rust-coding-style

Rust Coding Style

Project-Specific Patterns

• Use the 2024 edition of Rust
• Prefer derive_more over manual trait implementations
• Feature flags in this codebase use the #[cfg(feature = "...")] pattern
• Invoke cargo clippy with --all-features, --all-targets, and --no-deps from the root
• Use cargo doc --no-deps --all-features for checking documentation
• Use rustfmt to format the code
• Use #[expect(lint, reason = "...")] over #[allow(lint)]

Type System

• Create strong types with newtype patterns for domain entities
• Consider visibility carefully (avoid unnecessary pub)

#[derive(Debug, Copy, Clone, Eq, Hash, PartialEq, derive_more::Display)]
pub struct UserId(Uuid);

Async Patterns

• Use impl Future<Output = T> + Send in trait definitions:

fn get_data(
    &self,
    id: String,
) -> impl Future<Output = Result<Data, Report<DataError>>> + Send {
    async move {
        // Implementation
    }
}

Function Arguments

• Functions should never take more than 7 arguments. If a function requires more than 7 arguments, encapsulate related parameters in a struct.
• Functions that use data immutably should take a reference to the data, while functions that modify data should take a mutable reference. Never take ownership of data unless the function explicitly consumes it.
• Make functions const whenever possible.
• Prefer the following argument types when applicable, but only if this does not reduce performance:
  • impl AsRef<str> instead of &str or &String
  • impl AsRef<Path> instead of &Path or &PathBuf
  • impl IntoIterator<Item = &T> when only iterating over the data
  • &[T] instead of &Vec<T>
  • &mut [T] instead of &mut Vec<T> when the function doesn't need to resize the vector
  • impl Into<Cow<T>> instead of Cow<T>
  • impl Into<Arc<T>> instead of Arc<T>
  • impl Into<Rc<T>> instead of Rc<T>
  • impl Into<Box<T>> instead of Box<T>
• Never use impl Into<Option<_>> as from reading the caller site, it's not visible that None could potentially be passed

From and Into

• Generally prefer From implementations over Into implementations. The Rust compiler will automatically derive Into from From, but not vice versa.
• When converting between types, prefer using the from method over into for clarity. The from method makes the target type explicit in the code, while into requires type inference.
• For wrapper types like Cow, Arc, Rc, Report, and Box, prefer using explicit constructors (e.g., Cow::from, Arc::new) instead of .into(). This improves readability by clearly indicating the target type.

Smart Pointers

• When cloning smart pointers such as Arc and Rc, always use Arc::clone(&pointer) and Rc::clone(&pointer) instead of pointer.clone(). This explicitly indicates you're cloning the reference, not the underlying data.

Instrumentation

• Annotate functions that perform significant work with #[tracing::instrument]
• Use tracing macros (e.g., trace!, debug!, info!, warn!, error!) instead of println! or eprintln! for logging

Allocations

• Minimize allocations when possible. For example, reuse a Vec in a loop instead of creating a new one in each iteration.
• Prefer borrowed data over owned data where appropriate.
• Balance performance and readability—if an allocation makes code significantly more readable or maintainable, the trade-off may be worthwhile.

Types

• Use newtypes when a value should carry specific semantics beyond its underlying type. This improves type safety and code clarity.

For example:

struct UserId(u64);  // instead of `type UserId = u64;` or `u64`

Naming Conventions

When suggesting names for variables, functions, or types:

• Do not prefix test-function names with test_, this would otherwise result in test::test_<name> names.
• Provide a concise list of naming options with brief explanations of why each fits the context
• Choose names of appropriate length—avoid names that are too long or too short
• Avoid abbreviations unless they are widely recognized in the domain (e.g., Http or Json is acceptable, but Ctx instead of Context is not)
• Do not suffix names with their types (e.g., use users instead of usersList)
• Do not repeat the type name in variable names (e.g., use user instead of userUser)

Crate Preferences

• Use similar_asserts for test assertions
• Use insta for snapshot tests
• Use test_log for better test output (#[test_log::test])
• Use tracing macros, not log macros
• Prefer tracing::instrument for function instrumentation

Import Style

• Don't use local imports within functions, or blocks
• Avoid wildcard imports like use super::*;, or use crate::module::*;
• Never use a prelude use crate::prelude::*
• Prefer explicit imports to make dependencies clear and improve code readability
• We prefer core over alloc over std for imports to minimize dependencies
  • Use core for functionality that doesn't require allocation
  • Use alloc when you need allocation but not OS-specific features
  • Only use std when necessary for OS interactions or when using core/alloc would be unnecessarily complex
• Prefer qualified imports (use foo::Bar; let x = Bar::new()) over fully qualified paths (let x = foo::Bar::new()) for frequently used types
• Use pub use re-exports in module roots to create a clean public API
• Avoid importing items with the same name from different modules; use qualified imports
• Import traits using use module::Trait as _; when you only need the trait's methods and not the trait name itself
  • This pattern brings trait methods into scope without name conflicts
  • Use this especially for extension traits or when implementing foreign traits on local types

// Good - Importing a trait just for its methods:
use std::io::Read as _;

// Example with trait methods:
fn read_file(file: &mut File) -> Result<String, std::io::Error> {
    // Read methods available without importing the Read trait name
    let mut content = String::new();
    file.read_to_string(&mut content)?;
    Ok(content)
}

// Bad - Directly importing trait when only methods are needed:
use std::io::Read;

// Good - Importing trait for implementing it:
use std::io::Write;
impl Write for MyWriter { /* implementation */ }

// Bad - Wildcard import:
mod tests {
    use super::*;  // Wildcard import

    #[test]
    fn test_something() {
        // Test implementation
    }
}

// Good - Explicit imports:
mod tests {
    use crate::MyStruct;
    use crate::my_function;

    #[test]
    fn test_something() {
        // Test implementation
    }
}

// Bad - Local import:
fn process_data() {
    use std::collections::HashMap;  // Local import
    let map = HashMap::new();
    // Implementation
}

// Good - Module-level import:
use std::collections::HashMap;

fn process_data() {
    let map = HashMap::new();
    // Implementation
}

// Bad - Using std when core would suffice:
use std::fmt::Display;

// Good - Using core for non-allocating functionality:
use core::fmt::Display;

// Bad - Using std when alloc would suffice:
use std::collections::BTreeSet;

// Good - Using alloc for allocation without full std dependency:
use alloc::vec::Vec;

// Appropriate - Using std when needed:
use std::fs::File;  // OS-specific functionality requires std

Libraries and Components

• Abstract integrations with third-party systems behind traits to maintain clean separation of concerns

Comments and Assertions

• Do not add comments after a line of code; place comments on separate lines above the code they describe
• When using assertions, include descriptive messages using the optional description parameter rather than adding a comment
• All expect() messages should follow the format "should ..." to clearly indicate the expected behavior

For example:

// Bad:
assert_eq!(result, expected); // This should match the expected value

// Good:
assert_eq!(result, expected, "Values should match expected output");

// Bad:
some_value.expect("The value is not None"); // This should never happen

// Good:
some_value.expect("should contain a valid value");