Monday, January 2, 2012

Scala for 2012? Deciding Whether to Invest In a Programming Language

I have found it both interesting and rewarding to learn a new programming language or major framework on a roughly yearly basis. If forced to self-identify with any single programming language, it would be Java. However, over the years, I've used C and C++ fairly extensively and have used and learned enough to be dangerous about several other languages including shell scripting languages, Perl, JavaScript, Pascal, C#, Ruby, JRuby, Groovy, PHP, and Python. Of the latter group of languages (Ruby, JRuby, Groovy, PHP, and Python), Groovy has had the most practical benefit for me, but I have learned valuable idioms, best practices, and different ways of thinking from using the other languages.

In some years, I've not been as quick to learn a language in a year in which I've been learning either a major framework for that language or in a year in which a language I am familiar with undergoes significant changes. For example, Struts and the Spring Framework dominated my time as I learned each of them. JavaFX has similarly dominated my interest in recent weeks. When not working with a new language, I tend to focus on libraries and frameworks of the languages I am comfortable with. I have spent more time on Guava this year, for example.

Learning a new language does provide many benefits. However, these benefits don't come for free. There is always an opportunity cost associated with learning anything new. If a programming language is particularly different than what one is used to, this opportunity cost can be great. The opportunity cost can be manifest as many different things. It might be lower productivity than could be had using a known language. It might be missing out on learning a new framework, library, or approach in the more familiar language. The opportunity cost might be having to settle on fewer or choosing different features that better fit the new language. The opportunity cost may be as simple as not being able to do other things one would want to do and might have time to do if using a familiar language.

Because there are so many potential opportunity costs associated with learning a new programming language, I try to be careful about which I invest my time in. I typically have a compelling reason for learning a new language. Compelling reasons might include specific advantages of a language (such as PHP for many Web 2.0-centric projects) or widespread use and "employability" of that language. Other reasons might be to learn new techniques that can be adapted to more familiar languages. Perhaps the most compelling reason I've learned a new language has been to read and maintain code or scripts that I have handed to me and am assigned responsibility for.

For several years now, I've been somewhat curious about Scala, but have not yet committed myself to using it and learning it because other languages, frameworks, and tools have grabbed my attention. Typically, I've had some motivation that has made these tools, languages, or frameworks seem most worth my investment of time and energy. For example, the need to have a nice scripting language that meshes well in my Java development environment led me to Groovy. My attendance at JavaOne 2010 and JavaOne 2011, coupled with my interest in a modern GUI technology, has led to my interest in JavaFX. I spent time with Python after being in a position where I needed to read and modify Python scripts.

I recently welcomed the opportunity to pose some questions to Scala creator Martin Odersky related to what is in Scala that might motivate me (and others) to invest time and energy into learning Scala. As I articulated the questions I had for Martin, I realized that these are really the things that I informally look at when investigating a new language. I typically spend an hour to two finding a language's highest-level motivations first and only invest more time in that language if it seems to be a good potential fit for me. Martin has agreed to me posting my questions and his answers and they are shown next (I have added hyperlinks).

Question: What is the most compelling/motivating reason or reasons that one might want to invest time in learning Scala as opposed to continuing use of Java (mostly for applications in my case) + Groovy (mostly for development environment scripting in my case)? For example, Groovy appealed to me at a high level as a way to script with libraries, idioms, and syntax that I was comfortable with from Java application development experience.
There are actually quite a few different facets of Scala that individuals attach to for different reasons. Some are attracted by the succinct syntax and resulting productivity. Others gravitate to the stability of the JVM and runtime performance of a compiled language (versus an interpreted language, like Groovy), or the ability of a sophisticated type system to help programmers avoid errors that would otherwise crop up at runtime. Others find the functional style of programming to be a more natural way to reason about their application logic.

One of the strongest attractors, from a practical point of view, is that Scala (and the rest of the Typesafe Stack -- including Akka and Play) are designed to provide better tools to address the dual challenges of parallel and concurrent programming. With the advent of mainstream multicore/manycore hardware, and the increasing scale of applications that developers are charged to build, many industry developers are looking for higher level abstractions than threads and locks for building at this next scale. Many find that the functional style, immutable state, actor concurrency model, and other concepts at the heart of Scala make it simpler to build parallel and concurrent applications.

Question: What are Scala's biggest strengths, advantages, and innovative features?
At a high level, Scala seeks to be a pragmatic language that scales from the smallest scripts to the largest distributed systems.

One major thread of innovation in Scala is its unique blend of object-oriented Java with functional programming concepts. Scala's libraries build on this foundation to provide outstanding support for concurrency and parallelism, for example through the actor programming model and the built-in parallel collections introduced in Scala 2.9.

Scala's expressive type system and syntax helps developers build more reliable code and greatly increase extensibility, especially for library developers and those building domain-specific languages (DSLs).

Finally, it's important not to overlook the fact that Scala is deeply integrated with Java, supporting blended Scala/Java projects and allowing developers to apply their skills and investments in Java immediately when they start working with Scala.

Question: What are Scala's biggest weaknesses, disadvantages, and plans for improvement?
One of the challenges for a relatively young language like Scala is the maturity of tools. In particular, the Scala IDE for Eclipse has had its rough edges in the past -- one of the reasons that Typesafe, as the leading commercial contributor to Scala, has invested substantial resources in overhauling the IDE with version 2.0 (just released in December 2011).

Another challenge for adoption is that Scala does introduce with functional programming a new mode of thinking about programs, which takes some time to learn. It makes the transition gentle, because one can start writing Scala code like more concise Java code. But as Scala's native library ecosystem grows chances are that newcomers to the language will come across to some of its more foreign features before they have developed a good understanding. To avoid culture shock, we need to develop a set of best practices and good tutorials that help the transition. "Programming in Scala", which I have co-authored, is a comprehensive tutorial of the object/functional style. Cay Horstmann's "Scala for the Impatient", available as a free preview on the Typesafe site, is a pragmatic, fast-paced introduction.

Question: What situations/scenarios/use cases is Scala best and worst suited for?
As described above, Scala and its frameworks like Akka and Play really shine for building systems that need to scale on multiple fronts -- across cores, across machines in a cloud environment, and across large software teams.

Traditionally, one area where Scala or other JVM-hosted languages would not be considered well suited would be lower-level systems programming. But interestingly, we see evidence of forward-looking systems developers increasingly embracing managed runtime languages like Scala because they face fundamental challenges in building reliable systems for the era of multicore hardware and distributed deployments.

Martin's responses validate some of my own conclusions about Scala from reading posts by Scala enthusiasts and even some of the detractors. In terms of motivation, I have a difficult time believing that learning Scala primarily as a scripting language will be very motivating because I'm pretty happy with Groovy for scripting. However, for development of applications, I tend to use Java and not Groovy and I wonder if it's in that area where I'd be most likely to benefit from learning and using Scala.

Once I determine that a language is worth investing in, the next step is deciding how to best learn it. Reading about it is a necessity, but using it is what really helps me learn it and also what helps me identify the things I don't like about it. The trick is to come up with a somewhat realistic example that is easy enough to implement, but interesting enough to prove out some concepts. A "Hello World" is okay to get one's feet wet, but doesn't really test how the language fares for a developer's specific needs. My favorite initial examples are ones that actually provide benefit in addition to being a mechanism for learning. For example, when I was learning Groovy, I developed several scripts early on that were helpful to me as scripts in their own right regardless of the language they were written in. In those cases, I gained familiarity with Groovy while also receiving other utilitarian benefits.

The JavaWorld article Learn Scala with Specs2 Spring describes how a Java developer who uses the Spring Framework can use the author's company's Specs2 Spring for integration testing and also benefit from "an efficient and safe way to learn the patterns of object-functional programming with Scala." The entire premise of this article is exactly the kind of thing I like to do when learning a new language: combine legitimate benefit with learning of the new language.

One other thing to think about when trying out a new language is to ensure that one is trying it out for the correct situations. This was easy for me with Groovy: I tried out Groovy first in situations in which I wanted the power of the JVM or the scope of the JDK, but wanted a scripting-friendly language. A developer can quickly decide a language is "not good" simply because the situation in which the language used is not a great fit for that language. Related to this, another issue I try to keep in mind when learning a new language is that it's not fair to compare a language I know well and have spent thousands of hours with to a language that I've spent a few hours with. Unless I encounter some real deal-breakers early in the process, I try to not let "little problems" or things I don't like about the new language prevent me from giving it a real chance. An excellent recent post on this is Rob Pike's Esmerelda's Imagination. All that being stated, there are times I run into a true deal-breaker that makes me realize I should not invest any more time in a particular language because it doesn't fit my needs. That doesn't necessarily mean there's anything wrong with the language, but simply that it doesn't fit my needs well. An example of this would be using Java for a real-time system in Java's early days.

I think I'm almost ready to commit to spending more time with Scala. I'm not the type to make new-year resolutions, but it just so happens that it seems like the right time to give Scala a closer look. If I really do start to invest more time in Scala, my plan is to first re-read Bruce Eckel's Scala: The Static Language that Feels Dynamic and then read the A1 chapters of Scala for the Impatient, trying out and adapting examples. If I'm still interested in Scala after that, I can invest more at that time.

Do I still have reservations about spending time on Scala? Of course. One of my biggest concerns is best articulated by someone who actually seems to have tried out Scala. Cédric Beust states, "In my experience with Scala, it's hard not to like the language in the first week and it's hard to still be in love with it after reading the 700+ pages of a book about it." On the other hand, Casper Bang articulates well why I think I maybe I should spend time with Scala despite any other obvious motivations: "So I guess my point is, even if I do find Scala hyperbolish and biting over a bit too much; the majority of identifiable alpha-geeks that I track, are moving this way and as a practicing professional, I can not afford to ignore this."

The post Offbeat: Scala by the end of 2011 – No Drama but Frustration is Growing and the feedback comments related to that post are insightful and seem to reiterate some of the issues that Martin pointed out that Scala must deal with. In particular, when I look at the issue most likely to deter me from spending time on Scala, it is the risk that Scala may never take hold in mainstream development. If that turns out to be the case, then the primary advantage of learning Scala would be to change my way of thinking about things and that's not always necessarily worth the opportunity cost and other costs. This post and the feedback comments contain multiple sides of the same issue and are another reminder that I probably need to do more with Scala to decide for myself how I feel about it.

My plan as of right now is to invest significant time and effort into learning basics of Scala and applying it to some "realistic" examples. I even have plans to blog on what I learn. But, I have had these types of plans before and been distracted by some other shiny thing that has come my way. I think this time will be different, but I should know for certain by the end of 2012.

6 comments:

Orest Ivasiv said...

Hi,
I understand your dilemma.
I have the same situation :)
However, I've solved it with simple checklist developed for own usage.

Have a look: How to decide which technology hype is worth personal investments?

@DustinMarx said...

Orest,

Thanks for providing the link to your post. It's a nice succinct summary of things that I too consider when deciding whether to invest time and effort into learning a new language. I also like Grygoriy Mykhalyunyo's feedback comment because I do sometimes play with a new technology just because it's interesting, but tend to only do that when the amount of effort and time required is relatively small.

Dustin

@DustinMarx said...

opensas has posted a comment on the JavaWorld-syndicated version of this post in which he references his post First steps with Scala, say goodbye to bash scripts....

Dustin

@DustinMarx said...

Jelle Victoor has posted First Steps Into Scala.

Alastair Rae said...

The lack of decent IDE support is not just a weakness but a deal-breaker for professional adoption. The existing Eclipse plug-ins have been more than just a bit rough. I'll wait to see reviews of the newer IDE support before I look at Scala again.

sunjune said...

Its almost a year now. Would like to know your opinion on Scala language so far. I have started around 3-4 months back and need to know whether I should invest in learning further. Right now it seems the Scala IDE by Typesafe is good enough and the language seems to become bit more matured with latest version. Please share your experiences.

Thanks in advance
Mohan Kumar Muddana