#aoc #parse #puzzle #parser #advent


A little library for parsing your Advent of Code puzzle input

17 releases

0.2.17 Dec 21, 2022
0.2.16 Dec 20, 2022
0.1.9 Dec 10, 2022

#217 in Rust patterns

Download history 17/week @ 2023-06-09 23/week @ 2023-06-16 28/week @ 2023-06-23 12/week @ 2023-06-30 6/week @ 2023-07-07 6/week @ 2023-07-14 10/week @ 2023-07-21 7/week @ 2023-07-28 24/week @ 2023-08-04 11/week @ 2023-08-11 23/week @ 2023-08-18 40/week @ 2023-08-25 24/week @ 2023-09-01 11/week @ 2023-09-08 10/week @ 2023-09-15

85 downloads per month

MIT license

2.5K SLoC


A parser library designed for Advent of Code.

This library mainly provides a macro, parser!, that lets you write a custom parser for your AoC puzzle input in seconds.

For example, my puzzle input for December 2, 2015 looked like this:


The parser for this format is a one-liner: parser!(lines(u64 "x" u64 "x" u64)).

How to use aoc-parse

It's pretty easy.

use aoc_parse::{parser, prelude::*};

let p = parser!(lines(u64 "x" u64 "x" u64));
    vec![(4, 23, 21), (22, 29, 19)]

If you're using aoc-runner, it might look like this:

use aoc_runner_derive::*;
use aoc_parse::{parser, prelude::*};

fn parse_input(text: &str) -> Vec<(u64, u64, u64)> {
    let p = parser!(lines(u64 "x" u64 "x" u64));


The argument you need to pass to the parser! macro is a pattern; all aoc-parse does is match strings against your chosen pattern and convert them into Rust values.

Here are some examples of patterns:

lines(i32)      // matches a list of integers, one per line
                // converts them to a Vec<i32>

line(lower+)    // matches a single line of one or more lowercase letters
                // converts them to a Vec<char>

lines({         // matches lines made up of the characters < = >
    "<" => -1,  // converts them to a Vec<Vec<i32>> filled with -1, 0, and 1
    "=" => 0,
    ">" => 1

Here are the pieces that you can use in a pattern:

  • i8, i16, i32, i64, i128, isize, big_int - These match an integer, written out using decimal digits, with an optional + or - sign at the start, like 0 or -11474.

    It's an error if the string contains a number too big to fit in the type you chose. For example, parser!(i8).parse("1000") is an error. (It matches the string, but fails during the "convert" phase.)

    big_int parses a num_bigint::BigInt.

  • u8, u16, u32, u64, u128, usize, big_uint - The same, but without the sign.

  • i8_bin, i16_bin, i32_bin, i64_bin, i128_bin, isize_bin, big_int_bin, u8_bin, u16_bin, u32_bin, u64_bin, u128_bin, usize_bin, big_uint_bin, i8_hex, i16_hex, i32_hex, i64_hex, i128_hex, isize_hex, big_int_hex, u8_hex, u16_hex, u32_hex, u64_hex, u128_hex, usize_hex, big_uint_hex - Match an integer in base 2 or base 16. The _hex parsers allow both uppercase and lowercase digits A-F.

  • bool - Matches either true or false and converts it to the corresponding bool value.

  • 'x' or "hello" - A Rust character or string, in quotes, is a pattern that matches that exact text only.

    Exact patterns don't produce a value.

  • pattern1 pattern2 pattern3... - Patterns can be concatenated to form larger patterns. This is how parser!(u64 "x" u64 "x" u64) matches the string 4x23x21. It simply matches each subpattern in order. It converts the match to a tuple if there are two or more subpatterns that produce values.

  • parser_var - You can use previously defined parsers that you've stored in local variables.

    For example, the amount parser below makes use of the fraction parser defined on the previous line.

    let fraction = parser!(i64 "/" u64);
    let amount = parser!(fraction " tsp");
    assert_eq!(amount.parse("1/4 tsp").unwrap(), (1, 4));

    An identifier can also refer to a string or character constant.

Repeating patterns:

  • pattern* - Any pattern followed by an asterisk matches that pattern zero or more times. It converts the results to a Vec. For example, parser!("A"*) matches the strings A, AA, AAAAAAAAAAAAAA, and so on, as well as the empty string.

  • pattern+ - Matches the pattern one or more times, producing a Vec. parser!("A"+) matches A, AA, etc., but not the empty string.

  • pattern? - Optional pattern, producing a Rust Option. For example, parser!("x=" i32?) matches x=123, producing Some(123); it also matches x=, producing the value None.

    These behave just like the *, +, and ? special characters in regular expressions.

  • repeat_sep(pattern, separator) - Match the given pattern any number of times, separated by the separator. This converts only the bits that match pattern to Rust values, producing a Vec. Any parts of the string matched by separator are not converted.

Matching single characters:

  • alpha, alnum, upper, lower - Match single characters of various categories. (These use the Unicode categories, even though Advent of Code historically sticks to ASCII.)

  • digit, digit_bin, digit_hex - Match a single ASCII character that's a digit in base 10, base 2, or base 16, respectively. The digit is converted to its numeric value, as a usize.

  • any_char - Match the next character, no matter what it is (like . in a regular expression, except that any_char matches newline characters too).

  • char_of(str) - Match the next character if it's one of the characters in str. For example, char_of(">^<v") matches exactly one character, either >, ^, <, or v. Returns the index of the character within the list of options (in this case, 0, 1, 2, or 3).

Matching multiple characters:

  • string(pattern) - Matches the given pattern, but instead of converting it to some value, simply return the matched characters as a String.

    By default, alpha+ returns a Vec<char>, and sometimes that is handy in AoC, but often it's better to have it return a String.

Custom conversion:

  • ... name1:pattern1 ... => expr - On successfully matching the patterns to the left of =>, evaluate the Rust expression expr to convert the results to a single Rust value.

    Use this to convert input to structs. For instance, suppose your puzzle input contains each elf's name and height:


    and you'd like to turn this into a vector of struct Elf values. The code you need is:

    struct Elf {
        name: String,
        height: u32,
    let p = parser!(lines(
        elf:string(alpha+) '=' ht:u32 => Elf { name: elf, height: ht }

    The name elf applies to the pattern string(alpha+) and the name ht applies to the pattern i32. The bit after the => is plain old Rust code.

    The names are in scope only for the following expr in the same set of matching parentheses or braces.


  • {pattern1, pattern2, ...} - Matches any one of the patterns. First try matching pattern1; if it matches, stop. If not, try pattern2, and so on. All the patterns must produce the same type of Rust value.

    This is sort of like a Rust match expression.

    For example, parser!({"<" => -1, ">" => 1}) either matches <, returning the value -1, or matches >, returing 1.

    Alternatives are handy when you want to convert the input into an enum. For example, my puzzle input for December 23, 2015 was a list of instructions that looked (in part) like this:

    jie a, +4
    tpl a
    inc a
    jmp +2
    hlf a
    jmp -7

    This can be easily parsed into a vector of beautiful enums, like so:

    enum Reg {
    enum Insn {
        Jie(Reg, isize),
        Jio(Reg, isize),
    use Reg::*;
    use Insn::*;
    let reg = parser!({"a" => A, "b" => B});
    let p = parser!(lines({
        "hlf " r:reg => Hlf(r),
        "tpl " r:reg => Tpl(r),
        "inc " r:reg => Inc(r),
        "jmp " offset:isize => Jmp(offset),
        "jie " r:reg ", " offset:isize => Jie(r, offset),
        "jio " r:reg ", " offset:isize => Jio(r, offset),

Rule sets:

  • rule name1: type1 = pattern1; - Introduce a "rule", a named subparser.

    This supports parsing text with nesting brackets.

    enum Formation {
    let p = parser!(
        // First rule: A "formation" has return type Formation and is either
        // a letter or a stack.
        rule formation: Formation = {
            s:alpha => Formation::Elf(s),
            v:stack => Formation::Stack(v),
        // Second rule: A "stack" is one or more formations, wrapped in
        // matching parentheses.
        rule stack: Vec<Formation> = '(' v:formation+ ')' => v;
        // After all rules, the pattern that .parse() will actually match.
    assert!(p.parse("x(fo))\n").is_err());  // parens not balanced

    Ordinarily let suffices for parsers used by other parsers; but rule is needed for parsers that refer to themselves or to each other, cyclically, like formation and stack above. Rust's let doesn't support that.

    Note: Left-recursive grammars don't work, as usual for PEG parsers.

Lines and sections:

  • line(pattern) - Matches a single line of text that matches pattern, and the newline at the end of the line.

    This is like ^pattern\n in regular expressions, with two minor differences:

    • line(pattern) will only ever match exactly one line of text, even if pattern could match more newlines.

    • If your input does not end with a newline, line(<var<pattern) can still match the non-newline-terminated "line" at the end.

    line(string(any_char+)) matches a line of text, strips off the newline character, and returns the rest as a String.

    line("") matches a blank line.

  • lines(pattern) - Matches any number of lines of text matching pattern. Each line must be terminated by a newline, '\n'.

    Equivalent to line(pattern)*.

    let p = parser!(lines(repeat_sep(digit, " ")));
        p.parse("1 2 3\n4 5 6\n").unwrap(),
        vec![vec![1, 2, 3], vec![4, 5, 6]],
  • section(pattern) - Matches zero or more nonblank lines, followed by either a blank line or the end of input. The nonblank lines must match pattern.

    section() consumes the blank line. pattern should not expect to see it.

    It's common for an AoC puzzle input to have several lines of data, then a blank line, and then a different kind of data. You can parse this with section(p1) section(p2).

    section(lines(u64)) matches a section that's a list of numbers, one per line.

  • sections(pattern) - Matches any number of sections matching pattern. Equivalent to section(pattern)*

    Bringing it all together to parse a complex example:

    let example = "\
    Wiring Diagram #1:
    Wiring Diagram #2:
    let p = parser!(sections(
        line("Wiring Diagram #" usize ":")
        lines(repeat_sep(alpha, "->"))
            (1, vec![vec!['a', 'q', 'E', 'z', 'J'], vec!['D', 'f', 'D']]),
            (2, vec![vec!['g', 'r', 'f'], vec!['g', 'B']]),

License: MIT


~85K SLoC