Codeguidelines.md

Code guidelines

Automated linting

We use automated linting tools where possible and these are configured using .editorconfig and rustfmt.toml. Code is linted using the following three commands:

RUSTFLAGS="-Dwarnings" cargo build --all --all-targets
cargo +nightly fmt
cargo clippy

NOTE: We require a nightly build of rustfmt.

External guidelines

We follow the

• "Rust style guide",
• "Rust API Guidelines",
• "Rust Book", and
• "Libra Coding Guidelines".

Except,

• Unit tests are in the same file as the code covered and not in separate files.
• We use quickcheck for property based testing.

In addition to the above, we use Criterion extensively for statistical benchmarking. See the section below for what is expected.

Making exceptions

Sometimes clippy will give a false positive, where compliance will decrease the safety and readability of the code. In this case a localized exception is in order.

Clippy exceptions should be added on the smallest reasonable scope. Often, this will be a single function. An exception can be made for unit tests, where it is okay to disable it for all tests in the file.

All exceptions need to have a comment line preceding explaining why the rule leads to false positives.

Testing

Be liberal with debug_assert statements.

In unit tests, the correct values are named expected and computed values are named actual. When comparing, they are ordered actual then expected like so:

let actual = square(2);
let expected = 4;
assert_eq!(actual, expected);

The expressions can be inlined as long as the above order is maintained.

assert_eq!(square(2), 4);

Benchmarking

Benchmarks should be added for functions that are critical to performance.