Code Style Guide

Formatting
Indent using tabs.
Lines should be longer than 100 characters long only in exceptional circumstances and certainly no longer than 120. For this purpose, tabs are considered 4 characters wide.
Indent levels should be greater than 5 only in exceptional circumstances and certainly no greater than 8. If they are greater than 5, then consider using let or auxiliary functions in order to strip out complex inline expressions.
Never have spaces on a line prior to a non-whitespace character
Follow-on lines are only ever a single indent from the original line.
1
fn calculation(some_long_variable_a: i8, some_long_variable_b: i8) -> bool {
2
    let x = some_long_variable_a * some_long_variable_b
3
        - some_long_variable_b / some_long_variable_a
4
        + sqrt(some_long_variable_a) - sqrt(some_long_variable_b);
5
    x > 10
6
}
Copied!
Indent level should follow open parens/brackets, but should be collapsed to the smallest number of levels actually used:
1
fn calculate(
2
    some_long_variable_a: f32,
3
    some_long_variable_b: f32,
4
    some_long_variable_c: f32,
5
) -> f32 {
6
    (-some_long_variable_b + sqrt(
7
        // two parens open, but since we open & close them both on the
8
        // same line, only one indent level is used
9
        some_long_variable_b * some_long_variable_b
10
        - 4 * some_long_variable_a * some_long_variable_c
11
    // both closed here at beginning of line, so back to the original indent
12
    // level
13
    )) / (2 * some_long_variable_a)
14
}
Copied!
where is indented, and its items are indented one further.
Argument lists or function invocations that are too long to fit on one line are indented similarly to code blocks, and once one param is indented in such a way, all others should be, too. Run-on parameter lists are also acceptable for single-line run-ons of basic function calls.
1
// OK
2
fn foo(
3
    really_long_parameter_name_1: SomeLongTypeName,
4
    really_long_parameter_name_2: SomeLongTypeName,
5
    shrt_nm_1: u8,
6
    shrt_nm_2: u8,
7
) {
8
   ...
9
}
10
​
11
// NOT OK
12
fn foo(really_long_parameter_name_1: SomeLongTypeName, really_long_parameter_name_2: SomeLongTypeName,
13
    shrt_nm_1: u8, shrt_nm_2: u8) {
14
    ...
15
}
Copied!
1
{
2
    // Complex line (not just a function call, also a let statement). Full
3
    // structure.
4
    let (a, b) = bar(
5
        really_long_parameter_name_1,
6
        really_long_parameter_name_2,
7
        shrt_nm_1,
8
        shrt_nm_2,
9
    );
10
​
11
    // Long, simple function call.
12
    waz(
13
        really_long_parameter_name_1, 
14
        really_long_parameter_name_2,
15
        shrt_nm_1, 
16
        shrt_nm_2,
17
    );
18
​
19
    // Short function call. Inline.
20
    baz(a, b);
21
}
Copied!
Always end last item of a multi-line comma-delimited set with , when legal:
1
struct Point<T> {
2
    x: T,
3
    y: T,    // <-- Multiline comma-delimited lists end with a trailing ,
4
}
5
​
6
// Single line comma-delimited items do not have a trailing `,`
7
enum Meal { Breakfast, Lunch, Dinner };
Copied!
Avoid trailing ;s where unneeded.
1
if condition {
2
    return 1    // <-- no ; here
3
}
Copied!
match arms may be either blocks or have a trailing , but not both.
Blocks should not be used unnecessarily.
1
match meal {
2
    Meal::Breakfast => "eggs",
3
    Meal::Lunch => { check_diet(); recipe() },
4
//    Meal::Dinner => { return Err("Fasting") }   // WRONG
5
    Meal::Dinner => return Err("Fasting"),
6
}
Copied!
Style
Panickers require explicit proofs they don't trigger. Calling unwrap is discouraged. The exception
to this rule is test code. Avoiding panickers by restructuring code is preferred if feasible.
1
let mut target_path = 
2
    self.path().expect(
3
        "self is instance of DiskDirectory;\
4
        DiskDirectory always returns path;\
5
        qed"
6
    );
Copied!
Unsafe code requires explicit proofs just as panickers do. When introducing unsafe code,
consider tradeoffs between efficiency on one hand and reliability, maintenance costs, and security on the other.
Here is a list of questions that may help evaluating the tradeoff while preparing or reviewing a PR:
how much more performant or compact the resulting code will be using unsafe code,
how likely is it that invariants could be violated,
are issues stemming from the use of unsafe code caught by existing tests/tooling,
what are the consequences if the problems slip into production.

Build - Previous

Development

Next - Non Fungible Tokens

Cherry Originals

Last modified 2mo ago

Copy link

Contents

Formatting

Style