Crates.io | exhaustive |
lib.rs | exhaustive |
version | |
source | src |
created_at | 2024-04-05 16:28:25.836425 |
updated_at | 2024-11-24 18:47:39.374668 |
description | The trait for generating all values of a type, and a property-based test macro. |
homepage | |
repository | https://github.com/JonathanBrouwer/exhaustive |
max_upload_size | |
id | 1197502 |
Cargo.toml error: | TOML parse error at line 18, column 1 | 18 | autolib = false | ^^^^^^^ unknown field `autolib`, expected one of `name`, `version`, `edition`, `authors`, `description`, `readme`, `license`, `repository`, `homepage`, `documentation`, `build`, `resolver`, `links`, `default-run`, `default_dash_run`, `rust-version`, `rust_dash_version`, `rust_version`, `license-file`, `license_dash_file`, `license_file`, `licenseFile`, `license_capital_file`, `forced-target`, `forced_dash_target`, `autobins`, `autotests`, `autoexamples`, `autobenches`, `publish`, `metadata`, `keywords`, `categories`, `exclude`, `include` |
size | 0 |
Exhaustive
The trait for generating all values of a type, and a property-based test macro.
This crate provides a trait to generate all values (up to a certain depth) of a type. It also provides a derive macro that will derive this trait for you. Finally, it provides a test macro that run a property-based test for all values of a type.
#[derive(Debug, Exhaustive)]
enum Test1 { A, B, }
#[derive(Debug, Exhaustive)]
struct Test2 { a: bool }
#[exhaustive_test]
fn test(v: Test1, w: Test2) {
println!("{v:?} {w:?}");
}
A Test2 { a: false }
A Test2 { a: true }
B Test2 { a: false }
B Test2 { a: true }
In order to implement Exhaustive
for recursive types or arbitrary-length types such as Vec
s, the crate supports limiting the size of types.
This is implemented by limiting the amount of choices that can be made during the generation of the type.
When using the exhaustive_test
macro, you can limit the amount of choices using:
#[exhaustive_test(20)]
The amount of choices that is made during generation of a type is not guaranteed to be stable between releases, though it is always guaranteed to be linear with the size of type in memory. This means that for most types, the amount of choices is exponential with the runtime of the test. The recommended way of choosing this value is to pick it experimentally based on the desired runtime of the test.
The trait is implemented for most standard library types, such as Vec
, Result
, Option
, Box
, HashMap
, HashSet
, tuples, arrays, etc.
The primary exception is the number types, since these grow too quickly to exhaustively test.
The derive macro can derive Exhaustive
for any struct
or enum
, whose fields all implement Exhaustive
.
For this to work, the macros
feature of this crate needs to be enabled. This feature is enabled by default.
The trait can be implemented by hand. The signature of the trait is:
trait Exhaustive: Sized {
fn generate(u: &mut DataSourceTaker) -> Result<Self>;
}
The DataSourceTaker
has two important functions that can be used to generate your type:
fn choice(&mut self, range: usize) -> Result<usize>;
fn iter_of<T: Exhaustive>(&mut self) -> Result<DataSourceTakerIter<T>>;
choice
will generate a number in 0..range
.
This can be used to choose a variant of an enum, or to make any other choice you need to make while generating your datatype.iter_of
will generate an iterator of T
.
This iterator will always be finitely-sized, and should be fully consumed.
iter_of
makes 1 choice to determine the size of the iterator, after which the generation of the elements may make more choices.Any type that implements the Exhaustive
trait will have an auto-implementation of the following function:
fn iter_exhaustive(max_choices: Option<usize>) -> impl Iterator<Item=Self>;
This function provides an iterator of all generated values of the type with optionally a maximum number of choices.
The arbitrary crate was a big inspiration for this crate. It provides similar functionality to this crate, but it made a few different design decisions that make it unusable for efficient exhaustive testing.
Types are always generated from a byte slice, which means that it is not possible to make exact choices, only to read a byte and make a choice depending on that.
This is fine for the intended use case of arbitrary
, which is generating arbitrary values.
But to generate all values exhaustively this means we'd have to generate all byte sequences up to possibly a big size. This ends up generating a LOT of duplicate values, and therefore scales poorly.