The design side of programming language design

The word "design" is often used when talking about programming languages. In fact, it even made it into the name of one of the most prestigious academic programming conferences, Programming Language Design and Implementation (PLDI). Yet, it is almost impossible to come across a paper about programming languages that uses design methods to study its subject. We intuitively feel that "design" is an important aspect of programming languages, but we never found a way to talk about it and instead treat programming languages as mathematical puzzles or as engineering problems.

This is a shame. Applying design thinking, in the sense used in applied arts, can let us talk about, explore and answer important questions about programming languages that are ignored when we limit ourselves to mathematical or engineering methods. I think the programming language community is, perhaps unconsciously, aware of this - one of the reviews of my recent PLDI paper said "this is a nice, novel design paper, and the community often wants more design papers in our conferences". The problem is that we we do not know how to write and evaluate work that follows design methodology.

To better understand how design works, I recently read The Philosophy of Design by Glenn Parsons. The book perhaps did not answer many of my questions about design, but it did give me a number of ideas about what design is, what questions it can explore and how those could be relevant for the study of programming languages...

If you are interested in new ways of thinking about programming, please join our new Shift Society mailing list. We'll use it to share announcements about events like Salon des Refusés and provide space for discussions on the future of programming.

Is "programming language design" a design activity?

First of all, is it useful to explore how design thinking applies to programming languages? Perhaps the word "design" in the term "programming language design" is misleading and it really is just engineering and mathematical task. In the first section, Parsons tries to answer the question "What is design?" and many of the points made in the section made me think of programming languages, so I believe (some aspects of) programming language design are, indeed, design activities.

How design changes the world

One of the very broad definitions of design says that "design [is] the intentional creation of plans for a new kind of thing". This is perhaps too broad, but it captures an important fact - design is relevant when you are trying to create something innovative and new that did not exist before. For me, this is the most interesting part of programming languages, but it is the one that we rarely talk about. Instead, we evaluate (prove, measure, compare) different aspects of a language once it is created, but we almost never discuss how that creation happened.

Design produces items that have the primary function of altering the world, rather than explaining it, whereas in sciences, the primary function is explanation.

The iPhone was a successful design work, because it changed the world by creating a new kind of thing. It changed how most of us live today. For me, programming language design is interesting for the same reasons. It has the potential to transform how we interact with computers and, possibly, the world.

How programming languages change the world

Object-oriented programming certainly changed how we think about and construct programs. In case of Smalltalk, this was a part of more ambitious Dynabook design vision, yet, when you read the Wikipedia page on Smalltalk, it spends half of the page on syntax and control structures and says very little about the design thinking behind it. Similarly, I find type providers interesting because they change how we interact with data and can perhaps democratizing access to data, yet, my aforementioned paper was accepted because it had sufficiently convincing proofs.

Once we start looking at programming languages through the design perspective, we can start thinking about the principles behind a particular design and the subtle influences that such design decision has on the world. Let's look at two examples.

Exhibit 1: Abstraction and design

Design may seem merely a matter of aesthetics, but this is not the case. The way everyday things are designed can have subtle influence. A nice example mentioned by Parsons is the design of a dining table:

The design of a dining table might seem to be a mere aesthetic choice, but this is not quite so. At a rectangular table, someone sits at the "head position," whereas at a round table, everyone has an equal position. The effect is subtle, something that the people involved may not even be aware of, but differences of hierarchy and status are often shaped and maintained by just such subtle influences.

This is a great example of an influence of design that you will not realize unless you explicitly think about design. Exploring the consequences of design choices on how people think about and use programming languages seems like one interesting area of work.

Subtle influence of design

The fact that a table is rectangular does not mean that people involved cannot be equal partners, but its subtle influence makes this unlikely. Similarly, a programming language can make certain things possible, but it can be designed in a way that makes people unlikely to use them. As a simple example, if you declare mutable variables using let mutable as opposed to just let, you will probably use immutable variables more often.

This is a simple obvious example, but it is an example of design aspect of programming language that we rarely talk about. Are there more notable examples of the subtle influence of design in programming languages? I think abstraction is one of them...

Abstraction and power

Many programming languages provide some mechanism for abstraction. In object-oriented languages, you can make some members of an object private and you control what is exposed via a public interface. Similarly, a function in most languages exposes just a public interface - you can call it - but it hides the internals of how it works. The idea is that the functioning of a function or internals of an object are irrelevant and the caller should not see them.

This design decision has an important consequence. It means that the library author becomes the person at the "head position." They get to decide what is the right use of their components and what is not. The class distinction caused by abstraction has been hugely influential and people nowadays often say that a certain complex language feature (e.g. fancy types) need to be understood only by "library authors" and not by "library users".

Would an alternative design be possible and reasonable? In a paper Tracing a Paradigm for Externalization: Avatars and the GPII Nexus (PDF) from Salon des Refusés, Colin Clark and Antranig Basman very convincingly argue that systems which do not hide internals behind abstraction can be surprisingly long-living, precisely because they do not dictate how the components should be used.

Perhaps then, thinking about programming language design from the design perspective would let us discover more of such subtle influences of design, explore the alternatives and advance the discipline.

Exhibit 2: Modernist ideas and programming

Mathematical models can be used to prove properties about models of programs. Engineering methods can be used to make the construction process more robust. Design thinking complements these two by letting us talk about how programming languages, libraries and programs are created. What are the guiding principles that made them look the way they look? And what are these principles aiming to achieve?

Form follows function

Parson's book frames the discussion about design using modernist ideas:

Modernism's "reconceptualization" of Design problems is nicely summed up in the famous slogan "Form follows Function." The central idea is that if the Designer merely constructs the object to perform its function, then expression, aesthetic value and mediation become, as it were, "spin-off" values that follow effortlessly.

What would the design of a programming language or a library look if it followed the modernist slogan "Form follows Function?" Whether this is a design principle worth following for programming or not is another question - just like you might prefer Art Deco architecture over modernist architecture, perhaps you prefer more decorative programming languages over modernist ones.

However, framing discussion about programming languages in terms of guiding design ideas might give make the design process easier and clearer and it can also help us avoid some of the pointless discussions about languages. For example, very few people argue whether "modernism is better than postmodernism". People discuss the differences and their personal preference, but the question of which one is better makes little sense.

Ornament and modernism in programming

As an example of what I have in mind with different design principles, let's look at two small examples of parser combinator code snippets. I think the topic is worth a separate post, but I wanted a quick illustration. First, here is a part of the JSON parser using the FParsec library for F#:

let listBetweenStrings sOpen sClose pElement f =
  between (str sOpen) (str sClose)
          (ws >>. sepBy (pElement .>> ws) (str "," >>. ws) |>> f)

let jlist =
  listBetweenStrings "[" "]" jvalue JList

let keyValue =
  stringLiteral .>>. (ws >>. str ":" >>. ws >>. jvalue)

In this example, there is more to the "Form" than just the "Function". To make the code shorter and nicer to read, the form of it uses a number of custom ornamental operators such as .>>., >>. and |>>. You need to learn what is behind those ornaments, but then you might appreciate reading code written in this style.

If you prefer a modernist design and believe that ornament is a crime, you might prefer to write the code in a way where the form directly follows function. As an example consider the following Brainfuck parser in Clojure:

(defparser instruction []
  (choice (char \>) (char \<) (char \+)
          (char \-) (char \.) (char \,)
          (between (char \[) (char \]) (many (instruction)))))

(defparser bf []
  (many (instruction))

Here, the form follows directly the function given the material (language) constraints. Two notes are in order. First, I'm not saying that one is better than the other. I have my personal preference, but that is just a personal preference. Second, I don't think the language dictates the style in this case. I can perfectly imagine ornamental parser library for Clojure and modernist parser library for F#.

Just like with the subtle influence of design, I think the above example is merely scratching the surface. I believe the modernist principle that form follows function could be used as a building block for a new, interesting, kind of programming languages. This way of thinking about programming is very much unexplored. As far as I'm aware, the only relevant paper here is Notes on postmodern programming (PDF) by James Noble and Robert Biddle. I think there should be many more!

Side-note: How design lost to mathematics and science

Before wrapping up, there was one more remark in Parsons book that I found very interesting in the context of programming languages. One of the design movements mentioned in the book aimed to create "design science", partly in reply to the fact that design was not recognized as a reputable academic discipline:

Simon (...) argued that the reliance on intuition had led to a marginalization of Design within the contemporary university. In the university setting, Simon argued, the highest prestige accrues to the pure sciences, due to their much vaunted "scientific method" (...). Situated next to the pure sciences, Design (...) looked flimsy and amateurish.

I believe this also explains why design thinking got lost from academic thinking about programming languages. Compared to aspects that can be measured or formally proved, the design issues do perhaps look flimsy, but I believe they are equally important - they simply look at questions that cannot be answered with engineering, scientific or mathematical ways of thinking. In contrast to design, computer science took the right approach in order to succeed in the modern university. As noted by Nathan Ensmenger in The Computer Boys Take Over:

The rise of theoretical computer science was anything but inevitable. (...) Advocates of theoretical computer science pursued a strategy that served them well within the university, but increasingly alienated them from their colleagues in the industry.

Thanks to the success of theoretical computer science, there are now academic departments that study things like programming languages. Sadly, it also means that the methods of study are limited to those that the are prestigious in the academic environment. I believe that there is a room for looking at programming languages from the perspective of intuitive, (flimsy and amateurish) design perspective, because it is the best we have for answering a number of important programming language questions.

Summary: Towards resurrecting design

In this blog post, I looked at two aspects of design thinking that can be applied to programming language design. The first was the subtle influence of design - seemingly standard and uninteresting features of design such as rectangular shape of a table have important consequences. The same is the case with, for example, abstraction in programming languages. The second theme was design principles such the modernist mantra that form follows function. How can those principles be exhibited in programming language design and what other guiding principles can we find?

I believe that asking those questions would enrich the discussion about programming languages. As mentioned earlier, design is the intentional creation of plans for a new kind of thing with the primary function of altering the world - I think that there is a lot of unexplored space in programming languages and treating them from the design perspective is one way to exploring those new areas.

I also believe that interesting programming language research is not about finding a keyword that will make writing for loops 3x easier, but about changing how we interact with the world. For example, can programming language research help democratize access to data and fight "fake news"?

One initiative in this direction has been the recent Salon des Refusés workshop that I helped to organize. We hope to host another one at @programmingconf in Nice in 2018 and we will also be launching a new web page for the Salon soon, so if you made it to the end, follow me at @tomaspetricek to hear more soon! I would also love to hear your ideas about how design thinking can be applied to programming languages and what good materials on "design thinking" are there!

If you are interested in new ways of thinking about programming, please join our new Shift Society mailing list. We'll use it to share announcements about events like Salon des Refusés and provide space for discussions on the future of programming.

val listBetweenStrings : sOpen:'a -> sClose:'b -> pElement:'c -> f:'d -> 'e
val sOpen : 'a
val sClose : 'b
val pElement : 'c
val f : 'd
val jlist : obj
val keyValue : obj

Published: Tuesday, 12 September 2017, 5:42 PM
Author: Tomas Petricek
Typos: Send me a pull request!
Tags: academic, research, programming languages, philosophy, design