28 releases (17 breaking)

0.18.1 Apr 11, 2024
0.17.0 Mar 4, 2024
0.16.0 Jul 25, 2022
0.14.0 Jun 29, 2021
0.1.0 Dec 28, 2016

#28 in #validator

Download history 36067/week @ 2023-12-23 75743/week @ 2023-12-30 101147/week @ 2024-01-06 105498/week @ 2024-01-13 104147/week @ 2024-01-20 102635/week @ 2024-01-27 92785/week @ 2024-02-03 99196/week @ 2024-02-10 100143/week @ 2024-02-17 118054/week @ 2024-02-24 127889/week @ 2024-03-02 117227/week @ 2024-03-09 118087/week @ 2024-03-16 119154/week @ 2024-03-23 133176/week @ 2024-03-30 101442/week @ 2024-04-06

488,826 downloads per month
Used in 237 crates (15 directly)

MIT license



Macros 1.1 custom derive to simplify struct validation inspired by marshmallow and Django validators.

The minimum supported version is Rust 1.42.


validator = { version = "0.16", features = ["derive"] }

A short example:

use serde::Deserialize;

// A trait that the Validate derive will impl
use validator::{Validate, ValidationError};

#[derive(Debug, Validate, Deserialize)]
struct SignupData {
    mail: String,
    site: String,
    #[validate(length(min = 1), custom(function = "validate_unique_username"))]
    #[serde(rename = "firstName")]
    first_name: String,
    #[validate(range(min = 18, max = 20))]
    age: u32,
    #[validate(range(exclusive_min = 0.0, max = 100.0))]
    height: f32,

fn validate_unique_username(username: &str) -> Result<(), ValidationError> {
    if username == "xXxShad0wxXx" {
        // the value of the username will automatically be added later
        return Err(ValidationError::new("terrible_username"));


match signup_data.validate() {
  Ok(_) => (),
  Err(e) => return e;

A validation on an Option<_> field will be executed on the contained type if the option is Some. The validate() method returns a Result<(), ValidationErrors>. In the case of an invalid result, the ValidationErrors instance includes a map of errors keyed against the struct's field names. Errors may be represented in three ways, as described by the ValidationErrorsKind enum:

#[derive(Debug, Serialize, Clone, PartialEq)]
pub enum ValidationErrorsKind {
    List(BTreeMap<usize, Box<ValidationErrors>>),

In the simple example above, any errors would be of the Field(Vec<ValidationError>) type, where a single ValidationError has the following structure:

#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
pub struct ValidationError {
  pub code: Cow<'static, str>,
  pub message: Option<Cow<'static, str>>,
  pub params: HashMap<Cow<'static, str>, Value>,

The value of the field will automatically be added to the params with a key of value.

Note that validator works in conjunction with serde: in the example we can see that the first_name field is renamed from/to firstName. Any error on that field will be in the firstName key of the hashmap, not first_name.

The other two ValidationErrorsKind types represent errors discovered in nested (vectors of) structs, as described in this example:

use serde::Deserialize;
// A trait that the Validate derive will impl
use validator::Validate;

#[derive(Debug, Validate, Deserialize)]
struct SignupData {
   contact_details: ContactDetails,
   preferences: Vec<Preference>,
   allow_cookies: Option<bool>,

#[derive(Debug, Validate, Deserialize)]
struct ContactDetails {
   mail: String,

#[derive(Debug, Validate, Deserialize)]
struct Preference {
   #[validate(length(min = 4))]
   name: String,
   value: bool,

match signup_data.validate() {
 Ok(_) => (),
 Err(e) => return e;

Here, the ContactDetails and Preference structs are nested within the parent SignupData struct. Because these child types also derive Validate, the fields where they appear can be tagged for inclusion in the parent struct's validation method.

Any errors found in a single nested struct (the contact_details field in this example) would be returned as a Struct(Box<ValidationErrors>) type in the parent's ValidationErrors result.

Any errors found in a vector of nested structs (the preferences field in this example) would be returned as a List(BTreeMap<usize, Box<ValidationErrors>>) type in the parent's ValidationErrors result, where the map is keyed on the index of invalid vector entries.


You will need to import the Validate trait.

The validator crate can also be used without the custom derive as it exposes all the validation functions and types.


The crate comes with some built-in validators and you can have several validators for a given field.


Tests whether the String is a valid email according to the HTML5 regex, which means it will mark some esoteric emails as invalid that won't be valid in a email input as well. This validator doesn't take any arguments: #[validate(email)].


Tests whether the String is a valid URL. This validator doesn't take any arguments: #[validate(url)];


Tests whether a String or a Vec match the length requirement given. length has 3 integer arguments:

  • min
  • max
  • equal

Using equal excludes the min or max and will result in a compilation error if they are found.

At least one argument is required with a maximum of 2 (having min and max at the same time).


const MIN_CONST: u64 = 1;
const MAX_CONST: u64 = 10;

#[validate(length(min = 1, max = 10))]
#[validate(length(min = 1))]
#[validate(length(max = 10))]
#[validate(length(equal = 10))]
#[validate(length(min = "MIN_CONST", max = "MAX_CONST"))]


Tests whether a number is in the given range. range takes 1 or 2 arguments, and they can be normal (min and max) or exclusive (exclusive_min, exclusive_max, unreachable limits). These can be a number or a value path.


const MAX_CONSTANT: i32 = 10;
const MIN_CONSTANT: i32 = 0;

#[validate(range(min = 1))]
#[validate(range(min = "MIN_CONSTANT"))]
#[validate(range(min = 1, max = 10))]
#[validate(range(min = 1.1, max = 10.8))]
#[validate(range(max = 10.8))]
#[validate(range(min = "MAX_CONSTANT"))]
#[validate(range(min = "crate::MAX_CONSTANT"))]
#[validate(range(exclusive_min = 0.0, max = 100.0))]
#[validate(range(exclusive_max = 10))]


Tests whether the 2 fields are equal. must_match takes 1 string argument. It will error if the field mentioned is missing or has a different type than the field the attribute is on.


#[validate(must_match = "password2")]
#[validate(must_match(other = "password2"))]


Tests whether the string contains the substring given or if a key is present in a hashmap. contains takes 1 string argument.


#[validate(contains = "gmail")]
#[validate(contains(pattern = "gmail"))]


Pretty much the opposite of contains, provided just for ease-of-use. Tests whether a container does not contain the substring given if it's a string or if a key is NOT present in a hashmap. does_not_contain takes 1 string argument.


#[validate(does_not_contain = "gmail")]
#[validate(does_not_contain(pattern = "gmail"))]


Tests whether the string matches the regex given. regex takes 1 string argument: the path to a static Regex instance.


use once_cell::sync::Lazy;

static RE_TWO_CHARS: Lazy<Regex> = Lazy::new(|| {

#[validate(regex = "RE_TWO_CHARS")]
#[validate(regex(path = "RE_TWO_CHARS"))]


Test whether the string is a valid credit card number.




Calls one of your functions to perform a custom validation. The field reference will be given as a parameter to the function, which should return a Result<(), ValidationError>.


#[validate(custom(function = "validate_something"))]
#[validate(custom(function = "::utils::validate_something"))]

You can also parse arguments from the validation function to your custom validation by setting the arg parameter. arg can only be set to one type but you can set it to a tuple to pass multiple types at once. Defining the arg parameter will implement the ValidateArgs trait with the corresponding function types like this:

use validator::{Validate, ValidateArgs, ValidationError};

fn validate(value: &str, arg: (i64, i64)) -> Result<(), ValidationError> {

#[derive(Debug, Validate)]
struct TestStruct {
    #[validate(custom(function = "validate", arg = "(i64, i64)"))]
    value: String,

let test_struct: TestStruct = [...]
test_struct.validate_args((77, 555)).is_ok();

It is also possible to pass references by using the lifetime 'v_a note that this lifetime should only be used for the function parameters like this:

fn validate_value(_: &str, arg: &mut Database) -> Result<(), ValidationError> {

#[derive(Debug, Validate)]
struct TestStruct {
    //                                                     vvvv This is the lifetime for references
    #[validate(custom(function = "validate_value", arg = "&'v_a mut Database"))]
    value: String,

let mut database: Database = [...]
let test_struct: TestStruct = [...]
test_struct.validate_args(&mut database).is_ok();

Custom validation with arguments doesn't work on nested validation. See validator_derive_tests/tests/custom.rs and validator_derive_tests/tests/custom_args.rs for more examples.


Performs validation on a field with a type that also implements the Validate trait (or a vector of such types).




Tests whether the String has any utf-8 control characters, fails validation if it does. To use this validator, you must enable the unic feature for the validator crate. This validator doesn't take any arguments: #[validate(non_control_character)];


Tests whether the Option<T> field is Some;

Struct level validation

Often, some error validation can only be applied when looking at the full struct, here's how it works here:

#[derive(Debug, Validate, Deserialize)]
#[validate(schema(function = "validate_category", skip_on_field_errors = false))]
struct CategoryData {
    category: String,
    name: String,

The function mentioned should return a Result<(), ValidationError> and will be called after validation is done for all fields.

The skip_on_field_errors defaults to true if not present and will ensure that the function is not called if an error happened while validating the struct fields.

Any error on the struct level validation will appear in the key __all__ of the hashmap of errors.

Message and code

Each validator can take 2 optional arguments in addition to their own arguments:

  • message: a message to go with the error, for example if you want to do i18n
  • code: each validator has a default error code (for example the regex validator code is regex) but it can be overridden if necessary, mainly needed for the custom validator

Note that these arguments can't be applied to nested validation calls with #[validate].

For example, the following attributes all work:

// code attribute
#[validate(email(code = "code_str"))]
#[validate(credit_card(code = "code_str"))]
#[validate(length(min = 5, max = 10, code = "code_str"))]

#[validate(regex(path = "static_regex", code = "code_str"))]
#[validate(custom(function = "custom_fn", code = "code_str"))]
#[validate(contains(pattern = "pattern_str", code = "code_str"))]
#[validate(does_not_contain(pattern = "pattern_str", code = "code_str"))]
#[validate(must_match(other = "match_value", code = "code_str"))]

// message attribute
#[validate(url(message = "message_str"))]
#[validate(length(min = 5, max = 10, message = "message_str"))]

#[validate(regex(path = "static_regex", message = "message_str"))]
#[validate(custom(function = "custom_fn", message = "message_str"))]
#[validate(contains(pattern = "pattern_str", message = "message_str"))]
#[validate(does_not_contain(pattern = "pattern_str", message = "message_str"))]
#[validate(must_match(other = "match_value", message = "message_str"))]

// both attributes
#[validate(url(message = "message", code = "code_str"))]
#[validate(email(code = "code_str", message = "message"))]
#[validate(custom(function = "custom_fn", code = "code_str", message = "message_str"))]


~26K SLoC