Spanda

Lesson 4 — Types, units, and errors

Goal: Model success and failure with Result and Option instead of exceptions.

Example: examples/basics/04_result_and_option.sd

---

Result for fallible operations

Navigation, planning, and hardware calls can fail. Spanda uses Result<T, E>:

enum NavError {
  Blocked,
  Timeout
}

export fn plan_route() -> Result<Path, NavError> {
  return Err(Blocked);
}

Handle outcomes with match:

match plan_route() {
  Ok => wheels.drive(linear: 0.2 m/s, angular: 0.0 rad/s);
  Err => wheels.stop();
};

No exceptions — the type system forces you to handle both paths.

---

Option for missing values

When a value may not exist (optional sensor reading, cached map tile):

export fn latest_scan() -> Option<Scan> {
  return None();
}

match latest_scan() {
  Some => wheels.stop();
  None => wheels.stop();
};

Use Option<T> when absence is normal, Result<T, E> when failure carries a reason.

---

Structs and typed fields

Group related data:

struct NavGoal {
  x: Distance;
  y: Distance;
  heading: Angle;
}

let goal = NavGoal { x: 1.0 m, y: 0.0 m, heading: 0.0 rad };

Built-in aliases like Distance, Angle, and Path keep robotics code readable.

---

Serialization (preview)

For telemetry and IPC, serialize values to JSON or YAML:

let data = serialize(pose, "json");
let restored = deserialize(data, "json");

Full walkthrough: examples/basics/06_serialize_telemetry.sd and spanda-language.md.

---

Try it

spanda check examples/basics/04_result_and_option.sd
spanda run examples/basics/04_result_and_option.sd

---

Exercise

In a new module function check_clearance() -> Result<Path, NavError>:

1. Return Err(Blocked) when a lidar read would be below 0.5 m (use a stub condition for now)
2. Return Ok(...) with a dummy path otherwise
3. In your robot behavior, match the result and stop on Err

---

Next: Lesson 5 — Modules and traits

This site is open source. Improve this page.