Attribute Macro rstest::fixture

#[fixture]

Define a fixture that you can use in all rstest’s test arguments. You should just mark your function as #[fixture] and then use it as a test’s argument. Fixture functions can also use other fixtures.

Let’s see a trivial example:

use rstest::*;

#[fixture]
fn twenty_one() -> i32 { 21 }

#[fixture]
fn two() -> i32 { 2 }

#[fixture]
fn injected(twenty_one: i32, two: i32) -> i32 { twenty_one * two }

#[rstest]
fn the_test(injected: i32) {
    assert_eq!(42, injected)
}

If the fixture function is an async function your fixture become an async fixture.

Default values

If you need to define argument default value you can use #[default(expression)] argument’s attribute:

use rstest::*;

#[fixture]
fn injected(
    #[default(21)]
    twenty_one: i32,
    #[default(1 + 1)]
    two: i32
) -> i32 { twenty_one * two }

#[rstest]
fn the_test(injected: i32) {
    assert_eq!(42, injected)
}

The expression could be any valid rust expression, even an async block if you need. Moreover, if the type implements FromStr trait you can use a literal string to build it.

#[fixture]
fn db_connection(
    #[default("127.0.0.1:9000")]
    addr: SocketAddr
) -> DbConnection {
    // create connection
}

Async

If you need you can write async fixtures to use in your async tests. Simply use async keyword for your function and the fixture become an async fixture.

use rstest::*;

#[fixture]
async fn async_fixture() -> i32 { 42 }


#[rstest]
async fn the_test(#[future] async_fixture: i32) {
    assert_eq!(42, async_fixture.await)
}

The #[future] argument attribute helps to remove the impl Future<Output = T> boilerplate. In this case the macro expands it in:

#[rstest]
async fn the_test(async_fixture: impl std::future::Future<Output = i32>) {
    assert_eq!(42, async_fixture.await)
}

If you need, you can use #[future] attribute also with an implicit lifetime reference because the macro will replace the implicit lifetime with an explicit one.

Rename

Sometimes you want to have long and descriptive name for your fixture but you prefer to use a much shorter name for argument that represent it in your fixture or test. You can rename the fixture using #[from(short_name)] attribute like following example:

use rstest::*;

#[fixture]
fn long_and_boring_descriptive_name() -> i32 { 42 }

#[rstest]
fn the_test(#[from(long_and_boring_descriptive_name)] short: i32) {
    assert_eq!(42, short)
}

#[once] Fixture

Especially in integration tests there are cases where you need a fixture that is called just once for every tests. rstest provides #[once] attribute for these cases.

If you mark your fixture with this attribute, then rstest will compute a static reference to your fixture result and return this reference to all your tests that need this fixture.

In follow example all tests share the same reference to the 42 static value.

use rstest::*;

#[fixture]
#[once]
fn once_fixture() -> i32 { 42 }

// Take care!!! You need to use a reference to the fixture value

#[rstest]
#[case(1)]
#[case(2)]
fn cases_tests(once_fixture: &i32, #[case] v: i32) {
    // Take care!!! You need to use a reference to the fixture value
    assert_eq!(&42, once_fixture)
}

#[rstest]
fn single(once_fixture: &i32) {
    assert_eq!(&42, once_fixture)
}

There are some limitations when you use #[once] fixture. rstest forbid to use once fixture for:

• async function
• Generic function (both with generic types or use impl trait)

Take care that the #[once] fixture value will never be dropped.

Partial Injection

You can also partially inject fixture dependency using #[with(v1, v2, ..)] attribute:

use rstest::*;

#[fixture]
fn base() -> i32 { 1 }

#[fixture]
fn first(base: i32) -> i32 { 1 * base }

#[fixture]
fn second(base: i32) -> i32 { 2 * base }

#[fixture]
fn injected(first: i32, #[with(3)] second: i32) -> i32 { first * second }

#[rstest]
fn the_test(injected: i32) {
    assert_eq!(-6, injected)
}

Note that injected value can be an arbitrary rust expression. #[with(v1, ..., vn)] attribute will inject v1, ..., vn expression as fixture arguments: all remaining arguments will be resolved as fixtures.

Sometimes the return type cannot be inferred so you must define it: For the few times you may need to do it, you can use the #[default(type)], #[partial_n(type)] function attribute to define it:

use rstest::*;

#[fixture]
pub fn i() -> u32 {
    42
}

#[fixture]
pub fn j() -> i32 {
    -42
}

#[fixture]
#[default(impl Iterator<Item=(u32, i32)>)]
#[partial_1(impl Iterator<Item=(I,i32)>)]
pub fn fx<I, J>(i: I, j: J) -> impl Iterator<Item=(I, J)> {
    std::iter::once((i, j))
}

#[rstest]
fn resolve_by_default(mut fx: impl Iterator<Item=(u32, i32)>) {
    assert_eq!((42, -42), fx.next().unwrap())
}

#[rstest]
fn resolve_partial(#[with(42.0)] mut fx: impl Iterator<Item=(f32, i32)>) {
    assert_eq!((42.0, -42), fx.next().unwrap())
}

partial_i is the fixture used when you inject the first i arguments in test call.

Old compact syntax

There is also a compact form for all previous features. This will maintained for a long time but for fixture I strongly recommend to migrate your code because you’ll pay a little verbosity but get back a more readable code.

Follow the previous examples in old compact syntax.

Default

#[fixture(twenty_one=21, two=2)]
fn injected(twenty_one: i32, two: i32) -> i32 { twenty_one * two }

Rename

#[fixture]
fn long_and_boring_descriptive_name() -> i32 { 42 }

#[rstest(long_and_boring_descriptive_name as short)]
fn the_test(short: i32) {
    assert_eq!(42, short)
}

Partial Injection

#[fixture(second(-3))]
fn injected(first: i32, second: i32) -> i32 { first * second }

Partial Type Injection

#[fixture(::default<impl Iterator<Item=(u32, i32)>>::partial_1<impl Iterator<Item=(I,i32)>>)]
pub fn fx<I, J>(i: I, j: J) -> impl Iterator<Item=(I, J)> {
    std::iter::once((i, j))
}