2 unstable releases
0.2.0 | May 12, 2024 |
---|---|
0.1.0 | May 9, 2024 |
#330 in Debugging
25 downloads per month
Used in 2 crates
14KB
126 lines
close-to-rs
Provides a function, Tolerate, to verify the equivalence of floating-point numbers at arbitrary precision.
任意の精度で浮動小数点数の等価性を検証するための関数、トレイトを提供します。
An alias also exists for compatibility with assert-be-close, so changing the crate does not require changing the code.
assert-be-closeとの互換性を保つためのエイリアスも存在するため、クレートを変更してもコードを変更する必要がありません。
使い方(Usage)
Compare values with arbitrary precision.
任意の精度で数値を比較する。
use close_to::{close_to, far_from};
fn example() {
let (is_close, expected_diff, received_diff) = close_to(1.0, 1.0001, 3);
assert!(is_close);
let (is_close, expected_diff, received_diff) = close_to(1.0, 1.0001, 4);
assert!(is_close); // panic!
let (is_close, expected_diff, received_diff) = far_from(1.0, 1.0001, 4);
assert!(is_close);
let (is_close, expected_diff, received_diff) = far_from(1.0, 1.0001, 3);
assert!(is_close); // panic!
}
Different types can be compared if the first argument T
is a floating-point number and the second argument U
implements Into<T>
.
異なる型同士でも、第1引数T
が浮動小数点数かつ、第2引数U
がInto<T>
を実装していれば比較できます。
use close_to::{close_to, far_from};
fn example() {
close_to(1.0_f64, 1.0001_f32, 3);
close_to(1.0001_f64, 1_u8, 3);
close_to(1.0_f32, 1.0001_f64, 3); // the trait bound `f32: std::convert::From<f64>` is not satisfied [E0277]
}
There is also a function that panics if the result of the comparison is false, along with a debug expression.
また、比較した結果がfalseである場合、デバック表現とともにpanicする関数もあります。
use close_to::{assert_close_to, assert_far_from};
fn example() {
assert_close_to(1.0, 1.0001, 4); // panic with the following message
// assertion 'left ≈ right` failed
// left: 1
// right: 1.0001
// received_diff: 0.00009999999999998899
// expected_diff: 0.00005
assert_far_from(1.0, 1.0001, 3); // panic with the following message
// assertion 'left != right` failed
// left: 1
// right: 1.0001
// received_diff: 0.00009999999999998899
// expected_diff: 0.0005
}
Traits are also provided to implement these functions.
By default, they are implemented for f32
and f64
.
T.close_to(U, precision)
and close_to(T, U, precision)
have the same behavior.
これらの機能を実装するためのトレイトも提供しています。
デフォルトでは、f32
とf64
に対して実装されています。
T.close_to(U, precision)
とclose_to(T, U, precision)
は同じ挙動をします。
use close_to::{AssertCloseTo, CloseTo};
fn example() {
let is_close = 1_f64.close_to(1.0001, 3).0;
assert!(is_close);
let is_close = 1_f64.close_to(1.0001, 4).0;
assert!(is_close); // panic!
1_f64.assert_close_to(1.0001, 4); // panic with the same message as `assert_close_to`
1_f64.assert_far_from(1.0001, 3); // panic with the same message as `assert_far_from`
}
ライセンス(License)
Licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)
at your option.
Dependencies
~465KB