Similar to trybuild, but allows you to test how declarative or procedural macros are expanded.
Please refer to the documentation on docs.rs.
tryexpand requires cargo-expand to be installed.
Add tryexpand to your project as a dev-dependency by running
cargo install --dev tryexpand
Then under your crate's tests/ directory, create tests.rs file containing calls to tryexpand::expand() and populate the tests/expand/pass/, tests/expand/checked_pass/ and tests/expand/fail/ directories with corresponding Rust source files under test.
The tryexpand crate exposes the following functions test actions:
The tryexpand::expand(…) function runs cargo expand for each test file and snapshot the results:
tryexpand::expand(
// ...
)
.expect_pass();If you wish to also perform additional snapshot tests for all successfully expanded files you can do so via an additional call to either of these:
-
.and_check()to runcargo checkfor each test file and snapshot the results:tryexpand::expand( // ... ) .and_check() .expect_pass();
-
.and_run()to runcargo runfor each test file and snapshot the results:tryexpand::expand( // ... ) .and_run() .expect_pass();
-
.and_run_tests()to runcargo testfor each test file and snapshot the results:tryexpand::expand( // ... ) .and_run_tests() .expect_pass();
The tryexpand::check(…) function runs cargo check for each test file and snapshot the results:
tryexpand::check(
// ...
)
.expect_pass();The tryexpand::run(…) function runs cargo run for each test file and snapshot the results:
tryexpand::run(
// ...
)
.expect_pass();The tryexpand::run_tests(…) function runs cargo test for each test file and snapshot the results:
tryexpand::run_tests(
// ...
)
.expect_pass();The base of each tryexpand test suite is the corresponding test action function (we're using tryexpand::expand(…) here, but this applies to all actions), which you pass a list of file paths (or glob patterns) to:
#[test]
pub fn pass() {
tryexpand::expand(
["tests/expand/pass/*.rs"]
).expect_pass();
// or its short-hand (by default `.expect_pass()` is implied):
tryexpand::expand(
["tests/expand/pass/*.rs"]
);
}By default tryexpand's test action functions assert matched test files to pass their tests.
If instead you want to write tests for a failure's diagnostics, then ou can do so via an additional call to .expect_fail() (we're using tryexpand::expand(…) here, but this applies to all actions):
#[test]
pub fn fail() {
tryexpand::expand(
["tests/expand/fail/*.rs"]
).expect_fail();
}Additionally you can specify arguments to pass to the cargo command:
#[test]
tryexpand::expand(
// ...
)
// ...
.args(["--features", "test-feature"])
.expect_pass();As well as environment variables to set for the cargo command:
tryexpand::expand(
// ...
)
// ...
.envs([("MY_ENV", "my env var value")])
.expect_pass();The test can be run with:
cargo test
While it is possible to run parallel tests it is recommended to run them serially:
cargo test -- --test-threads=1
For debugging purposes you may want to see the output for all tests, not just the failing ones:
cargo test -- --no-capture
Each tryexpand test will invoke the cargo expand command (as well as any of the optional follow-up commands: cargo check, cargo run, cargo test) on each of the source files that matches the glob pattern and will compare the expansion result with the corresponding *.out.rs, *.out.txt or *.err.txt snapshot files.
If the environment variable TRYEXPAND=overwrite is provided (e.g. $ TRYEXPAND=overwrite cargo test), then snapshot files will be created, or overwritten, if one already exists. Snapshot files should get checked into version control.
Hand-writing snapshot files is not recommended.
When working with multiple expansion test files, it is recommended to specify wildcard (*.rs) instead of doing a multiple calls to the expand functions for individual files.
Usage of wildcards for multiple files will group them under a single temporary crate for which dependencies will be built a single time. In contrast, calling expand functions for each source file will create multiple temporary crates and that will reduce performance as dependencies will be build for each of the temporary crates.
More info on how glob patterns work.
See tests/macro-tests and tests/proc-macro-tests as a reference.
Since each rustc/cargo release might make changes to the emitted diagnostics
it is recommended to run tryexpand tests using a pinned toolchain, e.g.:
cargo +1.76.0 test <OPTIONS>
For each expand()-like method call within your tests a temporary and uniquely named Rust project will get generated within $CARGO_TARGET_DIR/target/tests/.
By default these projects will get deleted upon test completion (regardless of the outcome).
If you wish to take a look at the actual code/projects being expanded you can provide TRYEXPAND_KEEP_ARTIFACTS=1 and tryexpand will skip the cleanup:
TRYEXPAND_KEEP_ARTIFACTS=1 cargo test
By default tryexpand truncates console output that's longer than 100 lines.
If you wish to temporarily turn this behavior you can provide TRYEXPAND_TRUNCATE_OUTPUT=0 and tryexpand will produce the full console output:
TRYEXPAND_TRUNCATE_OUTPUT=0 cargo test
If you wish to see the exact commands that tryexpand is executing you can provide TRYEXPAND_DEBUG_LOG=1 (together with --nocapture, if you also wish to see logs of passing tests) and tryexpand will print additional debug logs:
TRYEXPAND_DEBUG_LOG=0 cargo test -- --nocapture
Please read CONTRIBUTING.md for details on our code of conduct,
and the process for submitting pull requests to us.
We use SemVer for versioning. For the versions available, see the tags on this repository.
This project is licensed under the MIT or Apache-2.0 – see the LICENSE-MIT.md/LICENSE-APACHE.md files for details.
The tryexpand crate originated as a fork of eupn's macrotest (crates.io).