JSON schema format
Topics on this page
This topic describes Cedar’s JSON schema format.
Schema format
A schema contains a declaration of one or more namespaces, each of which contains two mandatory JSON objects, entityTypes and actions. A namespace declaration can optionally include a third object, commonTypes, which defines types that can be referenced by the other two objects. We consider the format of namespaces and these three objects next.
Namespace
A namespace declaration contains a comma-separated list of JSON objects within braces { }. The following is an example of a namespace declaration:
A namespace declaration must contain two child elements, and may contain a third, appearing in any order:
entityTypesactionscommonTypes(optional)
You define the types of your application’s principal and resource entities within the entityTypes element, and the specific actions in the actions element. Principals and resources are separated from actions because actions are defined in the schema as individual discrete elements (each of which has type Action), whereas only the principal and resource entities’ types are defined. In your entity store you create individual principal and resource entities that have these types. Optionally, you can define type names in commonTypes and reference those names as types in the entityTypes and actions elements of your schema.
The declared namespace must be an identifier as specified in Cedar syntax, and it cannot contain the reserved __cedar namespace. It is automatically prepended to all types defined within the associated scope. For example, consider the following schema:
{
"ExampleCo::Database": {
"entityTypes": {
"Table": {
...
}
},
"actions": {
"createTable": {
...
}
}
}
}
Here, the schema is effectively defining the action entity ExampleCo::Database::Action::"createTable" and the entity type ExampleCo::Database::Table.
You can reference entity types and actions defined in other namespaces of the same schema by using their fully qualified names. For example, here is a schema that declares two namespaces, ExampleCo::Clients and ExampleCo::Furniture, where the second namespace’s entity type Table references the first’s entity type Manufacturer.
{
"ExampleCo::Clients": {
"entityTypes": {
"Manufacturer": { ... }
},
"actions": { ... }
},
"ExampleCo::Furniture": {
"entityTypes": {
"Table": {
"shape": {
"type": "Record",
"attributes": {
"manufacturer": {
"type": "Entity",
"name": "ExampleCo::Clients::Manufacturer"
}
}
}
}
},
"actions": { ... }
}
}
One special case: you cannot define entity types, actions, or common types in a namespace that would shadow definitions in the empty namespace. For example, if the entity type Table was also defined in the empty namespace, then the schema above would be invalid.
If you change a declared namespace in your schema you will need to change the entity types appearing in your policies and/or in other namespaces declared in your schema to instead reference the changed namespace.
entityTypes
A collection of the principal and resource entity types supported by your application. The entityTypes element contains a comma-separated list of JSON objects.
The high-level structure of an entityTypes entry looks like the following example.
"entityTypes": {
"EntityTypeName1": {
"memberOfTypes": [ "parentGroupTypeName1", "parentGroupTypeName2", … ],
"shape": { … }
},
"EntityTypeName2": {
"memberOfTypes": [ "parentGroupTypeName1", "parentGroupTypeName2", … ],
"shape": { … }
}
}
Each entry in the entityTypes is a JSON object with the following properties.
Entity type name
Specifies the name of the entity type as a string. This type name must be an identifier, which is defined in the Cedar grammar as a sequence of alphanumeric characters, omitting any Cedar reserved words.
The entity type name must be normalized and cannot include any embedded whitespace, such as spaces, newlines, control characters, or comments.
"My::Name::Space": {
"entityTypes": {
"UserGroup": { ... } // New entity type name
}
}
This type name is qualified by its namespace to form a fully qualified entity type which must be used when referencing this type in a policy.
"My::Name::Space::UserGroup"
memberOfTypes
Specifies a list of entity types that can be direct parents of entities of this type. Values in this list must be valid entity type names declared in the schema. If the memberOfTypes element is empty or not defined, then entities of that entity type can’t have any parents in the entity hierarchy. The following example shows that an entity of type User can have parents of type UserGroup.
"entityTypes": {
"UserGroup": {
…
},
"User": {
"memberOfTypes": [ "UserGroup" ],
…
}
}
If the parent type is part of the same namespace as the child type, then you can reference simply the parent type’s name. If the parent type is in a different namespace than the child type, then you must include the parent type’s namespace as part of the reference. The following example references two parent types, both named UserGroup. The first UserGroup is part of the same namespace as the child entity type that this entry is part of. The second UserGroup is defined in the namespace Aws::Cognito::UserPool_1.
"memberOfTypes": [
"UserGroup",
"Aws::Cognito::UserPool_1::UserGroup"
]
shape
Specifies the shape of the data stored in entities of this type. The shape element, if present, must have type Record, and be accompanied by a description of the entity’s attributes. The following example shows a simple specification of the User entity type.
"User" : {
"shape" : {
"type" : "Record",
"attributes" : {
"name" : {
"type" : "String"
},
"age" : {
"type" : "Long"
}
}
}
}
Each attribute in the attributes portion of the shape record must follow the format described next.
Note that if the shape element is omitted, then entities of the type being defined are assumed to have no data stored with them. This is equivalent to specifying a Record with {} for its attributes.
tags
Specifies the tag type for entities of this type. If this is not present, entities of this type are not allowed to have tags. Valid values for the tag type are the same as valid values for attribute types.
Attribute specifications
Each attribute in a Record is a JSON object that describes one attribute in the record associated with entities of this type. It has the form
"name" : {
"type" : "Type"
},
where name is the name of the attribute, and Type is one of the Cedar supported data types, discussed in detail below.
You can choose to specify whether an attribute is required or optional. By default, attributes that you define are required. This means that policies that reference this type can assume that the attribute is always present. You can make an attribute optional by adding "required": false to the attribute description. Here is an example:
"jobLevel": {
"type": "Long",
"required": false
},
You can choose to explicitly declare that an attribute is mandatory by including "required": true (but this is unnecessary as mandatory attributes are the default).
Attribute types
Attributes’ type components can be "String", "Long", "Boolean", "Record", "Set", "Entity", "Extension", "EntityOrCommon", or a common type name. The first three require no further information to be specified, the next five are described below, and the last is described in the commonTypes section.
Record
A record attribute has the same JSON format as the entity shape’s record’s attributes. As an example, the following refactors the User entity specification above to have a features attribute that is a Record containing some of the user’s physical features.
"User" : {
"shape" : {
"type" : "Record",
"attributes" : {
"name" : {
"type" : "String"
},
"features" : {
"type": "Record",
"attributes": {
"age" : {
"type" : "Long"
},
"height": {
"type" : "Long"
},
"eyecolor": {
"type": "String"
}
}
}
}
}
}
Entity
For attributes of type "Entity", you must also specify a name that identifies the entity type of this attribute. The entity type must be defined in the schema. For example, a resource entity might require an Owner element that specifies a User.
"Document" : {
"shape" : {
"type" : "Record",
"attributes" : {
"Owner": {
"type": "Entity",
"name": "User"
}
}
}
}
Set
For attributes with type "Set", you must also specify an element that defines the properties of the members of the set. Each element is a JSON object that describes what each member of the set looks like.
An element must contain a structure formatted according to the same rules as an attribute. As an example, consider the following Admins entry which could be part of the shape record of an Account entity type. This Admins element is a set of entities of type User and could be used to define which users have administrator permissions in the account.
"Group" : {
"shape" : {
"type" : "Record",
"attributes": {
"Admins": {
"type": "Set",
"element": {
"type": "Entity",
"name": "User"
}
}
}
}
}
Extension
For attributes of type "Extension", you must also specify the name of the specific extension type. There are two extension types: ipaddr for IP address values, and decimal for decimal values. For example, a Network entity may include the IP address of its gateway.
"Network": {
"shape": {
"type": "Record",
"attributes": {
"gateway": {
"type": "Extension",
"name": "ipaddr"
}
}
}
}
EntityOrCommon
Attributes with an entity or common type can also be referred to using an alternate "EntityOrCommon" syntax. This syntax more closely matches the Cedar schema format, and is output by the Cedar CLI when converting from the Cedar schema format to JSON. As an example, the following definition for the Document type is equivalent to the one given above in the Entity section if User is defined as an entity type.
"Document" : {
"shape" : {
"type" : "Record",
"attributes" : {
"Owner": {
"type": "EntityOrCommon",
"name": "User"
}
}
}
}
If User is defined as a common type, then Owner has this common type instead. If User is defined as both an entity and common type, the common type definition is preferred, matching the disambiguation rules for the Cedar schema format.
actions
A collection of the Action entities usable as actions in authorization requests submitted by your application. The actions element contains a comma-separated list of one or more JSON objects.
The high-level structure of an actions entry looks like the following.
"actions": {
"ActionName1": {
"memberOf": ["ActionGroupName1",...],
"appliesTo": {
"principalTypes": ["PrincipalEntityType1",...],
"resourceTypes": ["ResourceEntityType1",...],
}
},
"ActionName2": { ... },
...
}
You can add as many actions as your application requires.
Action name
Specifies the identifier for the action entity, as a string. The name of the action isn’t a value but the heading for its own JSON object. Since this is an entity identifier (rather than an entity type, as in the entityTypes section) it can contain anything that would be valid inside a Cedar string.
"actions": {
"ViewPhoto": {
...
},
"ListPhotos": {
...
},
...
}
You can then refer to these actions in your policies by using the following syntax. If the schema declares a namespace, then the entity type Action is qualified by that namespace.
MyApplicationNamespace::Action::"ViewPhoto"
memberOf
Specifies a list of action entity groups the action is a member of. The memberOf component is optional. If omitted, it means the action is a member of no action groups.
The following schema snippet shows an action named viewAlbum that is a member of the action group called viewImages.
"actions": {
"viewAlbum": {
…
"memberOf": [
{
"id": "viewImages",
"type": "PhotoFlash::Images::Action"
},
],
…
}
}
Action groups are themselves actions, and thus must also be defined in the schema in the actions section; we’ll see an example of that below.
appliesTo
Specifies a JSON object containing principalTypes, resourceTypes, and context which contain the principal, resources, and context entity types, respectively, that can accompany the action in an authorization request.
- If either the
principalTypesorresourceTypescomponents is given with an empty list[], the associated action is not permitted in an authorization request with any entities of that category. This effectively means that the action will not be used in an authorization request at all. This makes sense for actions that act as groups for other actions.
The following example actions snippet shows three actions. The first action, read, is an action group for the other two. It cannot appear in an authorization request because its principalTypes and resourceTypes components are []. The second action, viewPhoto, is a member of the read action group, and expects that any request with this action will have a principal entity of type User and a resource entity of type Photo. The third action, listAlbums, also a member of the read group, expects that a request with that action will have a principal entity of type User and a resource entity of type Account. Notice that for both of the latter two actions, the group membership only requires giving the ID of the action – "read" – and not the type. This is because the validator knows that all action groups must have type Action, and by default the action will be within the current namespace. To declare membership in an action group in a different namespace you need to include "type": "My::Namespace::Action" alongside the "id" portion, where My::Namespace is the different namespace.
"actions": {
"read": {
"appliesTo": {
"principalTypes": [],
"resourceTypes": []
}
},
"viewPhoto": {
"memberOf": [
{
"id": "read"
}
],
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Photo" ]
}
},
"listAlbums": {
"memberOf": [
{
"id": "read"
}
],
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Account" ]
}
}
}
context
Specifies a JSON object in the same format as an entity’s shape property, which defines the attributes that can be present in the context record in authorization requests made with this action. Specifying a context enables Cedar to validate policies that attempt to reference the context record’s attributes.
Each context entry consists of type and attributes objects. By default, the type object is of type Record, but can also be of type commonType. The attributes object has the same JSON format as the entity shape’s record’s attributes. For example, the following context snippet specifies that any request to perform the SomeAction action must include a Boolean attribute named field1 and a Longattribute named field2. Optionally, the context may include a third attribute field3 which, if present, is a String. The context entry is optional, and if excluded it is assumed to be an empty Record (in which case policies that try to access attributes in context will produce validation errors).
"actions": {
"SomeAction": {
"appliesTo": {
"principalTypes": [ ... ],
"resourceTypes": [ ... ],
"context": {
"type": "Record",
"attributes": {
"field1": { "type": "Boolean" },
"field2": { "type": "Long" },
"field3": { "type": "String",
"required": false }
}
}
}
}
}
commonTypes
Structure
Each JSON object within commonTypes consists of the name of a type being defined and its associated definition. The definition is specified just like an attribute type specification, i.e.,
"TypeName": {
// attribute type specification
}
Returning to our motivating example, we can define a record type called ReusedContext, which is then referenced by the view and upload actions.
...
"commonTypes": {
"ReusedContext": {
"type": "Record",
"attributes": {
"ip": { "type": "Extension", "name": "ipaddr" },
"is_authenticated": { "type": "Boolean" },
"timestamp": { "type": "Long" }
}
}
},
"actions": {
"view": {
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Photo" ],
"context": { "type": "ReusedContext" }
}
},
"upload": {
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Server" ],
"context": { "type": "ReusedContext" }
}
}
}
When referencing the attributes of a commonType object, you can reference the attributes directly. As an example, take the view action from the above schema. It is of type ReusedContent, which is defined as a commonTypes object. The following is how you would reference the is_authenticated attribute:
context.is_authenticated == "True"
We can also use type names defined in commonTypes within definitions in the entityTypes section. As a simple example, here we define a type name as a String, and then use the type (twice) in the User entity type’s attributes specifications:
...
"commonTypes": {
"name": {
"type": "String",
}
},
"entityTypes": {
"User": {
"shape": {
"type": "Record",
"attributes": {
"firstName": {
"type": "name"
},
"lastName": {
"type": "name"
}
}
}
}
}
As another example, we can use a defined record type for the shape of multiple entity types. In particular, we collect a set of attributes as a record named Person and use Person within the Employee and Customer entity type definitions.
...
"commonTypes": {
"Person": {
"type": "Record",
"attributes": {
"age": {"type": "Long"},
"name": {"type": "String"}
}
}
},
"entityTypes": {
"Employee": { "shape": { "type": "Person" } },
"Customer": { "shape": { "type": "Person" } }
}
If you then send an Employee entity as the principal in an authorization request, you could evaluate the attributes of that principal by using syntax similar to this example: principal.age.
Note that definitions of types appearing in commonTypes can refer to one another as long as they do not constitute cycles. For example, we can declare a common type Name as an alias for primitive type String and use it as the type of Person’s name attribute. However, Cedar schema parser raises an error if Name were declared as an alias of Person since they form a dependency cycle.
...
"commonTypes": {
"Person": {
"type": "Record",
"attributes": {
"age": {"type": "Long"},
"name": {"type": "Name"}
}
},
"Name": { "type": "String"}
},
"entityTypes": {
"Employee": { "shape": { "type": "Person" } },
"Customer": { "shape": { "type": "Person" } }
}
Example schema
The following schema is for a hypothetical application called PhotoFlash.
The schema defines a User entity that can have a department and a jobLevel. The user can be a member of a UserGroup.
The schema also defines the following three resource types:
- An
Accountcan have oneownerand zero or moreadminsthat are allUserentities. - An
Albumcan be nested inside anotherAlbum, and has a Booleanprivateattribute, and a reference to anAccount. - A
Photocan be placed in anAlbum, and also has aprivateattribute and a reference to anAccount.
{
"PhotoFlash": {
"entityTypes": {
"User": {
"memberOfTypes": [ "UserGroup" ],
"shape": {
"type": "Record",
"attributes": {
"department": { "type": "String" },
"jobLevel": { "type": "Long" }
}
}
},
"UserGroup": { },
"Photo": {
"memberOfTypes": [ "Album" ],
"shape": {
"type": "Record",
"attributes": {
"private": { "type": "Boolean" },
"account": {
"type": "Entity",
"name": "Account"
}
}
}
},
"Album": {
"memberOfTypes": [ "Album" ],
"shape": {
"type": "Record",
"attributes": {
"private": { "type": "Boolean" },
"account": {
"type": "Entity",
"name": "Account"
}
}
}
},
"Account": {
"memberOfTypes": [],
"shape": {
"type": "Record",
"attributes": {
"owner": {
"type": "Entity",
"name": "User"
},
"admins": {
"required": false,
"type": "Set",
"element": {
"type": "Entity",
"name": "User"
}
}
}
}
}
},
"actions": {
"viewPhoto": {
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Photo" ],
"context": {
"type": "Record",
"attributes": {
"authenticated": { "type": "Boolean" }
}
}
}
},
"listAlbums": {
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Account" ],
"context": {
"type": "Record",
"attributes": {
"authenticated": { "type": "Boolean" }
}
}
}
},
"uploadPhoto": {
"appliesTo": {
"principalTypes": [ "User" ],
"resourceTypes": [ "Album" ],
"context": {
"type": "Record",
"attributes": {
"authenticated": { "type": "Boolean" },
"photo": {
"type": "Record",
"attributes": {
"file_size": { "type": "Long" },
"file_type": { "type": "String" }
}
}
}
}
}
}
}
}
}