Originally posted internally 12 Jan 2016:

PC sends out a weekly update for things happening in the Workplace offering, and this week he touched on the topic of programming languages. Rather than replying to him 1:1 by email I’m writing here because I’d like to have a more open debate. Before we start… take a look at the RedMonk Programming Language Rankings (latest is June 2016), which PC referred to in his newsletter.

A couple of the languages that PC touched on are Scala and R. They’re both interesting, though for different reasons.

Scala caused my friend Kirk Wylie to write I Want A New Programming Language where he introduced a very interesting concept of a ‘Journeyman Language’. His general point was that Scala was too high end, and if he let (some of) his team develop in Scala then he’d end up with a part of the OpenGamma code base that could only be maintained by Scala experts.

Kirk brought up the concept of a ‘Journeyman Language’ before Go (or GoLang) was launched, but Go is now celebrating its 5th birthday, and in its short life just about all modern infrastructure software has been written or rewritten in Go (e.g. Docker, Cloud Foundry, Teraform, Kubernetes). I asked Kirk if Go fitted his requirements for a Journeyman Language, and he said it didn’t; though it seems to me that although it might not have all the bells and whistles Kirk was after it’s doing exactly that job at Google and elsewhere – allowing relatively inexperienced developers to be productive, and create code that’s comprehensible to others (after all the point of a programming language is to create stuff that’s understandable by other humans – compilers take care of making things work on the machines).

There are two other interesting things about Go:

1. It’s the first language since C that’s a systems language and an application development language.
2. It has great support for concurrency, having adopted the communicating sequential processes (CSP) approach that debuted in Occam – a language developed to work on Transputers, which were one of the first truly parallel processing substrates. Former Netflix Chief Architect (and all round tech superhero) Adrian Cockcroft does an excellent job of connecting the present to the past in his Gophercon 2016 presentation Communicating Sequential Goroutines. The key here is that Go (and its idioms) has made parallel processing easy and accessible, and since we now have multi core CPUs (and GPUs) all over the place that’s a good thing.

The language wonks’ language that’s appeared alongside Go is Rust, which comes from Mozilla (the Firefox browser people). Rust has a smaller footprint than Go (which might be a big deal for IoT applications), but it seems to have a steeper learning curve – so it’s less of a ‘Journeyman Language’. The Chef people recently launched their Habitat automation tool, which was written in Rust, but it’s a nightmare of complexity that will prevent the kind of grass roots uptake we’ve seen with Docker. Rust might not be to blame for that, but it’s a bad start for it in terms of mind share and impact.

The point to this post is that learning programming languages is an investment – it takes time to become proficient in a language, but that investment pays off when the language lets you solve a problem more quickly and efficiently. It’s easy to be reductionist about this and think too much about individual productivity, but Kirk’s point was about team productivity (and longevity) and there’s a broader point these days with open source with respect to community productivity (and longevity) – if you can bring other people into you project as collaborators then that becomes a force multiplier – so pick a language that’s popular and effective.

The final point is about context. R is great for statistics, which is why our data scientists use it as part of Operational Data Mining, though they also use a ton of Python. Scripting languages like Bash used to be where the action was at with systems operations, but the game has moved to Go and configuration management tools like Ansible (which is based on Python), Puppet and Chef (which are based on Ruby). The fact that Ruby has been squeezed out of tools like Cloud Foundry in favour of Go shows a meaningful direction of travel. For application development the RedMonk team have a saying that when apps grow up they get rewritten in Java, though it’s now clear that many adopters of microservices are doing so at least in part to allow some developers to use Go, Node.js or whatever else takes their fancy. I can see some reason for arguing that beginners should learn JavaScript, because it has utility on the client and server side (making it just about ubiquitous), but JavaScript has some awful idiosyncrasies and weirdness, which brings truth to the saying ‘In Ruby, everything is an object. In Clojure, everything is a list. In Javascript, everything is a terrible mistake.’

Retrospective

Another RedMonk Programming Language Rankings came along in January, which means we’re not that far from the next cut. Last time around it seems that Swift and Typescript were most worthy of comment.

I’m inclined towards thinking that the most important thing to learn is Markdown, but that’s a topic for another post.

Meanwhile we ran the DXC Codes competition for school kids, which we based around Scratch – a language that can trace its origins back to LISP via LOGO. There were some amazing entries.

Original Comments

NS

I can see where Kirk is coming from, but it appears he’s nuts. Who writes their own RDBMS?

“if you could write a reasonably high performance RDBMS system in the language, it has enough features. If you couldn’t, it’s not good enough. I like this particular test because I’ve done it several times”

These days we operate in a landscape where developers should use the best programming language / framework (very important) for the job, rather that what happens to be the current flavour of the month.

HS

R (and R Studio) is absolutely the way to go as entre to data science, with useful programming by-product. Programmers might come at it from the other direction, Python and Scikit-Learn, a popular machine learning library.

HS

It has always been somewhat amazing to me how many new languages emerge over time. Evolution of computer programming languages (original link lost on C3) And I know the dilemma of trying to decide which ‘next’ one to learn for those still active in the development of their software engineering career. I suppose it all depends where you start? An Excel macro person might like to try Python or R. A professional Java or C# programmer might be interested in understanding why they might be interested in Go, Scala, Hasskell, Julia, Lua, Erlang, Rust, Clojure, The ‘Full Stack’ developer will need to know how to mesh a bunch of approaches. For those who specialise in SQL, they might want to explore how that seques into machine learning frameworks. Etc.

I must have a stab at writing that ‘Why I should learn to program in <X> or can afford to ignore it?’  for all these languages, and more.