Developer guide

Getting started

You will first need:

One-time setup:

  1. Download the repository: git clone

  2. Grab the minimal amount of data to get started: cargo run --bin updater

  3. Run the game: RUST_BACKTRACE=1 cargo run --bin game --release. On Windows, set environment variables like this: set RUST_BACKTRACE=1 && cargo run --bin game --release

Development tips

  • Generated API documentation
  • Compile faster by just doing cargo run. The executable will have debug stack traces and run more slowly. You can do cargo run --release to build in optimized release mode; compilation will be slower, but the executable much faster.
  • Some in-game features are turned off by default or don't have a normal menu to access them. The list:
    • To toggle developer mode: press Control+S in game, or cargo run -- --dev
    • To warp to an object by numeric ID: press Control+j
    • To enter debug mode with all sorts of goodies: press Control+D
  • You can start the game in different modes using flags:
    • cargo run --bin game -- --dev data/system/us/seattle/maps/downtown.bin starts on a particular map
    • cargo run --bin game -- data/system/us/seattle/scenarios/downtown/weekday.bin starts with a scenario (which is tied to a certain map)
    • cargo run --bin game -- --challenge=trafficsig/tut2 starts on a particular challenge. See the list of aliases by passing in a bad value here.
    • cargo run --bin game -- data/player/saves/us/seattle/montlake/no_edits_unnamed/00h00m20.3s.bin restores an exact simulation state. Savestates are found in debug mode (Control+D) -- they're probably confusing for the normal player experience, so they're hidden for now.
    • cargo run --bin game -- --tutorial=12 starts somewhere in the tutorial
    • Adding --edits='name of edits' starts with edits applied to the map.

Downloading more cities

As data formats change over time, things in the data/ directory not under version control will get out of date. At any time, you can run cargo run --bin updater from the main repository directory to update only the files that have changed.

You can also opt into downloading updates for more cities by editing data/player/data.json. In the main UI, there's a button to download more cities that will help you manage this config file.

If you want to opt into absolutely everything: cargo run --bin updater -- --opt-into-all > data/player/data.json

Building map data

You can skip this section if you're just touching code in game, widgetry, and sim.

To run all pieces of the importer, you'll need some extra dependencies:

The first stage of the importer, --raw, will download input files from OSM, King County GIS, and so on. If the mirrors are slow or the files vanish, you could fill out data/config and use the updater described above to grab the latest input.

Building contraction hierarchies for pathfinding occurs in the --map stage. It can take a few minutes for larger maps. To view occasional progress updates, you can run the importer with

RUST_LOG="fast_paths=debug/contracted node [0-9]+0000 "

You can rerun specific stages of the importer:

  • If you're modifying the initial OSM data -> RawMap conversion in convert_osm, you need ./ --raw --map.
  • If you're modifying map_model but not the OSM -> RawMap conversion, then you just need ./ --map.
  • If you're modifying the demand model for Seattle, you can add --scenario to regenerate.
  • By default, all maps are regenerated. You can also specify a single map: ./ --map downtown.
  • By default, Seattle is assumed as the city. You have to specify otherwise: ./ --city=us/detroit --map downtown.

You can also make the importer import a new city.

Understanding stuff

The docs listed at explain things like map importing and how the traffic simulation works.

Code organization

If you're going to dig into the code, it helps to know what all the crates are. The most interesting crates are map_model, sim, and game.

Constructing the map:

  • convert_osm: extract useful data from OpenStreetMap and other data sources, emit intermediate map format
  • kml: extract shapes from KML and CSV shapefiles
  • map_model: the final representation of the map, also conversion from the intermediate map format into the final format
  • map_editor: GUI for modifying geometry of maps and creating maps from scratch. pretty abandoned as of June 2020
  • importer: tool to run the entire import pipeline
  • updater: tool to download/upload large files used in the import pipeline

Traffic simulation:

  • sim: all of the agent-based simulation logic
  • headless: tool to run a simulation without any visualization


  • game: the GUI and main gameplay
  • map_gui: common code to interact with map_model maps
  • widgetry: a GUI and 2D OpenGL rendering library, using glium + winit + glutin

Common utilities:

  • abstutil: a grab-bag timing and logging utilities
  • abstio: Reading/writing files on native/web
  • geom: types for GPS and map-space points, lines, angles, polylines, polygons, circles, durations, speeds


  • collisions: an experimental data format for real-world collision data
  • traffic_seitan: a bug-finding tool that randomly generates live map edits
  • tests: integration tests
  • santa: 15-minute Santa, an arcade game about delivering and zoning
  • parking_mapper: a standalone tool to help map street parking in OSM
  • osm_viewer: a standalone tool to render OSM in detail
  • fifteen_min: a standalone tool to explore 15-minute neighborhoods
  • popdat: use census data to produce traffic simulation input
  • traffic_signal_data: manual timing overrides for some traffic signals
  • sumo: interoperability with SUMO

Code conventions

All code is automatically formatted using; please run cargo +nightly fmt before sending a PR. (You have to install the nightly toolchain just for fmt)

cargo fmt can't yet organize imports, but we follow a convention to minimize conflict with what some IDEs do. Follow existing code to group imports: std, external crates, other crates in the project, the current crate, then finally any module declarations.

See the testing strategy page.

Error handling

The error handling is unfortunately inconsistent. The goal is to gracefully degrade instead of crashing the game. If a crash does happen, make sure the logs will have enough context to reproduce and debug. For example, giving up when some geometry problem happens isn't ideal, but at least make sure to print the road / agent IDs or whatever will help find the problem. It's fine to crash during map importing, since the player won't deal with this, and loudly stopping problems is useful. It's also fine to crash when initially constructing all of the renderable map objects, because this crash will consistently happen at startup-time and be noticed by somebody developing before a player gets to it.

Since almost none of the code ever needs to distinguish error cases, use anyhow. Most of the errors generated within A/B Street are just strings anyway; the bail! macro is a convenient way to return them.


Prefer using info!, warn!, error!, etc from the log crate rather than println.

Adjust the log level without recompiling via the RUST_LOG env variable.

RUST_LOG=debug cargo run --bin game

This can be done on a per lib basis:

RUST_LOG=my_lib=debug cargo run --bin game

Or a module-by-module basis:

RUST_LOG=my_lib::module=debug cargo run --bin game

You can mix and match:

# error logging by default, except the foo:bar module at debug level
# and the entire baz crate at info level
RUST_LOG=error,foo::bar=debug,baz=info cargo run --bin game

For some special cases, you might want to use regex matching by specifying a pattern with the "/":

# only log once every 10k
RUST_LOG="fast_paths=debug/contracted node [0-9]+0000 " mike import_la

See the env_logger documentation for more usage examples.


Use, just running it on the binaries you build normally.