@domir/bun-test

Helpers for testing Effect-based code with bun test. Provides an enhanced it function with support for scoped tests, test services such as TestClock, shared layers, and property testing. The API mirrors @effect/vitest as closely as bun's runner allows.

Installation

Ensure a current bun release is installed (tested with 1.3.14), then add the package as a dev dependency:

bun add -D @domir/bun-test@rc

Documentation

• Effect website
• @effect/vitest API reference (same API, minus the vitest test context)

Overview

The main entry point is the following import:

import { it } from '@domir/bun-test';

This import enhances the standard it function from bun:test with several powerful features, including:

Feature	Description
it.effect	Runs a scoped test with test services such as TestClock and TestConsole.
it.live	Runs a scoped test with the live Effect environment.
it.layer	Shares a Layer between multiple tests.
it.prop	Runs property tests using Effect Schema values or Arbitrary values.
it.flakyTest	Retries an Effect that might occasionally fail until it succeeds or reaches the configured timeout.

Writing Tests with it.effect

Here's how to use it.effect to write your tests:

Syntax

import { it } from "@domir/bun-test"

it.effect("test name", () => EffectContainingAssertions, timeout: number | TestOptions = 5_000)

it.effect automatically provides the Effect test services, including TestClock, and a fresh Scope for each test. The scope is closed when the test finishes.

Testing Successful Operations

To write a test, place your assertions directly within the main effect. This ensures that your assertions are evaluated as part of the test's execution.

Example (Testing a Successful Operation)

In the following example, we test a function that divides two numbers, but fails if the divisor is zero. The goal is to check that the function returns the correct result when given valid input.

import { expect, it } from '@domir/bun-test';
import { Effect } from 'effect';

// A simple divide function that returns an Effect, failing when dividing by zero
function divide(a: number, b: number) {
	if (b === 0) return Effect.fail('Cannot divide by zero');
	return Effect.succeed(a / b);
}

// Testing a successful division
it.effect('test success', () =>
	Effect.gen(function* () {
		const result = yield* divide(4, 2); // Expect 4 divided by 2 to succeed
		expect(result).toBe(2); // Assert that the result is 2
	})
);

Testing Successes and Failures as Exit

When you need to handle both success and failure cases in a test, you can use Effect.exit to capture the outcome as an Exit object. This allows you to verify both successful and failed results within the same test structure.

Example (Testing Success and Failure with Exit)

import { expect, it } from '@domir/bun-test';
import { Effect, Exit } from 'effect';

// A function that divides two numbers and returns an Effect.
// It fails if the divisor is zero.
function divide(a: number, b: number) {
	if (b === 0) return Effect.fail('Cannot divide by zero');
	return Effect.succeed(a / b);
}

// Test case for a successful division, using `Effect.exit` to capture the result
it.effect('test success as Exit', () =>
	Effect.gen(function* () {
		const result = yield* Effect.exit(divide(4, 2)); // Capture the result as an Exit
		expect(result).toStrictEqual(Exit.succeed(2)); // Expect success with the value 2
	})
);

// Test case for a failure (division by zero), using `Effect.exit`
it.effect('test failure as Exit', () =>
	Effect.gen(function* () {
		const result = yield* Effect.exit(divide(4, 0)); // Capture the result as an Exit
		expect(result).toStrictEqual(Exit.fail('Cannot divide by zero')); // Expect failure with the correct message
	})
);

Using the TestClock

When writing tests with it.effect, Effect test services are automatically provided. These include the TestClock, which allows you to simulate the passage of time in your tests.

Note: If you want to use the real-time clock (instead of the simulated one), you can switch to it.live.

Example (Using TestClock and it.live)

Here are examples that demonstrate how you can work with time in your tests using it.effect and TestClock:

1. Using it.live to show the current time: This will display the actual system time, since it runs in the live environment.

2. Using it.effect without adjustments: By default, the TestClock starts at 0, simulating the beginning of time for your test without any time passing.

3. Using it.effect and adjusting time: In this test, we simulate the passage of time by advancing the clock by 1000 milliseconds (1 second).

import { it } from '@domir/bun-test';
import { Clock, Effect } from 'effect';
import { TestClock } from 'effect/testing';

// Effect to log the current time
const logNow = Effect.gen(function* () {
	const now = yield* Clock.currentTimeMillis; // Fetch the current time from the clock
	console.log(now); // Log the current time
});

// Example of using the real system clock with `it.live`
it.live('runs the test with the live Effect environment', () =>
	Effect.gen(function* () {
		yield* logNow; // Prints the actual current time
	})
);

// Example of using `it.effect` with the default test environment
it.effect('run the test with the test environment', () =>
	Effect.gen(function* () {
		yield* logNow; // Prints 0, as the test clock starts at 0
	})
);

// Example of advancing the test clock by 1000 milliseconds
it.effect('run the test with the test environment and the time adjusted', () =>
	Effect.gen(function* () {
		yield* TestClock.adjust('1000 millis'); // Move the clock forward by 1000 milliseconds
		yield* logNow; // Prints 1000, reflecting the adjusted time
	})
);

Skipping Tests

If you need to temporarily disable a test but don't want to delete or comment out the code, you can use it.effect.skip. This is helpful when you're working on other parts of your test suite but want to keep the test for future execution.

Example (Skipping a Test)

import { expect, it } from '@domir/bun-test';
import { Effect, Exit } from 'effect';

function divide(a: number, b: number) {
	if (b === 0) return Effect.fail('Cannot divide by zero');
	return Effect.succeed(a / b);
}

// Temporarily skip the test for dividing numbers
it.effect.skip('test failure as Exit', () =>
	Effect.gen(function* () {
		const result = yield* Effect.exit(divide(4, 0));
		expect(result).toStrictEqual(Exit.fail('Cannot divide by zero'));
	})
);

Running a Single Test

When you're developing or debugging, it's often useful to run a specific test without executing the entire test suite. You can achieve this by using it.effect.only, which will run just the selected test and ignore the others.

Example (Running a Single Test)

import { expect, it } from '@domir/bun-test';
import { Effect, Exit } from 'effect';

function divide(a: number, b: number) {
	if (b === 0) return Effect.fail('Cannot divide by zero');
	return Effect.succeed(a / b);
}

// Run only this test, skipping all others
it.effect.only('test failure as Exit', () =>
	Effect.gen(function* () {
		const result = yield* Effect.exit(divide(4, 0));
		expect(result).toStrictEqual(Exit.fail('Cannot divide by zero'));
	})
);

Expecting Tests to Fail

When adding new failing tests, you might not be able to fix them right away. Instead of skipping them, you may want to assert it fails, so that when you fix them, you'll know and can re-enable them before it regresses.

Example (Asserting one test fails)

import { expect, it } from '@domir/bun-test';
import { Effect, Exit } from 'effect';

function divide(a: number, b: number) {
	if (b === 0) return Effect.fail('Cannot divide by zero');
	return Effect.succeed(a / b);
}

// Temporarily assert that the test for dividing by zero fails.
it.effect.fails('dividing by zero special cases', () =>
	Effect.gen(function* () {
		const result = yield* Effect.exit(divide(4, 0));
		expect(result).toStrictEqual(0);
	})
);

Logging

By default, it.effect suppresses log output, which can be useful for keeping test results clean. However, if you want to enable logging during tests, you can use it.live or provide a custom logger to control the output.

Example (Controlling Logging in Tests)

import { it } from '@domir/bun-test';
import { Effect, Logger } from 'effect';

// This test won't display the log message, as logging is suppressed by default in `it.effect`
it.effect('does not display a log', () =>
	Effect.gen(function* () {
		yield* Effect.log('it.effect'); // Log won't be shown
	})
);

// This test will display the log because a custom logger is provided
it.effect('providing a logger displays a log', () =>
	Effect.gen(function* () {
		yield* Effect.log('it.effect with custom logger'); // Log will be displayed
	}).pipe(
		Effect.provide(Logger.layer([Logger.consolePretty()])) // Providing a pretty logger for log output
	)
);

// This test runs using `it.live`, which enables logging by default
it.live('it.live displays a log', () =>
	Effect.gen(function* () {
		yield* Effect.log('it.live'); // Log will be displayed
	})
);

Resource Safety and Scope

Both it.effect and it.live provide a fresh Scope and close it after each test. Test bodies can therefore use scoped resources directly. Do not wrap the test body in Effect.scoped, because the test runner already manages its scope.

Example (Managing a Resource Lifecycle)

import { it } from '@domir/bun-test';
import { Console, Effect } from 'effect';

// Simulating the acquisition and release of a resource with console logging
const acquire = Console.log('acquire resource');
const release = Console.log('release resource');

// Defining a resource that requires proper management
const resource = Effect.acquireRelease(acquire, () => release);

it.effect('run with scope', () =>
	Effect.gen(function* () {
		yield* resource;
	})
);

Writing Tests with it.flakyTest

it.flakyTest is a utility designed to manage tests that may not succeed consistently on the first attempt. These tests, often referred to as "flaky," can fail due to factors like timing issues, external dependencies, or randomness. it.flakyTest allows for retrying these tests until they pass or a specified timeout is reached.

Example (Handling Flaky Tests with Retries)

Let's start by setting up a basic test scenario that has the potential to fail randomly:

import { it } from '@domir/bun-test';
import { Effect, Random } from 'effect';

// Simulating a flaky effect
const flaky = Effect.gen(function* () {
	const random = yield* Random.nextBoolean;
	if (random) {
		return yield* Effect.fail('Failed due to randomness');
	}
});

// Standard test that may fail intermittently
it.effect('possibly failing test', () => flaky);

In this test, the outcome is random, so the test might fail depending on the result of Random.nextBoolean.

To handle this flakiness, we use it.flakyTest to retry the test until it passes, or until a defined timeout expires:

// Retrying the flaky test with a 5-second timeout
it.effect('retrying until success or timeout', () => it.flakyTest(flaky, '5 seconds'));

Limitations

Bun's test runner does not pass an AbortSignal into the test function, only an optional done callback, so @domir/bun-test has no way to interrupt a running fiber when bun's own per-test timeout fires. If a test needs a guaranteed interruption after some duration, use Effect.timeout (or Effect.race against a delay) inside the effect itself rather than relying on the runner's timeout option.

Bun also passes no test context object to the test function, so the it.effect callbacks receive no ctx argument; import expect from @domir/bun-test instead of destructuring it from the context.

Publishing

Build the package first, then publish from the compiled output:

bun run build
cd dist
bun publish --tag rc

Pass --tag rc for prereleases such as 3.0.0-rc.0 so the latest dist-tag keeps pointing at the last stable release. The directory field under publishConfig in package.json is a pnpm convention that redirects pnpm publish to that folder automatically. Bun does not read it, so bun publish must be run from inside dist/ directly.