Day 2

Alright, today is day 2 of advent of code, we have a challenge slightly more difficult than day 1 but still very much accessible and fun (not overwhelming yet 😬).

Looking back on Day 1

Before we jump into day 2's puzzle and my proposed solution in Kotlin, I would like to share something I learned about yesterday:
in Kotlin List API there is a windowed method which allows iterating over a list with "sliding windows".
This method takes a size parameter which is the size of the "sliding window".

To give a concrete example, consider the following code:

listOf(1, 2, 3, 4).windowed(2) // returns [[1, 2], [2, 3], [3, 4]]

Also, it can be called with a "transform" function parameter which will be applied to each "sliding window" or sublist.
For example:

listOf(1, 2, 3, 4).windowed(2) { it.sum() }) // returns [3, 5, 7]

I imagine you could remember how this could have been useful for yesterday's problem 😅

Big thanks to Sebastian Aigner for sharing his solution for Day 1 and teaching me about this!
You can find an interesting blog post he wrote about advanced Kotlin Collection functionality.

Day 2 Problem

You can find the description for this puzzle here, I would summarize the problem as the following:

Given a list of commands, compute some aggregates based on the command's value.

Modeling the input

I have decided to use a data class to model the input, as well as an enum for the command type:

enum class Type {
    FORWARD, UP, DOWN
}

data class Command(val type: Type, val amount: Int)

The input to my functions will be a List<Command>.

Solution

The distinction between Part 1 and Part 2 is how each command affects the aggregate to be calculated.
I have taken two different approach in each part on how to compute these aggregates, which I will share with you below:

Part 1

For this part, there are two aggregates depth and horizontalPosition. To calculate each we need to iterate over the input list. We could either calculate both in the same function and loop, or create a function for each aggregate.
In Part 2 I will be doing the former, in Part 1 I am doing the latter.
Below are the two functions I have used:

fun horizontalPosition(commands: List<Command>) =
    commands
        .filter { it.type == Type.FORWARD }
        .map { it.amount }
        .sum()

fun depth(commands: List<Command>) =
    commands
        .filter { it.type == Type.DOWN || it.type == Type.UP }
        .map { if (it.type == Type.DOWN) it.amount else -it.amount }
        .sum()

These follow the pattern filter-map-reduce. It is the clean functional way of transforming a list into an aggregate.

• For the horizontalPosition we simply need to sum the commands of type FORWARD (while ignoring other types of commands).
• For the depth, we need to add the amount for DOWN commands and subtract it for UP commands (while ignoring FORWARD commands).

Part 2

For this part, there is an additional aggregate called aim which we need to compute depth.
In other words, the logic for calculating depth changes, so we need to ditch our previous function.
This time, we have to compute the aggregates together as they depend on each other (depth depends on aim).
To do this, we can leverage the fold method on the List API which is a generalized way of doing a reduce as instead of using the first element of the list as a starting point for the calculation, we can specify our own initial element (and its type).

I have again leveraged a data class here for my aggregates:

data class Attributes(val aim: Int, val horizontalPosition: Int, val depth: Int)

I will use this class for my accumulator when using fold:

fun attributes(commands: List<Command>): Attributes {
    return commands
        .fold(
            Attributes(0, 0, 0),
            { acc, command -> operation(acc, command) }
        )
}

private fun operation(attributes: Attributes, command: Command): Attributes {
    val (aim, horizontalPosition, depth) = attributes
    return when (command.type) {
        Type.FORWARD -> Attributes(aim, horizontalPosition + command.amount, depth + (aim * command.amount))
        Type.UP -> Attributes(aim - command.amount, horizontalPosition, depth)
        Type.DOWN -> Attributes(aim + command.amount, horizontalPosition, depth)
    }
}

Here my initial value for the acc variable is Attributes(0, 0, 0).
Then, I loop over the list and for each command, I transform my acc variable by calling the operation method.
The fold can be considered as doing the following:

fun fold(commands: List<Command>) {
    var acc = Attributes(0, 0, 0)
    for (command in commands) {
        acc = operation(acc, command)
    }
    return acc
}

In other words, the operation method is where the computation happens, and if you look more closely you can see I am using the when expression.
This can be considered an improved "switch" and allows partial "pattern matching".

Based on the command type, I create a new Attributes instance using the previous one, and the command amount.
Note here than I do not need an else clause in my when expression. This is because I am using an enum and not a string, hence the Kotlin compilers knows that I have all the cases covered (as my enum has only 3 fields).
This is very nice as it avoids having something like:

when (foo) { 
  // [...] 
  else -> throw Exception("this should never be reached")
}

Also note how data classes allow destructuring of properties:

val (aim, horizontalPosition, depth) = attributes

is equivalent to:

val aim = attributes.aim
val horizontalPosition = attributes.horizontalPosition
val depth = attributes.depth

Kotlin Concepts

Alright, I hope you liked my proposed solution and learned a few things along the way.
The key Kotlin concepts that we have covered are:

• filter-map-reduce for computing an aggregate from a list
• fold for computing multiple aggregates at once from a list
• when expression and enums work nicely hand in hand
• data class can easily help to build named tuples of data

Source code

You can find the source code for my solution here.

Thank you very much for reading!