (the list of persons to be sorted) and computation (the function to be applied to each person in the list). This idea is fundamental for many algorithms like for example the map/reduce: since data and computation are the same thing and the second is typically orders of magnitude smaller you are free to send them to the machine holding the data instead of the opposite.
- The higher level of abstraction: the possibility of encapsulating computations in functions and pass them around to other functions allows both a dramatic reduction of code duplication and the design of more generic and expressive API.
- Immutability and referential transparency: using immutable values and having side-effects programs makes far easier to reason on your code, test it and ensure its correctness.
- The parallelism friendliness: all the features listed above also enable the parallelization of your software in a simpler and more reliable way. It is not coincidence that functional programming started becoming more popular around 10 years ago that is also when multicore CPUs began to be available on commodity hardware.
Our readers love SQL (or at least, they use it frequently). How does functional programming compare to SQL?
The most evident thing that FP and SQL have in common is their declarative paradigm. To some extent SQL, or at least the data selection part, can be seen as a functional language specialized to manipulate data in tabular format.
The data modification part is a totally different story though. The biggest part of SQL users normally change data in a destructive way, overwriting or even deleting the existing data. This is clearly in contrast with the immutability mantra of functional programming. However this is only how SQL is most commonly used, but nothing dictates that it couldn’t be also employed in a non-destructive append-only way. I wish to see SQL used more often in this way in future.
In your day job, you’re working for Red Hat, on drools. Business rules sound enterprisey. How does that get along with your fondness of functional programming?
Under an user point of view a rule engine in general and drools in particular are the extreme form of declarative programming, second only to Prolog. For this reason developers who are only familiar with the imperative paradigm struggle to use it, because they also try to enforce it to work in an imperative way. Conversely programmers more used to think in functional (and then declarative) terms are more often able to use it correctly when they approach it for the first time.
For what regards me, my work as developer of both the core engine and the compiler of drools allows me to experiment every day in both fields of language design and algorithmic invention and optimization. To cut it short it’s a challenging job and there’s lot’s of fun in it: don’t tell this to my employer but I cannot stop being surprised that they allow me to play with this everyday and they also pay me for that.
You’re also on the board of VoxxedDays Ticino, Zurich, and CERN (wow, how geeky is that? A large hadron collider Java conference!). Why is Voxxed such a big success for you?
I must admit that, before being involved in this, I didn’t imagine the amount of work that organizing a conference requires. However this effort is totally rewarded. In particular the great advantage of VoxxedDays is the fact of being local 1-day events made by developers for developers that practically anybody can afford.
I remember that the most common feedback I received after the first VoxxedDays Ticino that we did 2 years ago was some like: “This has been the very first conference I attended in my life and I didn’t imagine it could have been a so amazing experience both under a technical and even more a social point of view. Thanks a lot for that, I eagerly wait to attend even next year”. Can you imagine something more rewarding for a conference organizer?
The other important thing for me is giving the possibility to speakers that aren’t rock stars (yet) to talk in public and share their experience with a competent audience. I know that for at least some of them this is only the first step to let themselves and others discover their capabilities as public speakers and launch them toward bigger conferences like the Devoxx.
Thank you very much Mario
If you want to learn more about Mario’s insights on functional programming, please do visit his interesting talks at Devoxx from the recent past: