#advent #challenge #framework #problem #advent-of-code #pretty-print

lib_aoc

A simple trait-based framework for the annual Advent of Code programming challenge

7 unstable releases

0.8.0 Dec 1, 2023
0.7.2 Dec 13, 2022
0.6.2 Dec 8, 2022
0.5.0 Dec 6, 2022

#54 in Value formatting

MIT/Apache

30KB
315 lines

lib_aoc

lib_aoc is a simple trait-based framework for the annual Advent of Code programming challenge.

Focus less on the boilerplate and more on the problem by automatically wiring up your solutions with input loading, pretty-printing and granular benchmarking.

Getting Started

Create a new binary crate and add lib_aoc as a dependency.

$ cargo new advent_of_code && cd advent_of_code
$ cargo add lib_aoc

Then, import the lib_aoc prelude and create a new struct to link your solutions.

use lib_aoc::prelude::*;

// Can be named whatever you'd like.
struct Solutions {}

fn main() { /* ... */ }

When solving a problem, you'll implement the Solution trait on this struct, and lib_aoc will take care of connecting everything together.

Before you can do that, however, you'll need to implement the Solver trait on the struct, which (among other, optional things) tells lib_aoc how you'd like puzzle inputs to be loaded.

The simple approach is to just read the input from disk, but more complex approaches (such as scraping the Advent of Code website directly) are certainly possible.

impl Solver for Solutions {
    fn load(day: u8) -> String {
        std::fs::read_to_string(format!("src/inputs/{day:02}.txt"))
            .expect("Puzzle input could not be read.")
    }
    
    // Note that a test loading implementation can be elided if one is not desired;
    // the default implementation will simply panic.
    fn load_test(day: u8, part: bool) -> String {
        std::fs::read_to_string(format!("src/inputs/test_{day:02}.txt"))
            .expect("Puzzle input could not be read.")
    }
}

With Solver implemented, you can now begin solving problems!

Implementing a Solution

For demonstration purposes, we'll assume a simple first problem:

  • The input is a list of integers, one per line.
  • Part one wants the sum of all the integers.
  • Part two wants us to square each integer, then sum them.

Start by implementing Solution<DAY_01> for your solutions struct; at minimum, you need to provide type definitions for Input and Output, as well as an implementation of parse.

impl Solution<DAY_01> for Solutions {
    type Input<'i> = Vec<u64>;
    type Output = u64;

    fn parse(puzzle: &str) -> Self::Input<'_> {
        puzzle
            .lines()
            .map(str::parse::<u64>())
            .map(Result::unwrap)
            .collect::<Vec<_>>() 
    }
}

At this point, the solution is technically ready to be run. You can use the solve_through macro to execute all solutions up to a certain day, like so:

fn main() {
    // Notes: 
    // - Due to macro limitations, you must use an integer literal for the day cap.
    // - Trying to solve through days you haven't implemented yet is a compile error.
    solve_through!(Solutions, 1);
}

Assuming your load implementation works, the program should output something like this:

--- DAY 1 ---
Part 1: unimplemented
Part 2: unimplemented

--- BENCH (RELEASE) ---
Parsing: 20 ns
Part 1: 20 ns
Part 2: 20 ns
Total: 60 ns

It looks like the actual solution logic is unimplemented! Fortunately, that's easy to fix - we just implement the part_one and part_two methods.

impl Solution<DAY_01> for Solutions {
    type Input<'i> = Vec<u64>;
    type Output = u64;

    fn parse(puzzle: &str) -> Self::Input<'_> {
        puzzle
            .lines()
            .map(str::parse::<u64>())
            .map(Result::unwrap)
            .collect::<Vec<_>>() 
    }

    fn part_one(input: &Self::Input<'_>) -> Self::Output {
        input.iter()
            .sum::<u64>()
            .into()
    }

    fn part_two(input: &Self::Input<'_>) -> Self::Output {
        input.iter()
            .map(|x| x.pow(2) )
            .sum::<u64>()
            .into()
    }
}

As you can see, the signatures of the solver methods are identical apart from their names - they take a shared reference to a value of type Input and return an Output.

The default implementations of these methods panic, which (by using std::panic::catch_unwind) is how lib_aoc knew to display unimplemented when the program was run earlier. By overriding them with implementations that don't panic and instead return a proper value, the result will be displayed instead:

--- DAY 1 ---
Part 1: 2506
Part 2: 95843

--- BENCH (RELEASE) ---
Parsing: 7.223 µs
Part 1: 73.838 µs
Part 2: 81.042 µs
Total: 162.244 µs

And that's it - you've implemented a solution!

Deriving Tests

Because Advent of Code provides a test case in the description of every problem, lib_aoc also comes with a macro for deriving tests from your Solution implementations.

Assuming your load_test implementation already correctly loads test cases, all you need to do is implement the Test trait on your solution to provide the expected results:

impl Test<DAY_01> for Solutions {
    fn expected(part: bool) -> Self::Output {
        // If you don't know the expected result for a part yet, you can just
        // substitute a panicking macro.
        match part {
            // PART_ONE and PART_TWO are constants from the prelude.
            PART_ONE => 24_000,
            PART_TWO => 34_000
        }
    }
} 

Then you can invoke the derive_tests macro to auto-generate the tests:

derive_tests!(Solutions, DAY_01);

This expands into a new module with a test function for each part of the solution, and can be run normally via cargo test.

Notes on Benchmarking

lib_aoc provides basic benchmarking of solution implementations via std::time::Instant. While the measurements it provides are good approximations, a crate like criterion is a better choice if you want a more rigorous solution.

Also note that execution clock is started after your Solver::load implementation returns, immediately before Solution::parse is invoked. This means the time spent loading the puzzle input is not considered by the benchmark.

Additional Customization Options

By overriding the Solver::display and Solver::finalize methods, it's possible to define custom behavior that is invoked once a solution finishes executing in a non-test context.

display has a default implementation that pretty-prints the solution outcome, while finalize defaults to a no-op. Both methods take a shared reference to an Outcome<impl Display>.

Want to add some awesome extra behavior like submitting your solution to AoC right from the command line? You can do that here!

Dependencies

~0–8.5MB
~51K SLoC