Thinking in Ramda: Declarative Programming
This post is Part 4 of a series about functional programming called Thinking in Ramda.
In Part 3, we talked about combining functions that take more than one argument by using the techniques of partial application and currying.
As we start writing small functional building blocks and combining them, we find that we have to write a lot of functions that wrap JavaScript’s operators such as arithmetic, comparison, logic, and control flow. This can feel tedious, but Ramda has our back.
But first, some background.
Imperative vs Declarative
There are many different ways to divide up the programming language/style landscape. There’s static typing vs dynamic typing, interpreted languages vs compiled languages, low-level vs high-level, etc.
Another such division is imperative programming vs declarative programming.
Without going too deep into this, imperative programming is a style of programming where the programmers tell the computer what to do by telling it how to do it. Imperative programming gives rise to a lot of the constructs we use every day: control flow (if-then-else statements and loops), arithmetic operators (+, -, *, /), comparison operators (===, >, <, etc.), and logical operators (&&, ||, !).
Declarative programming is a style of programming where the programmers tell the computer what to do by telling it what they want. The computer then has to figure out how to produce the result.
One of the classic declarative languages is Prolog. In Prolog, a program consists of a set of facts and a set of inference rules. You kick off the program by asking a question, and Prolog’s inference engine uses the facts and rules to answer your question.
Functional programming is considered a subset of declarative programming. In a functional program, we define functions and then tell the computer what to do by combining these functions.
Even in declarative programs, it is necessary to do similar tasks to those we do in imperative programs. Control flow, arithmetic, comparison, and logic are still the basic building blocks we have to work with. But we need to find a way to express these constructs in a declarative way.
Declarative Replacements
Since we’re programming in JavaScript, an imperative language, it’s fine to use the standard imperative constructs when writing “normal” JavaScript code.
But when we’re writing functional transformations using pipelines and similar constructs, the imperative constructs don’t play well.
Let’s look at some of these basic building blocks that Ramda provides to help us out of this jam.
Arithmetic
Back in Part 2, we implemented a series of arithmetic transformations to demonstrate a pipeline:
Arithmetic Pipeline
const multiply = (a, b) => a * b
const addOne = x => x + 1
const square = x => x * x
const operate = pipe(
multiply,
addOne,
square
)
operate(3, 4) // => ((3 * 4) + 1)^2 => (12 + 1)^2 => 13^2 => 169
Notice how we had to write functions for all of these basic building blocks that we wanted to use.
Ramda provides add, subtract, multiply, and divide functions to use in place of the standard arithmetic operators. So we can use Ramda’s multiply in place of the one we wrote ourselves, we can take advantage of Ramda’s curried add function to replace our addOne, and we can write square in terms of multiply as well:
Using Ramda‘s Arithmetic Functions
const square = x => multiply(x, x)
const operate = pipe(
multiply,
add(1),
square
)
add(1) is very similar to the increment operator (++), but the increment operator modifies the variable being incremented, so it is a mutation. As we learned in Part 1, immutability is a core tenet of functional programming so we don’t want to be using ++ or its cousin --.
We can use add(1) and subtract(1) for incrementing and decrementing, but because those two operations are so common, Ramda provides inc and dec instead.
So we can simplify our pipeline a bit more:
Using inc
const square = x => multiply(x, x)
const operate = pipe(
multiply,
inc,
square
)
subtract is the replacement for the binary - operator, but there’s also the unary - operator for negating a value. We could use multiply(-1), but Ramda provides the negate function to do this job.
Comparison
Also in Part 2, we wrote some functions for determining if a person was eligible to vote. The final version of that code looked like this:
Eligible Voters
const wasBornInCountry = person => person.birthCountry === OUR_COUNTRY
const wasNaturalized = person => Boolean(person.naturalizationDate)
const isOver18 = person => person.age >= 18
const isCitizen = either(wasBornInCountry, wasNaturalized)
const isEligibleToVote = both(isOver18, isCitizen)
Notice that some of our functions are using standard comparison operators (=== and >= in this case). As you might suspect by now, Ramda also provides replacements for these.
Let’s modify our code to use equals in place of === and gte in place of >=.
Using Ramda Comparison functions
const wasBornInCountry = person => equals(person.birthCountry, OUR_COUNTRY)
const wasNaturalized = person => Boolean(person.naturalizationDate)
const isOver18 = person => gte(person.age, 18)
const isCitizen = either(wasBornInCountry, wasNaturalized)
const isEligibleToVote = both(isOver18, isCitizen)
Ramda also provides gt for >, lt for <, and lte for <=.
Note that these functions take their arguments in what seems like normal order (is the first argument greater than the second?) That makes sense when used in isolation, but can be confusing when combining functions. These functions seem to violate Ramda’s “data-last” principle, so we’ll have to be careful when we use them in pipelines and similar situations. That’s when flip and the placeholder (__) will come in handy.
In addition to equals, there is identical for determining if two values reference the same memory.
There are a couple of common uses of ===: checking if a string or array is empty (str === ‘‘ or arr.length === 0), and checking if a variable is null or undefined. Ramda provides handy functions for both cases: isEmpty and isNil.
Logic
In Part 2 (and just above), we used the both and either functions in place of && and || operations. We also talked about complement in place of !.
These combined functions work great when the functions we are combining both operate on the same value. Above, wasBornInCountry, wasNaturalized, and isOver18 all apply to a person.
But sometimes we need to apply &&, ||, and ! to disparate values. For those cases, Ramda gives us and, or, and not functions. I think of it this way: and, or, and not work with values, while both, either, and complement work with functions.
A common use of || is for providing default values. For example, we might write something like this:
Defaulting
const lineWidth = settings.lineWidth || 80
This is a common idiom, and mostly works, but relies on JavaScript’s definition of “falsy”. What if 0 is a valid setting? Since 0 is falsy, we’ll end up with a line width of 80.
We could use the isNil function we just learned about above, but again Ramda has a nicer option for us: defaultTo.
Using defaultTo
const lineWidth = defaultTo(80, settings.lineWidth)
defaultTo checks if the second argument isNil. If it isn’t, it returns that as the value, otherwise it returns the first value.
Conditionals
Control flow is less necessary in functional programs, but still occasionally useful. The collection iteration functions we talked about in Part 1 take care of most looping situations, but conditionals are still quite important.
ifElse
Let’s write a function, forever21, that takes an age and returns the next age. But, as the name implies, once the age is 21, it stays that way.
Forever 21
const forever21 = age => age >= 21 ? 21 : age + 1
Notice that our conditional (age >= 21) and the second branch (age + 1) can both be written as functions of age. We could rewrite the first branch (21) as a constant function (() => 21) instead. Now we have three functions that take (or ignore) age.
We’re now in a position where we can use Ramda’s ifElse function, which is the function equivalent of the if...then...else structure or it’s shorter cousin, the ternary operator (?:).
Using ifElse
const forever21 = age => ifElse(gte(__, 21), () => 21, inc)(age)
As mentioned above, the comparison functions don’t work as we might like when combining functions, so we were forced to introduce the placeholder here (__). We could also switch over to lte instead:
Using lt instead of gte
const forever21 = age => ifElse(lte(21), () => 21, inc)(age)
In this case, we have to read this as “21 is less than or equal to age.” I’m going to stick with the placeholder version for the rest of this post because I find it more readable and less confusing.
Constants
Constant functions are quite useful in situations like this. As you might imagine, Ramda provides us a shortcut. In this case, the shortcut is named always.
Using always
const forever21 = age => ifElse(gte(__, 21), always(21), inc)(age)
Ramda also provides T and F as further shortcuts for always(true) and always(false).
Identity
Let’s try another function, alwaysDrivingAge. This function takes an age and returns it if it’s gte 16. But if it is less than 16, it returns 16. This allows anyone to pretend that they’re driving age, even if they’re not.
Driving Age
const alwaysDrivingAge = age => ifElse(lt(__, 16), always(16), a => a)(age)
The second branch of the conditional (a => a) is another common pattern in functional programming. It is known as the identity function. That is, a function that returns whatever argument it is given.
As you might suspect, Ramda provides an identity function for us.
Using identity
const alwaysDrivingAge = age => ifElse(lt(__, 16), always(16), identity)(age)
identity can take more than one argument, but it always returns its first argument. If we want to return something other than the first argument, there’s the more general nthArg function. It’s much less common than identity.
when and unless
Having an ifElse statement where one of the conditional branches is identity is also quite common, so Ramda provides more shortcuts for us.
If, as in our case, the second branch is identity, we can use when instead of ifElse:
Using when
const alwaysDrivingAge = age => when(lt(__, 16), always(16))(age)
If the first branch of the conditional is identity, we can use unless. If we reversed our condition to use gte(__, 16) instead, we could use unless.
Using unless
const alwaysDrivingAge = age => unless(gte(__, 16), always(16))(age)
cond
Ramda also provides the cond function which can replace a switch statement or a chain of if...then...else statements.
I’ll repeat the example from the Ramda documentation to show how it’s used:
Using cond
const water = temperature => cond([
[equals(0), always(‘water freezes at 0°C‘)],
[equals(100), always(‘water boils at 100°C‘)],
[T, temp => `nothing special happens at ${temp}°C`]
])(temperature)
I haven’t needed to use cond yet in my Ramda code, but I used to write Common Lisp code many years ago, so cond feels like an old friend.
Conclusion
We’ve now looked at a number of functions that Ramda gives us for turning our imperative code into declarative functional code.
Next
You may have noticed that the last few functions we wrote (forever21, drivingAge, and water) all take a parameter, build up a new function, and then apply that function to the parameter.
This is a common pattern, and once again Ramda provides the tools to clean this up. The next post, Pointfree Style looks at how to simplify functions that follow this pattern.