Testing
Mah has a built-in test runner. Tests live in *.test.mh files next to
the code they test, and mah test runs them.
# src/calc.test.mh
import "std:test"
fn add(a, b) { a + b }
test "adds two numbers" {
assert_eq(add(2, 3), 5)
assert(add(1, 1) > 1, "should grow")
}
test "division by zero throws" {
let e = assert_throws(fn() { 1 / 0 })
assert_eq(e.message, "Division by zero")
}
test "not ready yet" {
skip("needs real data")
} $ mah test
running 3 tests
test src/calc.test.mh::adds two numbers ... ok
test src/calc.test.mh::division by zero throws ... ok
test src/calc.test.mh::not ready yet ... skipped (needs real data)
test result: ok. 2 passed; 0 failed; 1 skipped; finished in 0.04s A new project from mah init already has one, src/main.test.mh.
Test files
- A test file holds only declarations (
import,fn,let,struct,enum,trait,impl) andtest "name" { ... }blocks. Anything that would run by itself, like a top-levelprint, is a compile error. testis only a keyword there: at the top level of a test file, right before the test’s name. Everywhere else it’s an ordinary name.mah runandmah buildrefuse a test file, and nothing can import one, so tests never end up in your program.x.test.mhimportingx.mhfrom the same directory sees all ofx.mh’s top-level names, exported or not, so private helpers can be tested directly. Every other importer still sees only exports.
Assertions: std:test
| Function | |
|---|---|
assert(cond, message = "") | fails unless cond is truthy |
assert_eq(actual, expected, message = "") | fails unless actual == expected; the report shows both |
assert_ne(a, b, message = "") | fails if a == b |
assert_throws(f) | calls f and returns what it threw; fails if it returned normally |
fail(message) | fails right away |
skip(reason = "") | stops the test and reports it as skipped |
A failing assertion throws an AssertionError, and skip throws SkipTest (both exported by std:test). A test body can throw anything else too: any other error
fails the test and is reported with its message and a Mah stack trace. assert_eq compares with ==, which compares Vectors, Maps, structs and
enums by identity, so assert_eq([1], [1]) fails. Compare their parts
instead.
Running tests
mah test # every *.test.mh under the project
mah test divide # tests whose name, or file::name, contains "divide"
mah test --file src/calc.test.mh
mah test --vm rust # run on mah-vm instead of the Python VM
mah test --timeout 500 # fail any test that takes longer than 500ms mah testsearches the whole project (the directory withmah-project.toml), skippingbuild/and hidden directories.- Every test runs in a fresh VM: the file’s declarations run again for each test, so no state leaks between tests.
- A test may
.awaitandsleep_async; it’s done when its body returns. Timers still pending then are dropped, with a warning next to the result. - A failure is reported at the line of the failing assertion in your test file, with the values involved and anything the test printed:
---- src/calc.test.mh::adds two numbers ----
assert_eq failed at src/calc.test.mh:8
actual: 4
expected: 5 The exit code is 1 if any test failed (or a test file doesn’t compile),
so mah test works as a CI step.