bun test --coverage: code coverage table and lcov report

bun test --coverage runs the tests and then prints a table with the share of functions and lines each source file executed, plus the line numbers that never ran. It needs no extra package (no Istanbul or c8). Add --coverage-reporter=lcov to write coverage/lcov.info for CI services and editors.

Example

math.ts has two functions; the test only calls one of them:

cat > math.ts <<'EOF'
export function add(a: number, b: number) {
  return a + b;
}
export function divide(a: number, b: number) {
  if (b === 0) throw new Error("division by zero");
  return a / b;
}
EOF
cat > math.test.ts <<'EOF'
import { expect, test } from "bun:test";
import { add } from "./math";
test("add", () => expect(add(1, 2)).toBe(3));
EOF
npx --yes bun@1 test --coverage 2>&1 | grep -E "\|"
echo "--- lcov"
npx --yes bun@1 test --coverage --coverage-reporter=lcov >/dev/null 2>&1
ls coverage

-----------|---------|---------|-------------------
File       | % Funcs | % Lines | Uncovered Line #s
-----------|---------|---------|-------------------
All files  |   50.00 |   40.00 |
 math.ts   |   50.00 |   40.00 | 4-6
-----------|---------|---------|-------------------
--- lcov
lcov.info

% Funcs counts functions that were called at least once and % Lines the executed lines; Uncovered Line #s points at the body of divide, which no test reached. The test file itself is excluded by default.

Options

bunfig.toml

[test]
coverage = true
coverageReporter = ["text", "lcov"]
coverageThreshold = { lines = 0.8, functions = 0.8 }

With the threshold set, bun test exits with a non-zero code when either value is below 80 percent, which fails a CI job.

Reading the numbers

• Coverage is computed only for files that were loaded. Add a test that imports a new module, otherwise it is invisible in the table.
• The All files row is an average over the listed files, not over the project.
• Line coverage does not mean every branch was tested: if (b === 0) counts as covered once the line ran, whatever the condition was.
• Running a subset of tests (bun test math) lowers the numbers, because fewer files and lines run.

More test runner options, such as filters and watch mode, are in bun test.

FAQ