Attributes
You can add attributes to your types to customize Schemars's derived JsonSchema implementation.
Serde allows setting #[serde(...)] attributes which change how types are serialized, and Schemars will generally respect these attributes to ensure that generated schemas will match how the type is serialized by serde_json. #[serde(...)] attributes can be overriden using #[schemars(...)] attributes, which behave identically (e.g. #[schemars(rename_all = "camelCase")]). You may find this useful if you want to change the generated schema without affecting Serde's behaviour, or if you're just not using Serde.
Validator allows setting #[validate(...)] attributes to restrict valid values of particular fields, many of which will be used by Schemars to generate more accurate schemas. These can also be overridden by #[schemars(...)] attributes.
TABLE OF CONTENTS
Supported Serde Attributes
#[serde(rename = "name")] / #[schemars(rename = "name")]
Set on a struct, enum, field or variant to use the given name in the generated schema instead of the Rust name. When used on a struct or enum, the given name will be used as the title for root schemas, and the key within the root's definitions property for subschemas.
If set on a struct or enum with generic type parameters, then the given name may contain them enclosed in curly braces (e.g. {T}) and they will be replaced with the concrete type names when the schema is generated.
Serde docs: container / variant / field
#[serde(rename_all = "...")] / #[schemars(rename_all = "...")]
Set on a struct, enum or variant to rename all fields according to the given case convention (see the Serde docs for details).
Serde docs: container / variant
#[serde(tag = "type")] / #[schemars(tag = "type")]
#[serde(tag = "t", content = "c")] / #[schemars(tag = "t", content = "c")]
#[serde(untagged)] / #[schemars(untagged)]
Set on an enum to generate the schema for the internally tagged, adjacently tagged, or untagged representation of this enum.
Serde docs: tag / tag+content / untagged
#[serde(default)] / #[schemars(default)] / #[serde(default = "path")] / #[schemars(default = "path")]
Set on a struct or field to give fields a default value, which excludes them from the schema's required properties. The default will also be set on the field's schema's default property, unless it is skipped by a skip_serializing_if attribute on the field. Any serialize_with or with attribute set on the field will be used to serialize the default value.
#[serde(skip)] / #[schemars(skip)]
Set on a variant or field to prevent it from appearing in any generated schema.
#[serde(skip_serializing)] / #[schemars(skip_serializing)]
Set on a field of a (non-tuple) struct to set the writeOnly property on that field's schema. Serde also allows this attribute on variants or tuple struct fields, but this will have no effect on generated schemas.
Serde docs: field
#[serde(skip_deserializing)] / #[schemars(skip_deserializing)]
Set on a variant or field. When set on a field of a (non-tuple) struct, that field's schema will have the readOnly property set. When set on a variant or tuple struct field Schemars will treat this the same as a skip attribute.
#[serde(flatten)] / #[schemars(flatten)]
Set on a field to include that field's contents as though they belonged to the field's container.
Serde docs: field
#[serde(with = "Type")] / #[schemars(with = "Type")]
Set on a variant or field to generate its schema as the given type instead of its actual type. Serde allows the with attribute to refer to any module path, but Schemars requires this to be an actual type which implements JsonSchema.
If the given type has any required generic type parameters, then they must all be explicitly specified in this attribute. Serde frequently allows you to omit them as it can make use of type inference, but unfortunately this is not possible with Schemars. For example, with = "Vec::<i32>" will work, but with = "Vec" and with = "Vec::<_>" will not.
#[serde(deny_unknown_fields)] / #[schemars(deny_unknown_fields)]
Setting this on a container will set the additionalProperties keyword on generated schemas to false to show that any extra properties are explicitly disallowed.
Serde docs: container
#[serde(transparent)] / #[schemars(transparent)]
Set on a newtype struct or a braced struct with one field to make the struct's generated schema exactly the same as that of the single field's.
Serde docs: container
#[schemars(bound = "...")]
Where-clause for the JsonSchema impl. This replaces any trait bounds inferred by schemars. Schemars does not use trait bounds from #[serde(bound)] attributes.
Serde docs: container
Supported Validator Attributes
#[validate(email)] / #[schemars(email)]
#[validate(phone)] / #[schemars(phone)]
#[validate(url)] / #[schemars(url)]
Sets the schema's format to email/phone/uri, as appropriate. Only one of these attributes may be present on a single field.
Validator docs: email / phone / url
#[validate(length(min = 1, max = 10))] / #[schemars(length(min = 1, max = 10))]
#[validate(length(equal = 10))] / #[schemars(length(equal = 10))]
Sets the minLength/maxLength properties for string schemas, or the minItems/maxItems properties for array schemas.
Validator docs: length
#[validate(range(min = 1, max = 10))] / #[schemars(range(min = 1, max = 10))]
Sets the minimum/maximum properties for number schemas.
Validator docs: range
#[validate(regex = "path::to::regex")] / #[schemars(regex = "path::to::regex")]
#[schemars(regex(pattern = r"^\d+$"))]
Sets the pattern property for string schemas. The path::to::regex will typically refer to a Regex instance, but Schemars allows it to be any value with a to_string() method.
Providing an inline regex pattern using regex(pattern = ...) is a Schemars extension, and not currently supported by the Validator crate. When using this form, you may want to use a r"raw string literal" so that \\ characters in the regex pattern are not interpreted as escape sequences in the string.
Validator docs: regex
#[validate(contains = "string")] / #[schemars(contains = "string")]
For string schemas, sets the pattern property to the given value, with any regex special characters escaped. For object schemas (e.g. when the attribute is set on a HashMap field), includes the value in the required property, indicating that the map must contain it as a key.
Validator docs: contains
#[validate(required)] / #[schemars(required)]
#[validate(required_nested)]
When set on an Option<T> field, this will create a schemas as though the field were a T.
Validator docs: required / required_nested
Other Attributes
#[schemars(schema_with = "some::function")]
Set on a variant or field to generate this field's schema using the given function. This function must be callable as fn(&mut schemars::gen::SchemaGenerator) -> schemars::schema::Schema.
#[schemars(title = "Some title", description = "Some description")]
Set on a container, variant or field to set the generated schema's title and/or description. If present, these will be used instead of values from any doc comments/attributes.
#[schemars(example = "some::function")]
Set on a container, variant or field to include the result of the given function in the generated schema's examples. The function should take no parameters and can return any type that implements serde's Serialize trait - it does not need to return the same type as the attached struct/field. This attribute can be repeated to specify multiple examples.
#[deprecated]
Set the Rust built-in deprecated attribute on a struct, enum, field or variant to set the generated schema's deprecated keyword to true.
#[schemars(crate = "other_crate::schemars")]
Set the path to the schemars crate instance the generated code should depend on. This is mostly useful for other crates that depend on schemars in their macros.
#[schemars(inner(...))]
Sets properties specified by validator attributes on items of an array schema. For example:
struct Struct {
#[schemars(inner(url, regex(pattern = "^https://")))]
urls: Vec<String>,
}
Doc Comments (#[doc = "..."])
If a struct, variant or field has any doc comments (or doc attributes), then these will be used as the generated schema's description. If the first line is an ATX-style markdown heading (i.e. it begins with a # character), then it will be used as the schema's title, and the remaining lines will be the description.