10 releases

0.2.5 Jan 15, 2022
0.2.4 Jan 9, 2022
0.2.1 Oct 26, 2021
0.1.4 Oct 24, 2021

#271 in Parser implementations

MIT license

1.5K SLoC

Shalrath   License Latest Version Documentation

A rusty, spiky, heat-seeking quake map parser

shalrath is a rust representation, [nom] parser and string serializer for Quake map files.

It's written in pure Rust, and enforces the use of safe code crate-wide via #![forbid(unsafe_code)].

Rust Representation

The Rust representation lives in the [repr] module, and is a set of structs that represent the contents of a map file.

The overall class structure - with some of the more specific innermost types omitted for simplicity - looks something like this:

└ Entity (1..*)
  ├ Properties (1..1)
  │ └ Property (0..*)
  └ Brushes (1..1)
    └ Brush (0..*)
      └ BrushPlane (4..*)

[Entity] is a game object that can contain [Property]s and [Brush]es.

[Property]s are key-value pairs stored as [String]s.

[Brush]es are convex shapes defined by the intersection of a set of [TexturePlane]s - 3D planes with associated texture mapping data.

At least one [Entity] - known as the worldspawn - must exist in any given map, and represents all of its structural [Brush]es. Structural [Brush]es are static geometry with no associated behavior.

In Quake terms, entities are given behavior by assigning them a classname property, which is used by the game code to assign an actor class that reads from other properties attached to the object.

These entities are separated into two categories:

  • Point Entities are [Entity]s that have a classname, but no [Brush]es.
    • These are used to represent actors like the player, enemies, or item pickups.
  • Brush Entities are [Entity]s that have both a classname and [Brush]es.
    • These are used to represent special world geometry like moving doors, elevators, and similar.

But, that's just for context, more of which can be found in the map file spec.

Ultimately, what you do with the Rust representation after parsing data into it is down to needs of your project. To that end, struct members are public in the case of named fields, and exposed via [Deref] for collection wrappers like [Properties] and [Brushes].


The simplest way to parse a map file into AST is by way of the [FromStr] trait:

use shalrath::repr::*;

let map =
    "{\"classname\" \"worldspawn\"\n{\n( 0 1 2 ) ( 3 4 5 ) ( 6 7 8 ) TEXTURE 0 0 0 1 1\n}\n}"
        .expect("Failed to parse map");

    Map::new(vec![Entity {
        properties: Properties::new(vec![Property {
            key: "classname".into(),
            value: "worldspawn".into()
        brushes: Brushes::new(vec![Brush::new(vec![
            BrushPlane {
                plane: Triangle {
                    v0: Point {
                        x: 0.0,
                        y: 1.0,
                        z: 2.0
                    v1: Point {
                        x: 3.0,
                        y: 4.0,
                        z: 5.0
                    v2: Point {
                        x: 6.0,
                        y: 7.0,
                        z: 8.0
                texture: "TEXTURE".into(),
                texture_offset: TextureOffset::Standard { u: 0.0, v: 0.0 },
                angle: 0.0,
                scale_x: 1.0,
                scale_y: 1.0,
                extension: Extension::Standard,

For a lower-level alternative, the [parser] module contains the [nom] functions used by the [FromStr] implementations, which can be used to parse plaintext data into individual AST structs.

Of these, [parse_map] is the primary entrypoint, and is equivalent to str::parse::<Map>():

use shalrath::parser::repr::parse_map;

let map_string = include_str!("../test_data/abstract-test.map");
let (_, map_ast) = parse_map(map_string).expect("Failed to parse map");
println!("{:#?}", map_ast);

String Serialization

The Rust representation can be serialized back into a text-based map representation via the [Display] or [ToString] traits:

use shalrath::repr::Map;

let map_string = include_str!("../test_data/abstract-test.map");
let map_ast = map_string.parse::<Map>().expect("Failed to parse map file");
let serialized_map_string = map_ast.to_string();
println!("{}", serialized_map_string);

In addition, round-trip parsing the resulting string back into the corresponding AST is a lossless operation, and is included as a standard part of shalrath's integration tests:

use shalrath::repr::Map;

let map_string = include_str!("../test_data/abstract-test.map");
let map_ast = map_string.parse::<Map>().expect("Failed to parse map file");
let serialized_map_string = map_ast.to_string();
let roundtrip_map_ast = serialized_map_string.parse::<Map>().expect("Failed to parse map file");
assert_eq!(map_ast, roundtrip_map_ast);

Format Support

Several variants of the base Quake 1 map format exist that retain the same core structure, but modify how brush planes are encoded.

shalrath supports these by categorizing them by UV format (represented by the [TextureOffset] enum):

UV Format Notes
Standard Faces project textures based on the closest world X/Y/Z plane.
Valve Faces project textures based on custom U/V axes, allowing for skewing and more accurate texturing of curved surfaces.

...and brush plane extension data, represented by the [Extension] enum:

Brush Plane Extension Notes
Standard Brush planes contain no extra data.
Hexen 2 Brush planes contain an extra numerical value whose usage is unknown.
Quake 2 Brush planes contain content_flags and surface_flags bitmasks, and a floating point value.
Daikatana Brush planes contain three unknown values, and floating point RGB values

Other formats like Quake 3 and Daikatana exist, but are effectively variants of the above, and will be handled transparently by the parser.

Serde Support

For cases where serializing and deserializing from non-map formats is required, shalrath includes [serde::Serialize] and [serde::Deserialize] derives for all types in the [repr] module.

These can be enabled by applying the serde feature flag to the shalrath dependency in Cargo.toml.


Currently shalrath only implements complete parsers that expect a full set of input data.

streaming implementations are planned, but currently pending further research.


~48K SLoC