#rule #non-empty #refined #type #string #target #define

refined_type

A library for imbuing rules into types and elevating them to more robust types

21 releases

0.5.4 May 29, 2024
0.5.3 May 29, 2024
0.4.16 Apr 3, 2024
0.4.14 Mar 31, 2024
0.1.5 Nov 12, 2023

#254 in Rust patterns

MIT license

86KB
2K SLoC

Refined-Type

refined-type is a library developed for Rust. It enhances your types, making them more robust and expanding the range of guarantees your applications can statically ensure.

Overview

You can create various rules for a certain type, such as phone numbers, addresses, times, and so on. Once you have established the rules, you can easily combine them. Specifically, if you create rules for 'non-empty strings' and 'strings composed only of alphabets,' you do not need to redefine a new rule for 'non-empty strings composed only of alphabets'. All rules can be arbitrarily combined and extended as long as the target type matches. Enjoy a wonderful type life!

Example Usage

As an example, let's convert from JSON to a struct.

// define the constraints you expect by combining 'Refined' and 'Rule'.
type MyNonEmptyString = Refined<NonEmptyRule<String>>;
type MyNonEmptyVec<T> = Refined<NonEmptyRule<Vec<T>>>;

// define a struct for converting from JSON.
#[derive(Debug, Eq, PartialEq, Deserialize)]
struct Human {
    name: MyNonEmptyString,
    friends: MyNonEmptyVec<String>,
}

fn example_1() -> anyhow::Result<()> {
    let json = json! {{
        "name": "john",
        "friends": ["tom", "taro"]
    }}
        .to_string();

    let actual = serde_json::from_str::<Human>(&json)?;
    let expected = Human {
        name: MyNonEmptyString::new("john".to_string())?,
        friends: MyNonEmptyVec::new(vec!["tom".to_string(), "taro".to_string()])?,
    };
    assert_eq!(actual, expected);
    Ok(())
}

// In the second example, while `friends` meets the rule, `name` does not, causing the conversion from JSON to fail
fn example_2() -> anyhow::Result<()> {
    let json = json! {{
        "name": "",
        "friends": ["tom", "taro"]
    }}
        .to_string();

    // because `name` is empty
    assert!(serde_json::from_str::<Human>(&json).is_err());
    Ok(())
}

// In the third example, while `name` satisfies the rule, `friends` does not, causing the conversion from JSON to fail.
fn example_3() -> anyhow::Result<()> {
    let json = json! {{
        "name": "john",
        "friends": []
    }}
        .to_string();

    // because `friends` is empty
    assert!(serde_json::from_str::<Human>(&json).is_err());
    Ok(())
}

Installation

cargo add refined_type

Custom Rule

There are many situations where you may want to define custom rules. To define rules for a specific target type, you first need to define a struct. In the struct, define fields for specifying detailed conditions. Once the definition is complete, all you need to do is implement the Rule trait. Add your preferred conditions as you like.

fn example_4() -> anyhow::Result<()> {
    let non_empty_string_result = Refined::<NonEmptyStringRule>::new("Hello World".to_string())?;
    assert_eq!(non_empty_string_result.into_value(), "Hello World");

    let empty_string_result = Refined::<NonEmptyStringRule>::new("".to_string());
    assert!(empty_string_result.is_err())
    Ok(())
}

Compose Rules

As mentioned earlier, it is possible to combine any rules as long as the target types match. In the example below, there are standalone rules for 'strings containing Hello' and 'strings containing World'. Since their target type is String, combining them is possible. I have prepared something called Rule Composer (And, Or, Not). By using Rule Composer, composite rules can be easily created.

Original Rules

struct ContainsHelloRule;

struct ContainsWorldRule;

impl Rule for ContainsHelloRule {
    type Item = String;

    fn validate(target: &Self::Item) -> Result<(), Error> {
        if target.contains("Hello") {
            Ok(())
        } else {
            Err(Error::new(format!("{} does not contain `Hello`", target)))
        }
    }
}

impl Rule for ContainsWorldRule {
    type Item = String;

    fn validate(target: &Self::Item) -> Result<(), Error> {
        if target.contains("World") {
            Ok(())
        } else {
            Err(Error::new(format!("{} does not contain `World`", target)))
        }
    }
}

1: And Rule Composer

And Rule Composer is a rule that satisfies both of the two rules. It is generally effective when you want to narrow down the condition range.

fn example_5() {
    type HelloAndWorldRule = And<ContainsHelloRule, ContainsWorldRule>;

    let rule_ok = Refined::<HelloAndWorldRule>::new("Hello! World!".to_string());
    assert!(rule_ok.is_ok());

    let rule_err = Refined::<HelloAndWorldRule>::new("Hello, world!".to_string());
    assert!(rule_err.is_err());
}

2: Or Rule Composer

Or Rule Composer is a rule that satisfies either of the two rules. It is generally effective when you want to expand the condition range.

fn example_6() {
    type HelloOrWorldRule = Or<ContainsHelloRule, ContainsWorldRule>;

    let rule_ok_1 = Refined::<HelloOrWorldRule>::new("Hello! World!".to_string());
    assert!(rule_ok_1.is_ok());

    let rule_ok_2 = Refined::<HelloOrWorldRule>::new("hello World!".to_string());
    assert!(rule_ok_2.is_ok());

    let rule_err = Refined::<HelloOrWorldRule>::new("hello, world!".to_string());
    assert!(rule_err.is_err());
}

3: Not Rule Composer

Not Rule Composer is a rule that does not satisfy a specific condition. It is generally effective when you want to discard only certain situations.

fn example_7() {
    type NotHelloRule = Not<ContainsHelloRule>;

    let rule_ok = Refined::<NotHelloRule>::new("hello! World!".to_string());
    assert!(rule_ok.is_ok());

    let rule_err = Refined::<NotHelloRule>::new("Hello, World!".to_string());
    assert!(rule_err.is_err());
}

4: Compose Rule Composer

Rule Composer is also a rule. Therefore, it can be treated much like a composite function

struct StartsWithHelloRule;

struct StartsWithByeRule;

struct EndsWithJohnRule;

impl Rule for StartsWithHelloRule {
    type Item = String;

    fn validate(target: &Self::Item) -> Result<(), Error> {
        if target.starts_with("Hello") {
            Ok(())
        } else {
            Err(Error::new(format!("{} does not start with `Hello`", target)))
        }
    }
}

impl Rule for StartsWithByeRule {
    type Item = String;

    fn validate(target: &Self::Item) -> Result<(), Error> {
        if target.starts_with("Bye") {
            Ok(())
        } else {
            Err(Error::new(format!("{} does not start with `Bye`", target)))
        }
    }
}

impl Rule for EndsWithJohnRule {
    type Item = String;

    fn validate(target: &Self::Item) -> Result<(), Error> {
        if target.ends_with("John") {
            Ok(())
        } else {
            Err(Error::new(format!("{} does not end with `John`", target)))
        }
    }
}

fn example_8() {
    type GreetingRule = And<Or<StartsWithHelloRule, StartsWithByeRule>, EndsWithJohnRule>;

    assert!(GreetingRule::validate(&"Hello! Nice to meet you John".to_string()).is_ok());
    assert!(GreetingRule::validate(&"Bye! Have a good day John".to_string()).is_ok());
    assert!(GreetingRule::validate(&"How are you? Have a good day John".to_string()).is_err());
    assert!(GreetingRule::validate(&"Bye! Have a good day Tom".to_string()).is_err());
}

JSON

refined_type is compatible with serde_json. This ensures type-safe communication and eliminates the need to write new validation processes. All you need to do is implement a set of rules once and implement serde’s Serialize and Deserialize.

Serialize

#[derive(Debug, Eq, PartialEq, Deserialize, Serialize)]
struct Human2 {
    name: NonEmptyString,
    age: u8,
}

fn example_9() -> anyhow::Result<()> {
    let john = Human2 {
        name: NonEmptyString::new("john".to_string())?,
        age: 8,
    };

    let actual = json!(john);
    let expected = json! {{
        "name": "john",
        "age": 8
    }};
    assert_eq!(actual, expected);
    Ok(())
}

Deserialize

fn example_10() -> anyhow::Result<()> {
    let json = json! {{
        "name": "john",
        "age": 8
    }}
        .to_string();

    let actual = serde_json::from_str::<Human2>(&json)?;

    let expected = Human2 {
        name: NonEmptyString::new("john".to_string())?,
        age: 8,
    };
    assert_eq!(actual, expected);
    Ok(())
}

Number

You can also represent the size of numbers as types. I have prepared macros that can easily define the size of numbers. Let’s use them to define a Age type that is narrowed down to ages 18 to 80.

greater_rule!((18, u8));
less_rule!((80, u8));
equal_rule!((18, u8), (80, u8));

type Age = Refined<TargetAgeRule>;

// 18 <= age
type TargetAge18OrMore = Or<EqualRule18u8, GreaterRule18u8>;

// age <= 80
type TargetAge80OrLess = Or<EqualRule80u8, LessRule80u8>;

// 18 <= age <= 80
type TargetAgeRule = And<TargetAge18OrMore, TargetAge80OrLess>;

Iterator

I have also prepared several useful refined types for Iterators.

ForAll

ForAll is a rule that applies a specific rule to all elements in the Iterator.

fn example_11() -> anyhow::Result<()> {
    let vec = vec!["Hello".to_string(), "World".to_string()];
    let for_all_ok = ForAll::<NonEmptyStringRule, _>::new(vec.clone())?;
    assert_eq!(vec, for_all_ok.into_value());

    let vec = vec!["Hello".to_string(), "".to_string()];
    let for_all_err = ForAll::<NonEmptyStringRule, _>::new(vec.clone());
    assert!(for_all_err.is_err());
    Ok(())
}

Exists

Exists is a rule that applies a specific rule to at least one element in the Iterator.

fn example_12() -> anyhow::Result<()> {
    let vec = vec!["Hello".to_string(), "".to_string()];
    let exists_ok = Exists::<NonEmptyStringRule, _>::new(vec.clone())?;
    assert_eq!(vec, exists_ok.into_value());

    let vec = vec!["".to_string(), "".to_string()];
    let exists_err = Exists::<NonEmptyStringRule, _>::new(vec.clone());
    assert!(exists_err.is_err());
    Ok(())
}

into_iter() and iter()

The Iterator I’ve prepared has into_iter and iter implemented. Therefore, you can easily map or convert it to a different Iterator using collect. Feel free to explore the capabilities of the Iterator you’ve been given!

into_iter()

fn example_11() -> anyhow::Result<()> {
    let ne_vec = NonEmptyVec::new(vec![1, 2, 3])?;
    let ne_vec: NonEmptyVec<i32> = ne_vec.into_iter().map(|n| n * 2).map(|n| n * 3).collect();
    assert_eq!(ne_vec.into_value(), vec![6, 12, 18]);
    Ok(())
}

iter()

fn example_12() -> anyhow::Result<()> {
    let ne_vec = NonEmptyVec::new(vec![1, 2, 3])?;
    let ne_vec: NonEmptyVec<i32> = ne_vec.iter().map(|n| n * 2).map(|n| n * 3).collect();
    assert_eq!(ne_vec.into_value(), vec![6, 12, 18]);
    Ok(())
}

NonEmptyVec to NonEmptyVecDeque using collect()

fn example_13() -> anyhow::Result<()> {
    let ne_vec = NonEmptyVec::new(vec![1, 2, 3])?;
    let ne_vec_deque: NonEmptyVecDeque<i32> = ne_vec.into_iter().collect();
    assert_eq!(ne_vec_deque.into_value(), vec![1, 2, 3]);
    Ok(())
}

Add Trait

I have implemented the Add trait for a part of the Refined that I provided. Therefore, operations can be performed without downgrading the type level.

NonEmptyString

fn example_14() -> anyhow::Result<()> {
    let non_empty_string_1 = NonEmptyString::new("Hello".to_string())?;
    let non_empty_string_2 = NonEmptyString::new("World".to_string())?;
    let non_empty_string = non_empty_string_1 + non_empty_string_2; // This is also `NonEmptyString` type

    assert_eq!(non_empty_string.into_value(), "HelloWorld");
    Ok(())
}

NonEmptyVec

fn example_15() -> anyhow::Result<()> {
    let ne_vec_1 = NonEmptyVec::new(vec![1, 2, 3])?;
    let ne_vec_2 = NonEmptyVec::new(vec![4, 5, 6])?;
    let ne_vec = ne_vec_1 + ne_vec_2; // This is also `NonEmptyVec` type

    assert_eq!(ne_vec.into_value(), vec![1, 2, 3, 4, 5, 6]);
    Ok(())
}

Length

You can impose constraints on objects that have a length, such as String or Vec.

String

fn example_16() -> Result<(), Error> {
    length_greater_than!(5);
    length_equal!(5, 10);
    length_less_than!(10);

    type Password = Refined<From5To10Rule<String>>;

    type From5To10Rule<T> = And<
        Or<LengthEqualRule5<T>, LengthGreaterThanRule5<T>>,
        Or<LengthLessThanRule10<T>, LengthEqualRule10<T>>,
    >;

    // length is 8. so, this is valid
    let raw_password = "password";
    let password = Password::new(raw_password.to_string())?;
    assert_eq!(password.into_value(), "password");

    // length is 4. so, this is invalid
    let raw_password = "pswd";
    let password = Password::new(raw_password.to_string());
    assert!(password.is_err());

    // length is 17. so, this is invalid
    let raw_password = "password password";
    let password = Password::new(raw_password.to_string());
    assert!(password.is_err());

    Ok(())
}

Vec

#[test]
fn example_17() -> anyhow::Result<()> {
    length_greater_than!(5);
    length_equal!(5, 10);
    length_less_than!(10);

    type Friends = Refined<From5To10Rule<Vec<String>>>;

    type From5To10Rule<T> = And<
        Or<LengthEqualRule5<T>, LengthGreaterThanRule5<T>>,
        Or<LengthLessThanRule10<T>, LengthEqualRule10<T>>,
    >;

    // length is 6. so, this is valid
    let raw_friends = vec![
        "Tom".to_string(),
        "Taro".to_string(),
        "Jiro".to_string(),
        "Hanako".to_string(),
        "Sachiko".to_string(),
        "Yoshiko".to_string(),
    ];
    let friends = Friends::new(raw_friends.clone())?;
    assert_eq!(friends.into_value(), raw_friends);

    // length is 2. so, this is invalid
    let raw_friends = vec!["Tom".to_string(), "Taro".to_string()];
    let friends = Friends::new(raw_friends.clone());
    assert!(friends.is_err());

    // length is 11. so, this is invalid
    let raw_friends = vec![
        "Tom".to_string(),
        "Taro".to_string(),
        "Jiro".to_string(),
        "Hanako".to_string(),
        "Sachiko".to_string(),
        "Yuiko".to_string(),
        "Taiko".to_string(),
        "John".to_string(),
        "Jane".to_string(),
        "Jack".to_string(),
        "Jill".to_string(),
    ];
    let friends = Friends::new(raw_friends.clone());
    assert!(friends.is_err());

    Ok(())
}

Custom Length

You can define a length for any type. Therefore, if you want to implement a length that is not provided by refined_type, you can easily do so using LengthDefinition.

#[test]
fn example_18() -> anyhow::Result<()> {
    length_equal!(5);

    #[derive(Debug, PartialEq)]
    struct Hello;
    impl LengthDefinition for Hello {
        fn length(&self) -> usize {
            5
        }
    }

    let hello = Refined::<LengthEqualRule5<Hello>>::new(Hello)?;
    assert_eq!(hello.into_value(), Hello);
    Ok(())
}

Tips

Directly writing And, Or, Not or Refined can often lead to a decrease in readability. Therefore, using type aliases can help make your code clearer.

type ContainsHelloAndWorldRule = And<ContainsHelloRule, ContainsWorldRule>;

type ContainsHelloAndWorld = Refined<ContainsHelloAndWorldRule>;

License

MIT License

Copyright (c) 2024 Tomoki Someya

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Dependencies

~2.5–4MB
~76K SLoC