# JSON Schema

In a simple way, JSON Schema is an object with validation keywords.

The keywords and their values define what rules the data should satisfy to be valid.

# JSON Schema versions

# draft-07 default

This version is provided as default export:

# draft-2019-09 NEW

Ajv supports all new keywords of JSON Schema draft-2019-09:

• unevaluatedProperties
• unevaluatedItems
• dependentRequired
• dependentSchemas
• maxContains/minContains
• $recursiveAnchor/$recursiveRef

To use draft-2019-09 schemas you need to import a different Ajv class:

You can use draft-07 schemas with this Ajv instance as well, draft-2019-09 is backwards compatible. If your schemas use $schema keyword, you need to add draft-07 meta-schema to Ajv instance:

# draft-2020-12 BREAKING

Ajv supports all keywords of JSON Schema draft-2020-12:

• prefixItems that replaced array form of items keyword
• changed items keyword that combined parts of functionality of items and additionalItems
• $dynamicAnchor/$dynamicRef

To use draft-2020-12 schemas you need to import a different Ajv class:

# draft-06

You can use JSON Schema draft-06 schemas with Ajv v7/8. If your schemas use $schema keyword, you need to add draft-06 meta-schema to Ajv instance. This example shows how to support both draft-06 and draft-07 schemas:

# draft-04

You can use JSON Schema draft-04 schemas with Ajv from v8.5.0 and the additional package ajv-draft-04 (opens new window) (both ajv and ajv-draft-04 should be installed).

# OpenAPI support

Ajv supports these additional OpenAPI specification (opens new window) keywords:

• nullable - to avoid using type keyword with array of types.
• discriminator - to optimize validation and error reporting of tagged unions

# JSON data type

# type

type keyword requires that the data is of certain type (or some of types). Its value can be a string (the allowed type) or an array of strings (multiple allowed types).

Type can be: number , integer , string , boolean , array , object or null .

Examples

1. schema : {type: "number"}

   valid : 1 , 1.5

   invalid : "abc" , "1" , [] , {} , null , true

2. schema : {type: "integer"}

   valid : 1 , 2

   invalid : "abc" , "1" , 1.5 , [] , {} , null , true

3. schema : {type: ["number", "string"]}

   valid : 1 , 1.5 , "abc" , "1"

   invalid : [] , {} , null , true

All examples above are JSON Schemas that only require data to be of certain type to be valid.

Most other keywords apply only to a particular type of data. If the data is of different type, the keyword will not apply and the data will be considered valid.

In v7 Ajv introduced Strict types mode that makes these mistakes less likely by requiring that types are constrained with type keyword whenever another keyword that applies to specific type is used.

# nullable OpenAPI

This keyword can be used to allow null value in addition to the defined type .

Ajv supports it by default, without additional options. These two schemas are equivalent, but the first one is better supported by some tools and is also compatible with strictTypes option (see Strict types )

{
  "type": "string",
  "nullable": true

and

{
  "type": ["string", "null"]

# Keywords for numbers

# maximum / minimum and exclusiveMaximum / exclusiveMinimum

The value of keyword maximum ( minimum ) should be a number. This value is the maximum (minimum) allowed value for the data to be valid.

The value of keyword exclusiveMaximum ( exclusiveMinimum ) should be a number. This value is the exclusive maximum (minimum) allowed value for the data to be valid (the data equal to this keyword value is invalid).

Examples

1. schema : {type: "number", maximum: 5}

   valid : 4 , 5

   invalid : 6 , 7

2. schema : {type: "number", minimum: 5}

   valid : 5 , 6

   invalid : 4 , 4.5

3. schema : {type: "number", exclusiveMinimum: 5}

   valid : 6 , 7

   invalid : 4.5 , 5

# multipleOf

The value of the keyword should be a number. The data to be valid should be a multiple of the keyword value (i.e. the result of division of the data on the value should be integer).

Examples

1. schema : {type: "number", multipleOf: 5}

   valid : 5 , 10

   invalid : 1 , 4

2. schema : {type: "number", multipleOf: 2.5}

   valid : 2.5 , 5 , 7.5

   invalid : 1 , 4

# Keywords for strings

# maxLength / minLength

The value of the keywords should be a number. The data to be valid should have length satisfying this rule. Unicode pairs are counted as a single character.

Examples

1. schema : {type: "string", maxLength: 5}

   valid : "abc" , "abcde"

   invalid : "abcdef"

2. schema : {type: "string", minLength: 2}

   valid : "ab" , "😀😀"

   invalid : "a" , "😀"

# pattern

The value of the keyword should be a string. The data to be valid should match the regular expression defined by the keyword value.

Ajv uses new RegExp(value, "u") to create the regular expression that will be used to test data.

Example

schema : {type: "string", pattern: "[abc]+"}

valid : "a" , "abcd" , "cde"

invalid : "def" , ""

# format

The value of the keyword should be a string. The data to be valid should match the format with this name.

Ajv does not include any formats, they can be added with ajv-formats (opens new window) plugin.

Example

schema : {type: "string", format: "ipv4"}

valid : "192.168.0.1"

invalid : "abc"

# Keywords for arrays

# maxItems / minItems

The value of the keywords should be a number. The data array to be valid should not have more (less) items than the keyword value.

Example

schema : {type: "array", maxItems: 3}

valid : [] , [1] , ["1", 2, "3"]

invalid : [1, 2, 3, 4]

# uniqueItems

The value of the keyword should be a boolean. If the keyword value is true , the data array to be valid should have unique items.

Example

schema : {type: "array", uniqueItems: true}

valid : [] , [1] , ["1", 2, "3"]

invalid : [1, 2, 1] , [{a: 1, b: 2}, {b: 2, a: 1}]

# items

# items in draft-04, -06, -07 and -2019-09

The value of the keyword should be a schema or an array of schemas.

If the keyword value is a schema, then for the data array to be valid each item of the array should be valid according to the schema. In this case the additionalItems keyword is ignored.

If the keyword value is an array, then items with indices less than the number of items in the keyword should be valid according to the schemas with the same indices. Whether additional items are valid will depend on additionalItems keyword.

Examples

1. schema : {type: "array", items: {type: "integer"}}

   valid : [1,2,3] , []

   invalid : [1,"abc"]

2. schema :

   {
     type: "array",
     items: [{type: "integer"}, {type: "string"}]
   

   valid : [1] , [1, "abc"] , [1, "abc", 2] , []

   invalid : ["abc", 1] , ["abc"]

The schema in example 2 will log warning by default (see strictTuples option), because it defines unconstrained tuple. To define a tuple with exactly 2 elements use minItems and additionalItems keywords (see example 1 in additionalItems ).

# items in draft-2020-12 NEW

The value of the keyword must be a schema.

For the data array to be valid:

• if prefixItems keyword is not used in the schema, then each item of the array must be valid according to the schema in items .
• if prefixItems keyword is used in the schema, then each item with the index starting from the size of prefixItems schema must be valid according to the schema in items

Examples

1. schema : {type: "array", items: {type: "integer"}}

   valid : [1,2,3] , []

   invalid : [1,"abc"]

2. schema :

   {
     type: "array",
     prefixItems: [{type: "integer"}, {type: "integer"}],
     minItems: 2
     items: false
   

   valid : [1, 2]

   invalid : [] , [1] , [1, 2, 3] , [1, "abc"] (any wrong number of items or wrong type)

3. schema :

   {
     type: "array",
     prefixItems: [{type: "integer"}, {type: "integer"}],
     items: {type: "string"}
   

   valid : [] , [1, 2] , [1, 2, "abc"]

   invalid : ["abc"] , [1, 2, 3]

   
   _valid_: `[1]`, `[1, "abc"]`, `[1, "abc", 2]`, `[]`
   _invalid_: `["abc", 1]`, `["abc"]`
   

The schema in example 3 will log warning by default (see strictTuples option), because it defines unconstrained tuple. To define a tuple with exactly 2 elements use minItems and items keywords (see example 2).

# prefixItems NEW: draft 2020-12

The value of the keyword must be an array of schemas.

For the data array to be valid, the items with indices less than the number of schemas in this keyword must be valid according to the schemas with the same indices. Whether additional items are valid will depend on items keyword.

Examples

schema :

{
  type: "array",
  prefixItems: [{type: "integer"}, {type: "string"}]

valid : [1] , [1, "abc"] , [1, "abc", 2] , []

invalid : ["abc", 1] , ["abc"]

The schema in example will log warning by default (see strictTuples option), because it defines unconstrained tuple. To define a tuple with exactly 2 elements use minItems and items keywords (see example 2 in items ).

# additionalItems

The value of the keyword should be a boolean or an object.

additionalItems keyword is ignored if items keyword is not present or is an object. By default Ajv will throw exception in this case - see Strict mode

additionalItems keyword is ignored if items keyword has more elements than data array.

If the data array has more elements than the items keyword value then the result of the validation depends on the value of additionalItems keyword:

• false : data is invalid
• true : data is valid
• an object: data is valid if all additional items (i.e. items with indices greater or equal than "items" keyword value length) are valid according to the schema in "additionalItems" keyword.

The schemas in examples 2-3 log warning by default, use option strictTuples: false to allow)

Examples

1. schema :

   {
     type: "array",
     items: [{type: "integer"}, {type: "integer"}],
     minItems: 2
     additionalItems: false
   

   valid : [1, 2]

   invalid : [] , [1] , [1, 2, 3] , [1, "abc"] (any wrong number of items or wrong type)

2. schema :

   {
     type: "array",
     items: [{type: "integer"}, {type: "integer"}],
     additionalItems: true
   

   valid : [] , [1, 2] , [1, 2, 3] , [1, 2, "abc"]

   invalid : ["abc"] , [1, "abc", 3]

3. schema :

   {
     type: "array",
     items: [{type: "integer"}, {type: "integer"}],
     additionalItems: {type: "string"}
   

   valid : [] , [1, 2] , [1, 2, "abc"]

   invalid : ["abc"] , [1, 2, 3]

# contains

The value of the keyword is a JSON Schema. The array is valid if it contains at least one item that is valid according to this schema.

Example

schema : {type: "array", contains: {type: "integer"}}

valid : [1] , [1, "foo"] , any array with at least one integer

invalid : [] , ["foo", "bar"] , any array without integers

# maxContains / minContains NEW: draft 2019-09

The value of these keywords should be an integer.

Without contains keyword they are ignored (logs error or throws exception in ajv strict mode ).

The array is valid if it contains at least minContains items and no more than maxContains items that are valid against the schema in contains keyword.

Example

schema :

{
  type: "array",
  contains: {type: "integer"},
  minContains: 2,
  maxContains: 3

valid : [1, 2] , [1, 2, 3, "foo"] , any array with 2 or 3 integers

invalid : [] , [1, "foo"] , [1, 2, 3, 4] , any array with fewer than 2 or more than 3 integers

# unevaluatedItems NEW: draft 2019-09

The value of this keyword is a JSON Schema (can be a boolean).

This schema will be applied to all array items that were not evaluated by other keywords for items ( items , additionalItems and contains ) in the current schema and all sub-schemas that were valid for this data instance. It includes:

• all subschemas schemas in allOf and $ref keywords
• valid sub-schemas in oneOf and anyOf keywords
• sub-schema in if keyword
• sub-schemas in then or else keywords that were applied based on the validation result by if keyword.

The only scenario when this keyword would be applied to some items is when items keyword value is an array of schemas and additionalItems was not present (or did not apply, in case it was present in some invalid subschema).

Some user-defined keywords can also make items "evaluated".

Example

schema :

{
  type: "array",
  items: [
    {type: "number"},
    {type: "number"}
  unevaluatedItems: false,
  anyOf: [
    {items: [true, true, {type: "number"}]},
    {items: [true, true, {type: "boolean"}]}

valid : [1, 2, 3] , [1, 2, true]

invalid :

• [1, 2] - the third item is not present
• [1, 2, "3"] - the third item is "unevaluated"

See tests (opens new window) for unevaluatedItems keyword for other examples.

# Keywords for objects

# maxProperties / minProperties

The value of the keywords should be a number. The data object to be valid should have not more (less) properties than the keyword value.

Example

schema : {type: "object", maxProperties: 2 }

valid : {} , {a: 1} , {a: "1", b: 2}

invalid : {a: 1, b: 2, c: 3}

# required

The value of the keyword should be an array of unique strings. The data object to be valid should contain all properties with names equal to the elements in the keyword value.

Example

schema : {type: "object", required: ["a", "b"]}

valid : {a: 1, b: 2} , {a: 1, b: 2, c: 3}

invalid : {} , {a: 1} , {c: 3, d: 4}

# properties

The value of the keyword should be a map with keys equal to data object properties. Each value in the map should be a JSON Schema. For data object to be valid the corresponding values in data object properties should be valid according to these schemas.

Example

schema :

{
  type: "object",
  properties: {
    foo: {type: "string"},
    bar: {
      type: "number",
      minimum: 2

valid : {} , {foo: "a"} , {foo: "a", bar: 2}

invalid : {foo: 1} , {foo: "a", bar: 1}

# patternProperties

The value of this keyword should be a map where keys should be regular expressions and the values should be JSON Schemas. For data object to be valid the values in data object properties that match regular expression(s) should be valid according to the corresponding schema(s).

When the value in data object property matches multiple regular expressions it should be valid according to all the schemas for all matched regular expressions.

Example

schema :

{
  type: "object",
  patternProperties: {
    "^fo.*$": {type: "string"},
    "^ba.*$": {type: "number"}

valid : {} , {foo: "a"} , {foo: "a", bar: 1}

invalid : {foo: 1} , {foo: "a", bar: "b"}

# additionalProperties

The value of the keyword should be either a boolean or a JSON Schema.

If the value is true the keyword is ignored.

If the value is false the data object to be valid should not have "additional properties" (i.e. properties other than those used in "properties" keyword and those that match patterns in "patternProperties" keyword).

If the value is a schema for the data object to be valid the values in all "additional properties" should be valid according to this schema.

Examples

1. schema :

   {
     type: "object",
     properties: {
       foo: {type: "number"}
     patternProperties: {
       "^.*r$": {type: "number"}
     additionalProperties: false
   

   valid : {} , {foo: 1} , {foo: 1, bar: 2}

   invalid : {a: 3} , {foo: 1, baz: 3}

2. schema :

   {
     type: "object",
     properties: {
       foo: {type: "number"}
     patternProperties: {
       "^.*r$": {type: "number"}
     additionalProperties: {type: "string"}
   

   valid : {} , {a: "b"} , {foo: 1} , {foo: 1, bar: 2} , {foo: 1, bar: 2, a: "b"}

   invalid : {a: 3} , {foo: 1, baz: 3}

3. schema :

   {
     type: "object",
     properties: {
       foo: {type: "number"}
     additionalProperties: false,
     anyOf: [
         properties: {
           bar: {type: "number"}
         properties: {
           baz: {type: "number"}
   

   valid : {} , {foo: 1}

   invalid : {bar: 2} , {baz: 3} , {foo: 1, bar: 2} , etc.

# dependencies deprecated in draft 2019-09

This keyword is deprecated. The same functionality is available with keywords dependentRequired and dependentSchemas .

The value of the keyword is a map with keys equal to data object properties. Each value in the map should be either an array of unique property names ("property dependency" - see dependentRequired keyword) or a JSON Schema ("schema dependency" - see dependentSchemas keyword).

For property dependency, if the data object contains a property that is a key in the keyword value, then to be valid the data object should also contain all properties from the array of properties.

For schema dependency, if the data object contains a property that is a key in the keyword value, then to be valid the data object itself (NOT the property value) should be valid according to the schema.

Examples

1. schema (property dependency) :

   {
     type: "object",
     dependencies: {
       foo: ["bar", "baz"]
   

   valid : {foo: 1, bar: 2, baz: 3} , {} , {a: 1}

   invalid : {foo: 1} , {foo: 1, bar: 2} , {foo: 1, baz: 3}

2. schema (schema dependency) :

   {
     type: "object",
     dependencies: {
       foo: {
         properties: {
           bar: {type: "number"}
   

   valid : {} , {foo: 1} , {foo: 1, bar: 2} , {a: 1}

   invalid : {foo: 1, bar: "a"}

# dependentRequired NEW: draft 2019-09

The value of this keyword should be a map with keys equal to data object properties. Each value in the map should be an array of unique property names.

If the data object contains a property that is a key in the keyword value, then to be valid the data object should also contain all properties from the corresponding array of properties in this keyword.

Example

schema :

{
  type: "object",
  dependentRequired: {
    foo: ["bar", "baz"]

valid : {foo: 1, bar: 2, baz: 3} , {} , {a: 1}

invalid : {foo: 1} , {foo: 1, bar: 2} , {foo: 1, baz: 3}

# dependentSchemas NEW: draft 2019-09

The value of the keyword should be a map with keys equal to data object properties. Each value in the map should be a JSON Schema.

If the data object contains a property that is a key in the keyword value, then to be valid the data object itself (NOT the property value) should be valid according to the corresponding schema in this keyword.

Example

schema :

{
  type: "object",
  dependentSchemas: {
    foo: {
      properties: {
        bar: {type: "number"}

valid : {} , {foo: 1} , {foo: 1, bar: 2} , {a: 1}

invalid : {foo: 1, bar: "a"}

# propertyNames

The value of this keyword is a JSON Schema.

For data object to be valid each property name in this object should be valid according to this schema.

Example

schema (requires email format from ajv-formats (opens new window) ):

{
  type: "object",
  propertyNames: {
    format: "email"

valid : {"[email protected]": "any", "[email protected]": "any"}

invalid : {foo: "any value"}

# unevaluatedProperties NEW: draft 2019-09

The value of this keyword is a JSON Schema (can be a boolean).

This schema will be applied to all properties that were not evaluated by other keywords for properties ( properties , patternProperties and additionalProperties ) in the current schema and all sub-schemas that were valid for this data instance. It includes:

• all subschemas schemas in allOf and $ref keywords
• valid sub-schemas in oneOf and anyOf keywords
• sub-schema in if keyword
• sub-schemas in then or else keywords that were applied based on the validation result by if keyword.

Some user-defined keywords can also make properties "evaluated".

Example

schema :

{
  type: "object",
  required: ["foo"],
  properties: {foo: {type: "number"}},
  unevaluatedProperties: false,
  anyOf: [
      required: ["bar"],
      properties: {bar: {type: "number"}}
      required: ["baz"],
      properties: {baz: {type: "number"}}

valid : {foo: 1, bar: 2} , {foo: 1, baz: 2} , {foo: 1, bar: 2, baz: 3}

invalid :

• {foo: 1} - neither bar nor baz are present
• {foo: 1, bar: 2, boo: 3} - boo is unevaluated
• {foo: 1, bar: 2, baz: "3"} - not valid against the 2nd subschema, so baz is "unevaluated".

See tests (opens new window) for unevaluatedProperties keyword for other examples.

# discriminator NEW: OpenAPI

Ajv has a limited support for discriminator keyword: to optimize validation, error handling, and modifying data with oneOf keyword.

Its value should be an object with a property propertyName - the name of the property used to discriminate between union members.

When using discriminator keyword only one subschema in oneOf will be used, determined by the value of discriminator property.

Example

schema :

{
  type: "object",
  discriminator: {propertyName: "foo"},
  required: ["foo"],
  oneOf: [
      properties: {
        foo: {const: "x"},
        a: {type: "string"},
      required: ["a"],
      properties: {
        foo: {enum: ["y", "z"]},
        b: {type: "string"},
      required: ["b"],

valid : {foo: "x", a: "any"} , {foo: "y", b: "any"} , {foo: "z", b: "any"}

invalid :

• {} , {foo: 1} - discriminator tag must be string
• {foo: "bar"} - discriminator tag value must be in oneOf subschema
• {foo: "x", b: "b"} , {foo: "y", a: "a"} - invalid object

From the perspective of validation result discriminator is defined as no-op (that is, removing discriminator will not change the validity of the data), but errors reported in case of invalid data will be different.

There are following requirements and limitations of using discriminator keyword:

• mapping in discriminator object is not supported.
• oneOf keyword must be present in the same schema.
• discriminator property should be required either on the top level, as in the example, or in all oneOf subschemas.
• each oneOf subschema must have properties keyword with discriminator property. The subschemas should be either inlined or included as direct references (only $ref keyword without any extra keywords is allowed).
• schema for discriminator property in each oneOf subschema must be const or enum , with unique values across all subschemas.

Not meeting any of these requirements would fail schema compilation.

# Keywords for all types

# enum

The value of the keyword should be an array of unique items of any types. The data is valid if it is deeply equal to one of items in the array.

Example

schema : {enum: [2, "foo", {foo: "bar" }, [1, 2, 3]]}

valid : 2 , "foo" , {foo: "bar"} , [1, 2, 3]

invalid : 1 , "bar" , {foo: "baz"} , [1, 2, 3, 4] , any value not in enum

# const

The value of this keyword can be anything. The data is valid if it is deeply equal to the value of the keyword.

Example

schema : {const: "foo"}

valid : "foo"

invalid : any other value

The same can be achieved with enum keyword using the array with one item. But const keyword is more than just a syntax sugar for enum . In combination with the $data reference it allows to define equality relations between different parts of the data. This cannot be achieved with enum keyword even with $data reference because $data cannot be used in place of one item - it can only be used in place of the whole array in enum keyword.

Example

schema :

{
  type: "object",
  properties: {
    foo: {type: "number"},
    bar: {const: {$data: "1/foo"}}

valid : {foo: 1, bar: 1} , {}

invalid : {foo: 1} , {bar: 1} , {foo: 1, bar: 2}

# Compound keywords

# not

The value of the keyword should be a JSON Schema. The data is valid if it is invalid according to this schema.

Example

schema : {type: "number", not: {minimum: 3}}

valid : 1 , 2

invalid : 3 , 4

# oneOf

The value of the keyword should be an array of JSON Schemas. The data is valid if it matches exactly one JSON Schema from this array. Validators have to validate data against all schemas to establish validity according to this keyword.

Example

schema :

{
  type: "number",
  oneOf: [{maximum: 3}, {type: "integer"}]

valid : 1.5 , 2.5 , 4 , 5

invalid : 2 , 3 , 4.5 , 5.5

# anyOf

The value of the keyword should be an array of JSON Schemas. The data is valid if it is valid according to one or more JSON Schemas in this array. Validators only need to validate data against schemas in order until the first schema matches (or until all schemas have been tried). For this reason validating against this keyword is faster than against "oneOf" keyword in most cases.

Example

schema :

{
  type: "number",
  anyOf: [{maximum: 3}, {type: "integer"}]

valid : 1.5 , 2 , 2.5 , 3 , 4 , 5

invalid : 4.5 , 5.5

# allOf

The value of the keyword should be an array of JSON Schemas. The data is valid if it is valid according to all JSON Schemas in this array.

Example

schema :

{
  type: "number",
  allOf: [{maximum: 3}, {type: "integer"}]

valid : 2 , 3

invalid : 1.5 , 2.5 , 4 , 4.5 , 5 , 5.5

# if / then / else

These keywords allow to implement conditional validation. Their values should be valid JSON Schemas (object or boolean).

If if keyword is absent, the validation succeeds.

If the data is valid against the sub-schema in if keyword, then the validation result is equal to the result of data validation against the sub-schema in then keyword (if then is absent, the validation succeeds).

If the data is invalid against the sub-schema in if keyword, then the validation result is equal to the result of data validation against the sub-schema in else keyword (if else is absent, the validation succeeds).

Examples

1. schema :

   {
     type: "object",
     if: {properties: {foo: {minimum: 10}}},
     then: {required: ["bar"]},
     else: {required: ["baz"]}
   

   valid :

   • {foo: 10, bar: true }
   • {}
   • {foo: 1, baz: true }

   invalid :

   • {foo: 10} ( bar is required)
   • {foo: 10, baz: true } ( bar is required)
   • {foo: 1} ( baz is required)

2. schema :

   {
     type: "integer",
     minimum: 1,
     maximum: 1000,
     if: {minimum: 100},
     then: {multipleOf: 100},
     else: {
       if: {minimum: 10},
       then: {multipleOf: 10}
   

   valid : 1 , 5 , 10 , 20 , 50 , 100 , 200 , 500 , 1000

   invalid :

   • -1 , 0 (<1)
   • 2000 (>1000)
   • 11 , 57 , 123 (any integer with more than one non-zero digit)
   • non-integers

# Metadata keywords

JSON Schema specification defines several metadata keywords that describe the schema itself but do not perform any validation.

• title and description : information about the data represented by that schema
• $comment : information for developers. With option $comment Ajv logs or passes the comment string to the user-supplied function. See Options .
• default : a default value of the data instance, see Assigning defaults .
• examples : an array of data instances. Ajv does not check the validity of these instances against the schema.
• readOnly and writeOnly : marks data-instance as read-only or write-only in relation to the source of the data (database, api, etc.).
• contentEncoding : RFC 2045 (opens new window) , e.g., "base64".
• contentMediaType : RFC 2046 (opens new window) , e.g., "image/png".