OData Common Schema Definition Language (CSDL) JSON Representation Version 4.02

Committee Specification Draft 01

28 February 2024

\(\hbox{}\)

This stage:

https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/csd01/odata-csdl-json-v4.02-csd01.md (Authoritative)
https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/csd01/odata-csdl-json-v4.02-csd01.html
https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/csd01/odata-csdl-json-v4.02-csd01.pdf

Previous stage:

N/A

Latest stage:

https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/odata-csdl-json-v4.02.md (Authoritative)
https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/odata-csdl-json-v4.02.html
https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/odata-csdl-json-v4.02.pdf

Technical Committee:

OASIS Open Data Protocol (OData) TC

Chairs:

Ralf Handl (ralf.handl@sap.com), SAP SE
Michael Pizzo (mikep@microsoft.com), Microsoft

Editors:

Ralf Handl (ralf.handl@sap.com), SAP SE
Michael Pizzo (mikep@microsoft.com), Microsoft
Heiko Theißen (heiko.theissen@sap.com), SAP SE

Additional artifacts:

This prose specification is one component of a Work Product that also includes:

• JSON schemas: OData CSDL JSON schema. https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/csd01/schemas/.

Related work:

This specification replaces or supersedes:

• OData Common Schema Definition Language (CSDL) JSON Representation Version 4.01. Edited by Michael Pizzo, Ralf Handl, and Martin Zurmuehl. OASIS Standard. Latest stage: https://docs.oasis-open.org/odata/odata-csdl-json/v4.01/odata-csdl-json-v4.01.html.

This specification is related to:

• OData Version 4.02. Edited by Michael Pizzo, Ralf Handl, and Heiko Theißen. A multi-part Work Product that includes:
  • OData Version 4.02 Part 1: Protocol. Latest stage: https://docs.oasis-open.org/odata/odata/v4.02/odata-v4.02-part1-protocol.html
  • OData Version 4.02 Part 2: URL Conventions. Latest stage: https://docs.oasis-open.org/odata/odata/v4.02/odata-v4.02-part2-url-conventions.html
  • ABNF components: OData ABNF Construction Rules Version 4.02 and OData ABNF Test Cases. https://docs.oasis-open.org/odata/odata/v4.02/csd01/abnf/
• OData Vocabularies Version 4.0. Edited by Michael Pizzo, Ralf Handl, and Ram Jeyaraman. Latest stage: https://docs.oasis-open.org/odata/odata-vocabularies/v4.0/odata-vocabularies-v4.0.html
• OData Common Schema Definition Language (CSDL) XML Representation Version 4.02. Edited by Michael Pizzo, Ralf Handl, and Heiko Theißen. Latest stage: https://docs.oasis-open.org/odata/odata-csdl-xml/v4.02/odata-csdl-xml-v4.02.html
• OData JSON Format Version 4.02. Edited by Ralf Handl, Mike Pizzo, and Heiko Theißen. Latest stage: https://docs.oasis-open.org/odata/odata-json-format/v4.02/odata-json-format-v4.02.html

Abstract:

OData services are described by an Entity Model (EDM). The Common Schema Definition Language (CSDL) defines specific representations of the entity data model exposed by an OData service, using XML, JSON, and other formats. This document (OData CSDL JSON Representation) specifically defines the JSON representation of CSDL.

Status:

This document was last revised or approved by the OASIS Open Data Protocol (OData) TC on the above date. The level of approval is also listed above. Check the “Latest stage” location noted above for possible later revisions of this document. Any other numbered Versions and other technical work produced by the Technical Committee (TC) are listed at https://www.oasis-open.org/committees/tc_home.php?wg_abbrev=odata#technical.

TC members should send comments on this specification to the TC’s email list. Others should send comments to the TC’s public comment list, after subscribing to it by following the instructions at the “Send A Comment” button on the TC’s web page at https://www.oasis-open.org/committees/odata/.

This specification is provided under the RF on RAND Terms Mode of the OASIS IPR Policy, the mode chosen when the Technical Committee was established. For information on whether any patents have been disclosed that may be essential to implementing this specification, and any offers of patent licensing terms, please refer to the Intellectual Property Rights section of the TC’s web page (https://www.oasis-open.org/committees/odata/ipr.php).

Note that any machine-readable content (Computer Language Definitions) declared Normative for this Work Product is provided in separate plain text files. In the event of a discrepancy between any such plain text file and display content in the Work Product’s prose narrative document(s), the content in the separate plain text file prevails.

Key words:

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in BCP 14 RFC2119 and RFC8174 when, and only when, they appear in all capitals, as shown here.

Citation format:

When referencing this specification the following citation format should be used:

[OData-CSDL-JSON-v4.02]

OData Common Schema Definition Language (CSDL) JSON Representation Version 4.02. Edited by Ralf Handl, Michael Pizzo, and Heiko Theißen. 28 February 2024. OASIS Committee Specification Draft 01. https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/csd01/odata-csdl-json-v4.02-csd01.html. Latest stage: https://docs.oasis-open.org/odata/odata-csdl-json/v4.02/odata-csdl-json-v4.02.html.

Notices

Copyright © OASIS Open 2024. All Rights Reserved.

Distributed under the terms of the OASIS IPR Policy.

The name “OASIS” is a trademark of OASIS, the owner and developer of this specification, and should be used only to refer to the organization and its official outputs.

For complete copyright information please see the full Notices section in an Appendix below.

Table of Contents

1 Introduction

OData services are described in terms of an Entity Model. The Common Schema Definition Language (CSDL) defines a representation of the entity model exposed by an OData service using the JavaScript Object Notation (JSON), see RFC8259.

This format is based on the OpenUI5 OData V4 Metadata JSON Format, see OpenUI5, with some extensions and modifications made necessary to fully cover OData CSDL Version 4.01.

1.1 Changes from Earlier Versions

1.2 Glossary

1.2.1 Definitions of Terms

1.2.2 Acronyms and Abbreviations

1.2.3 Document Conventions

Keywords defined by this specification use this monospaced font.

Some sections of this specification are illustrated with non-normative examples.

Non-normative examples use this paragraph style.

All examples in this document are non-normative and informative only.

Representation-specific text is indented and marked with vertical lines.

All other text is normative unless otherwise labeled.

2 JSON Representation

OData CSDL JSON is a full representation of the OData Common Schema Definition Language in the JavaScript Object Notation (JSON) defined in RFC8259. It additionally follows the rules for “Internet JSON” (I-JSON) defined in RFC7493 for e.g. objects, numbers, date values, and duration values.

It is an alternative to the CSDL XML representation defined in OData-CSDLXML and neither adds nor removes features.

2.1 Requesting the JSON Representation

The OData CSDL JSON representation can be requested using the $format query option in the request URL with the media type application/json, optionally followed by media type parameters, or the case-insensitive abbreviation json which MUST NOT be followed by media type parameters.

Alternatively, this representation can be requested using the Accept header with the media type application/json, optionally followed by media type parameters.

If specified, $format overrides any value specified in the Accept header.

The response MUST contain the Content-Type header with a value of application/json, optionally followed by media type parameters.

Possible media type parameters are:

• IEEE754Compatible
• metadata

The names and values of these parameters are case-insensitive.

2.1.1 Controlling the Representation of Numbers

The IEEE754Compatible=true parameter indicates that the service MUST serialize Edm.Int64 and Edm.Decimal numbers as strings. This is in conformance with RFC7493. If not specified, or specified as IEEE754Compatible=false, all numbers MUST be serialized as JSON numbers.

This enables support for JavaScript numbers that are defined to be 64-bit binary format IEEE 754 values ECMAScript (see section 4.3.1.9) resulting in integers losing precision past 15 digits, and decimals losing precision due to the conversion from base 10 to base 2.

Responses that format Edm.Int64 and Edm.Decimal values as strings MUST specify this parameter in the media type returned in the Content-Type header.

2.1.2 Controlling the Amount of Control Information

The representation of constant annotation values in CSDL JSON documents closely follows the representation of data defined in OData-JSON.

A client application can use the metadata format parameter in the Accept header when requesting a CSDL JSON document to influence how much control information will be included in the response.

Other Accept header parameters are orthogonal to the metadata parameter and are therefore not mentioned in this section.

2.1.2.1 metadata=minimal

The metadata=minimal format parameter indicates that the service SHOULD remove computable control information from the payload wherever possible.

This means that the @type control information is only included if the type of the containing object or targeted property cannot be heuristically determined, e.g. for

• Terms or term properties with an abstract declared type,
• Terms or term properties with a declared type that has derived types, or
• Dynamic properties of open types.

See OData-JSON for the exact rules.

2.1.2.2 metadata=full

The metadata=full format parameter indicates that the service MUST include all control information explicitly in the payload.

This means that the @type control information is included in annotation values except for primitive values whose type can be heuristically determined from the representation of the value, see OData-JSON for the exact rules.

2.1.2.3 metadata=none

The metadata=none format parameter indicates that the service SHOULD omit all control information.

2.2 Design Considerations

CSDL JSON documents are designed for easy and efficient lookup of model constructs by their name without having to know or guess what kind of model element it is. Thus, all primary model elements (entity types, complex types, type definitions, enumeration types, terms, actions, functions, and the entity container) are direct members of their schema, using the schema-unique name as the member name. Similarly, child elements of primary model elements (properties, navigation properties, enumeration type members, entity sets, singletons, action imports, and function imports) are direct members of the objects describing their parent model element, using their locally unique name as the member name.

To avoid name collisions, all fixed member names are prefixed with a dollar ($) sign and otherwise have the same name and capitalization as their counterparts in the CSDL XML representation OData-CSDLXML (with one exception: the counterpart of the EntitySet element’s EntityType member is $Type, to harmonize it with all other type references).

Additional fixed members introduced by this specification and without counterpart in OData-CSDLXML are also prefixed with a dollar ($) sign and use upper-camel-case names. One of these is $Kind which represents the kind of model element. Its value is the upper-camel-case local name of the XML element representing this kind of model element in OData-CSDLXML, e.g. EntityType or NavigationProperty.

While the XML representation of CSDL allows referencing model elements with alias-qualified names as well as with namespace-qualified names, this JSON representation requires the use of alias-qualified names if an alias is specified for an included or document-defined schema. Aliases are usually shorter than namespaces, so this reduces text size of the JSON document. Text size matters even if the actual HTTP messages are sent in compressed form because the decompressed form needs to be reconstructed, and clients not using a streaming JSON parser have to materialize the full JSON document before parsing.

To further reduce size the member $Kind is optional for structural properties as these are more common than navigation properties, and the member $Type is optional for string properties, parameters, and return types, as this type is more common than other primitive types.

In general, all members that have a default value SHOULD be omitted if they have the default value.

2.3 JSON Schema Definition

The structure of CSDL JSON documents can be verified with the JSON Schema OData-CSDL-Schema provided as an additional artifact of this prose specification. This schema only defines the shape of a well-formed CSDL JSON document but is not descriptive enough to define what a correct CSDL JSON document MUST be in every imaginable use case. This specification document defines additional rules that correct CSDL JSON documents MUST fulfill. In case of doubt on what makes a CSDL JSON document correct the rules defined in this specification document take precedence.

3 Entity Model

An OData service exposes a single entity model. This model may be distributed over several schemas, and these schemas may be distributed over several documents.

A service is defined by a single CSDL document which can be accessed by sending a GET request to <serviceRoot>/$metadata. This document is called the metadata document. It MAY reference other CSDL documents.

The metadata document contains a single entity container that defines the resources exposed by this service. This entity container MAY extend an entity container defined in a referenced document.

The model of the service consists of all CSDL constructs used in its entity containers.

The scope of a CSDL document is the document itself and all schemas included from directly referenced documents. All entity types, complex types and other named model elements in scope (that is, defined in the document itself or a schema of a directly referenced document) can be accessed from a referencing document by their qualified names. This includes the built-in primitive and abstract types.

Referencing another document may alter the model defined by the referencing document. For instance, if a referenced document defines an entity type derived from an entity type in the referencing document, then an entity set of the service defined by the referencing document may return entities of the derived type. This is identical to the behavior if the derived type had been defined directly in the referencing document.

Note: referencing documents is not recursive. Only named model elements defined in directly referenced documents can be used within the schema. However, those elements may in turn include or reference model elements defined in schemas referenced by their defining schema.

3.1 Nominal Types

A nominal type has a name that MUST be a simple identifier. Nominal types are referenced using their qualified name. The qualified type name MUST be unique within a model as it facilitates references to the element from other parts of the model.

Names are case-sensitive, but service authors SHOULD NOT choose names that differ only in case.

3.2 Structured Types

Structured types are composed of other model elements. Structured types are common in entity models as the means of representing entities and structured properties in an OData service. Entity types and complex types are both structured types.

Structured types are composed of zero or more structural properties and navigation properties. These properties can themselves be of a structured type.

An instance of a structured type must have a finite representation that includes all its properties. These properties are either integral parts of the instance or references to instances. In the first case the integral parts must be represented as part of the overall representation. These integral parts are modeled as structural properties and containment navigation properties. If an instance of a structured type contains a chain of these, this chain MUST be finite for the overall representation to be finite, even if the chain of types leads back to the structured type of the instance. Note that in this circular case finiteness is only possible if the chain of instances ends with a null value or an empty collection. In the second case the references are modeled as non-containment navigation properties. Chains of these can be infinite, for example, if an entity contains a self-reference.

Open entity types and open complex types allow properties to be added dynamically to instances of the open type.

3.3 Primitive Types

Structured types are composed of other structured types and primitive types. OData defines the following primitive types:

Edm.Date and Edm.DateTimeOffset follow XML-Schema-2 and use the proleptic Gregorian calendar, allowing the year 0000 (equivalent to 1 BCE) and negative years (year -0001 being equivalent to 2 BCE etc.). The supported date range is service-specific and typically depends on the underlying persistency layer, e.g. SQL only supports years 0001 to 9999.

Edm.Decimal with a Scale value of floating, Edm.Double, and Edm.Single allow the special numeric values -INF, INF, and NaN.

Edm.Stream is a primitive type that can be used as a property of an entity type or complex type, the underlying type for a type definition, or a binding or non-binding parameter or return type of an action or function. Edm.Stream, or a type definition whose underlying type is Edm.Stream, cannot be used in collections.

Some of these types allow facets, defined in section “Type Facets”.

See rule primitiveLiteral in OData-ABNF for the representation of primitive type values in URLs and OData-JSON for the representation in requests and responses.

3.4 Type Facets

The facets in the following subsections modify or constrain the acceptable values of primitive typed model elements, for example a structural property, action or function parameter, action or function return type, or term.

For single-valued model elements the facets apply to the value of the model element. For collection-valued model elements the facets apply to the items in the collection.

3.4.1 MaxLength

A positive integer value specifying the maximum length of a binary, stream or string value. For binary or stream values this is the octet length of the binary data, for string values it is the character length (number of code points for Unicode).

If no maximum length is specified, clients SHOULD expect arbitrary length.

3.4.2 Precision

For a decimal value: the maximum number of significant decimal digits of the model element’s value; it MUST be a positive integer.

For a temporal value (datetime-with-timezone-offset, duration, or time-of-day): the number of decimal places allowed in the seconds portion of the value; it MUST be a non-negative integer between zero and twelve.

Note: service authors SHOULD be aware that some clients are unable to support a precision greater than 28 for decimal values and 7 for temporal values. Client developers MUST be aware of the potential for data loss when round-tripping values of greater precision. Updating via PATCH and exclusively specifying modified values will reduce the risk for unintended data loss.

Note: model elements with duration values and a granularity less than seconds (e.g. minutes, hours, days) can be annotated with the term Measures.DurationGranularity, see OData-VocMeasures.

"SuggestedTimes": {
  "$Type": "Edm.DateTimeOffset",
  "$Collection": true,
  "$Precision": 6
}

3.4.3 Scale

A non-negative integer value specifying the maximum number of digits allowed to the right of the decimal point, or one of the symbolic values floating or variable.

The value floating means that the decimal value represents a decimal floating-point number whose number of significant digits is the value of the Precision facet. OData 4.0 responses MUST NOT specify the value floating.

The value variable means that the number of digits to the right of the decimal point can vary from zero to the value of the Precision facet.

An integer value means that the number of digits to the right of the decimal point may vary from zero to the value of the Scale facet, and the number of digits to the left of the decimal point may vary from one to the value of the Precision facet minus the value of the Scale facet. If Precision is equal to Scale, a single zero MUST precede the decimal point.

The value of Scale MUST be less than or equal to the value of Precision.

Note: if the underlying data store allows negative scale, services may use a Precision with the absolute value of the negative scale added to the actual number of significant decimal digits, and client-provided values may have to be rounded before being stored.

"Amount32": {
  "$Type": "Edm.Decimal",
  "$Precision": 3,
  "$Scale": 2
}

"Amount22": {
  "$Type": "Edm.Decimal",
  "$Precision": 2,
  "$Scale": 2
}

"Amount3v": {
  "$Type": "Edm.Decimal",
  "$Precision": 3
}

"Amount7f": {
  "$Type": "Edm.Decimal",
  "$Precision": 7,
  "$Scale": "floating"
}

3.4.4 Unicode

For a string-typed model element the Unicode facet indicates whether the it might contain and accept string values with Unicode characters (code points) beyond the ASCII character set. The value false indicates that the it will only contain and accept string values with characters limited to the ASCII character set.

If no value is specified, the Unicode facet defaults to true.

3.4.5 SRID

For a geometry- or geography-typed model element the SRID facet identifies which spatial reference system is applied to its values.

The value of the SRID facet MUST be a non-negative integer or the special value variable. If no value is specified, the facet defaults to 0 for Geometry types or 4326 for Geography types.

The valid values of the SRID facet and their meanings are as defined by the European Petroleum Survey Group EPSG.

3.5 Built-In Abstract Types

The following built-in abstract types can be used within a model:

• Edm.PrimitiveType
• Edm.ComplexType
• Edm.EntityType
• Edm.Untyped

Conceptually, these are the abstract base types for primitive types (including type definitions and enumeration types), complex types, entity types, or any type or collection of types, respectively, and can be used anywhere a corresponding concrete type can be used, except:

• Edm.EntityType
  • cannot be used as the type of a singleton in an entity container because it doesn’t define a structure, which defeats the purpose of a singleton.
  • cannot be used as the type of an entity set because all entities in an entity set must have the same key fields to uniquely identify them within the set.
  • cannot be the base type of an entity type or complex type.
• Edm.ComplexType
  • cannot be the base type of an entity type or complex type.
• Edm.PrimitiveType
  • cannot be used as the type of a key property of an entity type or as the underlying type of an enumeration type.
  • cannot be used as the underlying type of a type definition in a CSDL document with a version of 4.0.
  • can be used as the underlying type of a type definition in a CSDL document with a version of 4.01 or greater.
• Edm.Untyped
  • cannot be returned in a payload with an OData-Version header of 4.0. Services should treat untyped properties as dynamic properties in 4.0 payloads.
  • cannot be used as the type of a key property of an entity type.
  • cannot be the base type of an entity type or complex type.
  • cannot be used as the underlying type of a type definition or enumeration type.
• Collection(Edm.PrimitiveType)
  • cannot be used.
• Collection(Edm.Untyped)
  • cannot be returned in a payload with an OData-Version header of 4.0. Services should treat untyped properties as dynamic properties in 4.0 payloads.

3.6 Built-In Types for defining Vocabulary Terms

Vocabulary terms can, in addition, use

• Edm.AnnotationPath
• Edm.PropertyPath
• Edm.NavigationPropertyPath
• Edm.AnyPropertyPath (Edm.PropertyPath or Edm.NavigationPropertyPath)
• Edm.ModelElementPath (any model element, including Edm.AnnotationPath, Edm.NavigationPropertyPath, and Edm.PropertyPath)

as the type of a primitive term, or the type of a property of a complex type (recursively) that is exclusively used as the type of a term. See section “Path Expressions” for details.

3.7 Annotations

Many parts of the model can be decorated with additional information using annotations. Annotations are identified by their term name and an optional qualifier that allows applying the same term multiple times to the same model element.

A model element MUST NOT specify more than one annotation for a given combination of term and qualifier.

4 CSDL JSON Document

{
  "$Version": "4.01",
  "$EntityContainer": "org.example.DemoService",
  …
}

4.1 Reference

A reference to an external CSDL document allows to bring part of the referenced document’s content into the scope of the referencing document.

A reference MUST specify a URI that uniquely identifies the referenced document, so two references MUST NOT specify the same URI. The URI SHOULD be a URL that locates the referenced document. If the URI is not dereferencable it SHOULD identify a well-known schema. The URI MAY be absolute or relative URI; relative URLs are relative to the URL of the document containing the reference, or relative to a base URL specified in a format-specific way.

A reference MAY be annotated.

The Core.SchemaVersion annotation, defined in OData-VocCore, MAY be used to indicate a particular version of the referenced document. If the Core.SchemaVersion annotation is present, the $schemaversion system query option, defined OData-Protocol, SHOULD be used when retrieving the referenced schema document.

{
  …
  "$Reference": {
    "https://oasis-tcs.github.io/odata-vocabularies/vocabularies/Org.OData.Capabilities.V1.json": {
      …
    },
    "https://oasis-tcs.github.io/odata-vocabularies/vocabularies/Org.OData.Core.V1.json": {
      …
    },
    "http://example.org/display/v1": {
      …
    }
  },
  …
}

4.2 Included Schema

A reference MAY include zero or more schemas from the referenced document.

The included schemas are identified via their namespace. The same namespace MUST NOT be included more than once, even if it is declared in more than one referenced document.

When including a schema, a simple identifier value MAY be specified as an alias for the schema that is used in qualified names instead of the namespace. For example, an alias of display might be assigned to the namespace org.example.vocabularies.display. An alias-qualified name is resolved to a fully qualified name by examining aliases for included schemas and schemas defined within the document.

Aliases are document-global, so all schemas defined within or included into a document MUST have different aliases, and aliases MUST differ from the namespaces of all schemas defined within or included into a document.

The alias MUST NOT be one of the reserved values Edm, odata, System, or Transient.

An alias is only valid within the document in which it is declared; a referencing document may define its own aliases for included schemas.

{
  …
  "$Reference": {
    "https://oasis-tcs.github.io/odata-vocabularies/vocabularies/Org.OData.Capabilities.V1.json": {
      "$Include": [
        {
          "$Namespace": "Org.OData.Capabilities.V1",
          "$Alias": "Capabilities"
        }
      ]
    },
    "https://oasis-tcs.github.io/odata-vocabularies/vocabularies/Org.OData.Core.V1.json": {
      "$Include": [
        {
          "$Namespace": "Org.OData.Core.V1",
          "$Alias": "Core",
          "@Core.DefaultNamespace": true
        }
      ]
    },
    "http://example.org/display/v1": {
      "$Include": [
        {
          "$Namespace": "org.example.display",
          "$Alias": "UI"
        }
      ]
    }
  },
  …
}

4.3 Included Annotations

In addition to including whole schemas with all model constructs defined within that schema, a reference may include annotations.

Annotations are selectively included by specifying the namespace of the annotations’ term. Consumers can opt not to inspect the referenced document if none of the term namespaces is of interest for the consumer.

In addition, the qualifier of annotations to be included MAY be specified. For instance, a service author might want to supply a different set of annotations for various device form factors. If a qualifier is specified, only those annotations from the specified term namespace with the specified qualifier (applied to a model element of the target namespace, if present) SHOULD be included. If no qualifier is specified, all annotations within the referenced document from the specified term namespace (taking into account the target namespace, if present) SHOULD be included.

The qualifier also provides consumers insight about what qualifiers are present in the referenced document. If the consumer is not interested in that particular qualifier, the consumer can opt not to inspect the referenced document.

In addition, the namespace of the annotations’ target MAY be specified. If a target namespace is specified, only those annotations which apply a term form the specified term namespace to a model element of the target namespace (with the specified qualifier, if present) SHOULD be included. If no target namespace is specified, all annotations within the referenced document from the specified term namespace (taking into account the qualifier, if present) SHOULD be included.

The target namespace also provides consumers insight about what namespaces are present in the referenced document. If the consumer is not interested in that particular target namespace, the consumer can opt not to inspect the referenced document.

{
  …
  "$Reference": {
    "http://odata.org/ann/b": {
      "$IncludeAnnotations": [
        {
          "$TermNamespace": "org.example.validation"
        },
        {
          "$TermNamespace": "org.example.display",
          "$Qualifier": "Tablet"
        },
        {
          "$TermNamespace": "org.example.hcm",
          "$TargetNamespace": "com.example.Sales"
        },
        {
          "$TermNamespace": "org.example.hcm",
          "$Qualifier": "Tablet",
          "$TargetNamespace": "com.example.Person"
        }
      ]
    }
  },
  …
}

5 Schema

One or more schemas describe the entity model exposed by an OData service. The schema acts as a namespace for elements of the entity model such as entity types, complex types, enumerations and terms.

A schema is identified by a namespace. Schema namespaces MUST be unique within the scope of a document and SHOULD be globally unique. A schema cannot span more than one document.

The schema’s namespace is combined with the name of elements in the schema to create unique qualified names, so identifiers that are used to name types MUST be unique within a namespace to prevent ambiguity.

Services SHOULD NOT have actions and functions with the same name, and MUST NOT have actions and functions with the same name bound to the same type.

Names are case-sensitive, but service authors SHOULD NOT choose names that differ only in case.

The namespace MUST NOT be one of the reserved values Edm, odata, System, or Transient.

5.1 Alias

A schema MAY specify an alias which MUST be a simple identifier.

If a schema specifies an alias, the alias MUST be used instead of the namespace within qualified names throughout the document to identify model elements of that schema. A mixed use of namespace-qualified names and alias-qualified names is not allowed.

Aliases are document-global, so all schemas defined within or included into a document MUST have different aliases, and aliases MUST differ from the namespaces of all schemas defined within or included into a document. Aliases defined by a schema can be used throughout the containing document and are not restricted to the schema that defines them.

The alias MUST NOT be one of the reserved values Edm, odata, System, or Transient.

{
  …
  "org.example": {
    "$Alias": "self",
    "@Core.Description": "Example schema",
    …
  },
  …
}

5.2 Annotations with External Targeting

"org.example": {
  "$Alias": "self",
  "$Annotations": {
    "self.Person": {
      "@Core.Description#Tablet": "Dummy",
      …
    }
  }
}

6 Entity Type

Entity types are nominal structured types with a key that consists of one or more references to structural properties. An entity type is the template for an entity: any uniquely identifiable record such as a customer or order.

The entity type’s name is a simple identifier that MUST be unique within its schema.

An entity type can define two types of properties. A structural property is a named reference to a primitive, complex, or enumeration type, or a collection of primitive, complex, or enumeration types. A navigation property is a named reference to another entity type or collection of entity types.

All properties MUST have a unique name within an entity type. Properties MUST NOT have the same name as the declaring entity type. They MAY have the same name as one of the direct or indirect base types or derived types.

"Employee": {
  "$Kind": "EntityType",
  "$Key": [
    "ID"
  ],
  "ID": {},
  "FirstName": {},
  "LastName": {},
  "Manager": {
    "$Kind": "NavigationProperty",
    "$Nullable": true,
    "$Type": "self.Manager"
  }
}

6.1 Derived Entity Type

An entity type can inherit from another entity type by specifying it as its base type.

An entity type inherits the key as well as structural and navigation properties of its base type.

An entity type MUST NOT introduce an inheritance cycle by specifying a base type.

"Manager": {
  "$Kind": "EntityType",
  "$BaseType": "self.Employee",
  "AnnualBudget": {
    "$Nullable": true,
    "$Type": "Edm.Decimal",
    "$Scale": 0
  },
  "Employees": {
    "$Kind": "NavigationProperty",
    "$Collection": true,
    "$Type": "self.Employee"
  }
}

6.2 Abstract Entity Type

An entity type MAY indicate that it is abstract and cannot have instances.

For OData 4.0 responses a non-abstract entity type MUST define a key or derive from a base type with a defined key.

An abstract entity type MUST NOT inherit from a non-abstract entity type.

6.3 Open Entity Type

An entity type MAY indicate that it is open and allows clients to add properties dynamically to instances of the type by specifying uniquely named property values in the payload used to insert or update an instance of the type.

An entity type derived from an open entity type MUST indicate that it is also open.

Note: structural and navigation properties MAY be returned by the service on instances of any structured type, whether or not the type is marked as open. Clients MUST always be prepared to deal with additional properties on instances of any structured type, see OData-Protocol.

6.4 Media Entity Type

An entity type that does not specify a base type MAY indicate that it is a media entity type. Media entities are entities that represent a media stream, such as a photo. Use a media entity if the out-of-band stream is the main topic of interest and the media entity is just additional structured information attached to the stream. Use a normal entity with one or more properties of type Edm.Stream if the structured data of the entity is the main topic of interest and the stream data is just additional information attached to the structured data. For more information on media entities see OData-Protocol.

An entity type derived from a media entity type MUST indicate that it is also a media entity type.

Media entity types MAY specify a list of acceptable media types using an annotation with term Core.AcceptableMediaTypes, see OData-VocCore.

6.5 Key

An entity is uniquely identified within an entity set by its key. A key MAY be specified if the entity type does not specify a base type that already has a key declared.

In order to be specified as the type of an entity set or a collection-valued containment navigation property, the entity type MUST either specify a key or inherit its key from its base type.

In OData 4.01 responses entity types used for singletons or single-valued navigation properties do not require a key. In OData 4.0 responses entity types used for singletons or single-valued navigation properties MUST have a key defined.

An entity type’s key refers to the set of properties whose values uniquely identify an instance of the entity type within an entity set. The key MUST consist of at least one property.

Key properties MUST NOT be nullable and MUST be typed with an enumeration type, one of the following primitive types, or a type definition based on one of these primitive types:

• Edm.Boolean
• Edm.Byte
• Edm.Date
• Edm.DateTimeOffset
• Edm.Decimal
• Edm.Duration
• Edm.Guid
• Edm.Int16
• Edm.Int32
• Edm.Int64
• Edm.SByte
• Edm.String
• Edm.TimeOfDay

Key property values MAY be language-dependent, but their values MUST be unique across all languages and the entity-ids (defined in OData-Protocol) MUST be language independent.

A key property MUST be a non-nullable primitive property of the entity type itself, including non-nullable primitive properties of non-nullable single-valued complex properties, recursively.

In OData 4.01 the key properties of a directly related entity type MAY also be part of the key if the navigation property is single-valued and not nullable. This includes navigation properties of non-nullable single-valued complex properties (recursively) of the entity type. If a key property of a related entity type is part of the key, all key properties of the related entity type MUST also be part of the key.

If the key property is a property of a complex property (recursively) or of a directly related entity type, the key MUST specify an alias for that property that MUST be a simple identifier and MUST be unique within the set of aliases, structural and navigation properties of the declaring entity type and any of its base types.

An alias MUST NOT be defined if the key property is a primitive property of the entity type itself.

For key properties that are a property of a complex or navigation property, the alias MUST be used in the key predicate of URLs instead of the path to the property because the required percent-encoding of the forward slash separating segments of the path to the property would make URL construction and parsing rather complicated. The alias MUST NOT be used in the query part of URLs, where paths to properties don’t require special encoding and are a standard constituent of expressions anyway.

"Category": {
  "$Kind": "EntityType",
  "$Key": [
    "ID"
  ],
  "ID": {
    "$Type": "Edm.Int32"
  },
  "Name": {
    "$Nullable": true,
    "@Core.IsLanguageDependent": true
  }
}

"Category": {
  "$Kind": "EntityType",
  "$Key": [
    {
      "EntityInfoID": "Info/ID"
    }
  ],
  "Info": {
    "$Type": "self.EntityInfo"
  },
  "Name": {
    "$Nullable": true
  }
},
"EntityInfo": {
  "$Kind": "ComplexType",
  "ID": {
    "$Type": "Edm.Int32"
  },
  "Created": {
    "$Type": "Edm.DateTimeOffset",
    "$Precision": 0
  }
}

"OrderLine": {
  "$Kind": "EntityType",
  "$Key": [
    "OrderID",
    "LineNumber"
  ],
  "OrderID": {
    "$Type": "Edm.Int32"
  },
  "LineNumber": {
    "$Type": "Edm.Int32"
  }
}

GET http://host/service/Categories(EntityInfoID=1)

GET http://example.org/OData.svc/Categories?$filter=Info/ID le 100

7 Structural Property

A structural property is a property of a structured type that has one of the following types:

• Primitive type
• Complex type
• Enumeration type
• A collection of one of the above

A structural property MUST specify a unique name as well as a type.

The property’s name MUST be a simple identifier. It is used when referencing, serializing or deserializing the property. It MUST be unique within the set of structural and navigation properties of the declaring structured type, and MUST NOT match the name of any navigation property in any of its base types. If a structural property with the same name is defined in any of this type’s base types, then the property’s type MUST be a type derived from the type specified for the property of the base type and constrains this property to be of the specified subtype for instances of this structured type. The name MUST NOT match the name of any structural or navigation property of any of this type’s base types for OData 4.0 responses.

Names are case-sensitive, but service authors SHOULD NOT choose names that differ only in case.

"Measurement": {
  "$Kind": "ComplexType",
  "Dimension": {
    "$MaxLength": 50,
    "$DefaultValue": "Unspecified"
  },
  "Length": {
    "$Type": "Edm.Decimal",
    "$Precision": 18,
    "$Scale": 2
  }
}

7.1 Type

The property’s type MUST be a primitive type, complex type, or enumeration type in scope, or a collection of one of these types.

If the property is part of a chain of structural properties and containment navigation properties leading back to the property’s declaring type, the finiteness condition for structured types demands that at least one property in this chain MUST be nullable or collection-valued.

A collection-valued property MAY be annotated with the Core.Ordered term, defined in OData-VocCore, to specify that it supports a stable ordering.

A collection-valued property MAY be annotated with the Core.PositionalInsert term, defined in OData-VocCore, to specify that it supports inserting items into a specific ordinal position.

"Units": {
  "$Collection": true
}

7.2 Nullable

A Boolean value specifying whether the property can have the value null.

7.3 Default Value

A primitive- or enumeration-typed property MAY define a default value that is used if the property is not explicitly represented in an annotation or the body of a request or response.

If no value is specified, the client SHOULD NOT assume a default value.

8 Navigation Property

A navigation property allows navigation to related entities. It MUST specify a unique name as well as a type.

The navigation property’s name MUST be a simple identifier. It is used when referencing, serializing or deserializing the navigation property. It MUST be unique within the set of structural and navigation properties of the declaring structured type, and MUST NOT match the name of any structural property in any of its base types. If a navigation property with the same name is defined in any of this type’s base types, then the navigation property’s type MUST be a type derived from the type specified for the navigation property of the base type, and constrains this navigation property to be of the specified subtype for instances of this structured type. The name MUST NOT match the name of any structural or navigation property of any of this type’s base types for OData 4.0 responses.

Names are case-sensitive, but service authors SHOULD NOT choose names that differ only in case.

"Product": {
  "$Kind": "EntityType",
  …
  "Category": {
    "$Kind": "NavigationProperty",
    "$Type": "self.Category",
    "$Partner": "Products",
    "$ReferentialConstraint": {
      "CategoryID": "ID"
    }
  },
  "Supplier": {
    "$Kind": "NavigationProperty",
    "$Type": "self.Supplier"
  }
},
"Category": {
  "$Kind": "EntityType",
  …
  "Products": {
    "$Kind": "NavigationProperty",
    "$Collection": true,
    "$Type": "self.Product",
    "$Partner": "Category",
    "$OnDelete": "Cascade",
    "$OnDelete@Core.Description": "Delete all related entities"
  }
}

8.1 Navigation Property Type

The navigation property’s type MUST be an entity type in scope, the abstract type Edm.EntityType, or a collection of one of these types.

If the property is part of a chain of structural properties and containment navigation properties leading back to the property’s declaring type, the finiteness condition for structured types demands that at least one property in this chain MUST be nullable or collection-valued.

If the type is a collection, an arbitrary number of entities can be related. Otherwise there is at most one related entity.

The related entities MUST be of the specified entity type or one of its subtypes.

For a collection-valued containment navigation property the specified entity type MUST have a key defined.

A collection-valued navigation property MAY be annotated with the Core.Ordered term, defined in OData-VocCore, to specify that it supports a stable ordering.

A collection-valued navigation property MAY be annotated with the Core.PositionalInsert term, defined in OData-VocCore, to specify that it supports inserting items into a specific ordinal position.

8.2 Nullable Navigation Property

A Boolean value specifying whether the declaring type MAY have no related entity. If false, instances of the declaring structured type MUST always have a related entity.

Nullable MUST NOT be specified for a collection-valued navigation property, a collection is allowed to have zero items.

8.3 Partner Navigation Property

A navigation property of an entity type MAY specify a partner navigation property. Navigation properties of complex types MUST NOT specify a partner.

If specified, the partner navigation property is identified by a path relative to the entity type specified as the type of the navigation property. This path MUST lead to a navigation property defined on that type or a derived type. The path MAY traverse complex types, including derived complex types, but MUST NOT traverse any navigation properties. The type of the partner navigation property MUST be the declaring entity type of the current navigation property or one of its parent entity types.

If the partner navigation property is single-valued, it MUST lead back to the source entity from all related entities. If the partner navigation property is collection-valued, the source entity MUST be part of that collection.

If no partner navigation property is specified, no assumptions can be made as to whether one of the navigation properties on the target type will lead back to the source entity.

If a partner navigation property is specified, this partner navigation property MUST either specify the current navigation property as its partner to define a bi-directional relationship or it MUST NOT specify a partner navigation property. The latter can occur if the partner navigation property is defined on a complex type, or if the current navigation property is defined on a type derived from the type of the partner navigation property.

8.4 Containment Navigation Property

A navigation property MAY indicate that instances of its declaring structured type contain the targets of the navigation property, in which case the navigation property is called a containment navigation property.

Containment navigation properties define an implicit entity set for each instance of its declaring structured type. This implicit entity set is identified by the read URL of the navigation property for that structured type instance.

Instances of the structured type that declares the navigation property, either directly or indirectly via a property of complex type, contain the entities referenced by the containment navigation property. The canonical URL for contained entities is the canonical URL of the containing instance, followed by the path segment of the navigation property and the key of the contained entity, see OData-URL.

Entity types used in collection-valued containment navigation properties MUST have a key defined.

For items of an ordered collection of complex types (those annotated with the Core.Ordered term defined in OData-VocCore the canonical URL of the item is the canonical URL of the collection appended with a segment containing the zero-based ordinal of the item. Items within in an unordered collection of complex types do not have a canonical URL. Services that support unordered collections of complex types declaring a containment navigation property, either directly or indirectly via a property of complex type, MUST specify the URL for the navigation link within a payload representing that item, according to format-specific rules.

OData 4.0 responses MUST NOT specify a complex type declaring a containment navigation property as the type of a collection-valued property.

An entity cannot be referenced by more than one containment relationship, and cannot both belong to an entity set declared within the entity container and be referenced by a containment relationship.

Containment navigation properties MUST NOT be specified as the last path segment in the path of a navigation property binding.

When a containment navigation property navigates between entity types in the same inheritance hierarchy, the containment is called recursive.

Containment navigation properties MAY specify a partner navigation property. If the containment is recursive, the relationship defines a tree, thus the partner navigation property MUST be nullable (for the root of the tree) and single-valued (for the parent of a non-root entity). If the containment is not recursive, the partner navigation property MUST NOT be nullable.

An entity type inheritance chain MUST NOT contain more than one navigation property with a partner navigation property that is a containment navigation property.

Note: without a partner navigation property, there is no reliable way for a client to determine which entity contains a given contained entity. This may lead to problems for clients if the contained entity can also be reached via a non-containment navigation path.

8.5 Referential Constraint

A single-valued navigation property MAY define one or more referential constraints. A referential constraint asserts that the dependent property (the property defined on the structured type declaring the navigation property) MUST have the same value as the principal property (the referenced property declared on the entity type that is the target of the navigation).

The type of the dependent property MUST match the type of the principal property, or both types MUST be complex types.

If the principal property references an entity, then the dependent property must reference the same entity.

If the principal property’s value is a complex type instance, then the dependent property’s value must be a complex type instance with the same properties, each with the same values.

If the navigation property on which the referential constraint is defined is nullable, or the principal property is nullable, then the dependent property MUST also be nullable. If both the navigation property and the principal property are not nullable, then the dependent property MUST NOT be nullable.

"Product": {
  "$Kind": "EntityType",
  …
  "CategoryID": {},
  "CategoryKind": {},
  "Category": {
    "$Kind": "NavigationProperty",
    "$Type": "self.Category",
    "$Partner": "Products",
    "$ReferentialConstraint": {
      "CategoryID": "ID",
      "CategoryKind": "Kind",
      "CategoryKind@Core.Description": "Referential Constraint to non-key property"
    }
  }
},
"Category": {
  "$Kind": "EntityType",
  "$Key": [
    "ID"
  ],
  "ID": {},
  "Kind": {
    "$Nullable": true
  },
  …
}

8.6 On-Delete Action

A navigation property MAY define an on-delete action that describes the action the service will take on related entities when the entity on which the navigation property is defined is deleted.

The action can have one of the following values:

• Cascade, meaning the related entities will be deleted if the source entity is deleted,
• None, meaning a DELETE request on a source entity with related entities will fail,
• SetNull, meaning all properties of related entities that are tied to properties of the source entity via a referential constraint and that do not participate in other referential constraints will be set to null,
• SetDefault, meaning all properties of related entities that are tied to properties of the source entity via a referential constraint and that do not participate in other referential constraints will be set to their default value.

If no on-delete action is specified, the action taken by the service is not predictable by the client and could vary per entity.

"Category": {
  "$Kind": "EntityType",
  …
  "Products": {
    "$Kind": "NavigationProperty",
    "$Collection": true,
    "$Type": "self.Product",
    "$Partner": "Category",
    "$OnDelete": "Cascade",
    "$OnDelete@Core.Description": "Delete all products in this category"
  }
}

9 Complex Type

Complex types are keyless nominal structured types. The lack of a key means that instances of complex types cannot be referenced, created, updated or deleted independently of an entity type. Complex types allow entity models to group properties into common structures.

The complex type’s name is a simple identifier that MUST be unique within its schema.

A complex type can define two types of properties. A structural property is a named reference to a primitive, complex, or enumeration type, or a collection of primitive, complex, or enumeration types. A navigation property is a named reference to an entity type or a collection of entity types.

All properties MUST have a unique name within a complex type. Properties MUST NOT have the same name as the declaring complex type. They MAY have the same name as one of the direct or indirect base types or derived types.

"Dimensions": {
  "$Kind": "ComplexType",
  "Height": {
    "$Type": "Edm.Decimal",
    "$Scale": 0
  },
  "Weight": {
    "$Type": "Edm.Decimal",
    "$Scale": 0
  },
  "Length": {
    "$Type": "Edm.Decimal",
    "$Scale": 0
  }
},
"Product": {
  …
  "ProductDimensions": {
    "$Nullable": true,
    "$Type": "self.Dimensions"
  },
  "ShippingDimensions": {
    "$Nullable": true,
    "$Type": "self.Dimensions"
  }
},
"ShipmentBox": {
  …
  "Dimensions": {
    "$Nullable": true,
    "$Type": "self.Dimensions"
  }
}

9.1 Derived Complex Type

A complex type can inherit from another complex type by specifying it as its base type.

A complex type inherits the structural and navigation properties of its base type.

A complex type MUST NOT introduce an inheritance cycle by specifying a base type.

The rules for annotations of derived complex types are described in section 14.2.

9.2 Abstract Complex Type

A complex type MAY indicate that it is abstract and cannot have instances.

9.3 Open Complex Type

A complex type MAY indicate that it is open and allows clients to add properties dynamically to instances of the type by specifying uniquely named property values in the payload used to insert or update an instance of the type.

A complex type derived from an open complex type MUST indicate that it is also open.

Note: structural and navigation properties MAY be returned by the service on instances of any structured type, whether or not the type is marked as open. Clients MUST always be prepared to deal with additional properties on instances of any structured type, see OData‑Protocol.

10 Enumeration Type

Enumeration types are nominal types that represent a non-empty series of related values. Enumeration types expose these related values as members of the enumeration.

The enumeration type’s name is a simple identifier that MUST be unique within its schema.

Although enumeration types have an underlying numeric value, the preferred representation for an enumeration value is the member name. Discrete sets of numeric values should be represented as numeric values annotated with the AllowedValues annotation defined in OData-VocCore.

Enumeration types marked as flags allow values that consist of more than one enumeration member at a time.

"FileAccess": {
  "$Kind": "EnumType",
  "$UnderlyingType": "Edm.Int32",
  "$IsFlags": true,
  "Read": 1,
  "Write": 2,
  "Create": 4,
  "Delete": 8
}

10.1 Underlying Integer Type

An enumeration type MAY specify one of Edm.Byte, Edm.SByte, Edm.Int16, Edm.Int32, or Edm.Int64 as its underlying type.

If not explicitly specified, Edm.Int32 is used as the underlying type.

10.2 Flags Enumeration Type

An enumeration type MAY indicate that the enumeration type allows multiple members to be selected simultaneously.

If not explicitly specified, only one enumeration type member MAY be selected simultaneously.

"Pattern": {
  "$Kind": "EnumType",
  "$UnderlyingType": "Edm.Int32",
  "$IsFlags": true,
  "Plain": 0,
  "Red": 1,
  "Blue": 2,
  "Yellow": 4,
  "Solid": 8,
  "Striped": 16,
  "SolidRed": 9,
  "SolidBlue": 10,
  "SolidYellow": 12,
  "RedBlueStriped": 19,
  "RedYellowStriped": 21,
  "BlueYellowStriped": 22
}

10.3 Enumeration Type Member

Enumeration type values consist of discrete members.

Each member is identified by its name, a simple identifier that MUST be unique within the enumeration type. Names are case-sensitive, but service authors SHOULD NOT choose names that differ only in case.

Each member MUST specify an associated numeric value that MUST be a valid value for the underlying type of the enumeration type.

Enumeration types can have multiple members with the same value. Members with the same numeric value compare as equal, and members with the same numeric value can be used interchangeably.

Enumeration members are sorted by their numeric value.

For flag enumeration types the combined numeric value of simultaneously selected members is the bitwise OR of the discrete numeric member values.

"ShippingMethod": {
  "$Kind": "EnumType",
  "FirstClass": 0,
  "FirstClass@Core.Description": "Shipped with highest priority",
  "TwoDay": 1,
  "TwoDay@Core.Description": "Shipped within two days",
  "Overnight": 2,
  "Overnight@Core.Description": "Shipped overnight",
  "@Core.Description": "Method of shipping"
}

11 Type Definition

A type definition defines a specialization of one of the primitive types or of the built-in abstract type Edm.PrimitiveType.

The type definition’s name is a simple identifier that MUST be unique within its schema.

Type definitions can be used wherever a primitive type is used (other than as the underlying type in a new type definition) and are type-comparable with their underlying types and any type definitions defined using the same underlying type.

It is up to the definition of a term to specify whether and how annotations with this term propagate to places where the annotated type definition is used, and whether they can be overridden.

"Length": {
  "$Kind": "TypeDefinition",
  "$UnderlyingType": "Edm.Int32",
  "@Measures.Unit": "Centimeters"
},
"Weight": {
  "$Kind": "TypeDefinition",
  "$UnderlyingType": "Edm.Int32",
  "@Measures.Unit": "Kilograms"
},
"Size": {
  "$Kind": "ComplexType",
  "Height": {
    "$Nullable": true,
    "$Type": "self.Length"
  },
  "Weight": {
    "$Nullable": true,
    "$Type": "self.Weight"
  }
}

11.1 Underlying Primitive Type

The underlying type of a type definition MUST be a primitive type that MUST NOT be another type definition.

The type definition MAY specify facets applicable to the underlying type. Possible facets are: $MaxLength, $Unicode, $Precision, $Scale, or $SRID.

Additional facets appropriate for the underlying type MAY be specified when the type definition is used but the facets specified in the type definition MUST NOT be re-specified.

For a type definition with underlying type Edm.PrimitiveType no facets are applicable, neither in the definition itself nor when the type definition is used, and these should be ignored by the client.

Where type definitions are used, the type definition is returned in place of the primitive type wherever the type is specified in a response.

12 Action and Function

12.1 Action

Actions are service-defined operations that MAY have observable side effects and MAY return a single instance or a collection of instances of any type.

The action’s name is a simple identifier that MUST be unique within its schema.

Actions cannot be composed with additional path segments.

An action MAY specify a return type that MUST be a primitive, entity or complex type, or a collection of primitive, entity or complex types in scope.

An action MAY define parameters used during the execution of the action.

12.2 Action Overloads

Bound actions support overloading (multiple actions having the same name within the same schema) by binding parameter type. The combination of action name and the binding parameter type MUST be unique within a schema.

Unbound actions do not support overloads. The names of all unbound actions MUST be unique within a schema.

An unbound action MAY have the same name as a bound action.

12.3 Function

Functions are service-defined operations that MUST NOT have observable side effects and MUST return a single instance or a collection of instances of any type.

The function’s name is a simple identifier that MUST be unique within its schema.

Functions MAY be composable.

The function MUST specify a return type which MUST be a primitive, entity or complex type, or a collection of primitive, entity or complex types in scope.

A function MAY define parameters used during the execution of the function.

12.4 Function Overloads

Bound functions support overloading (multiple functions having the same name within the same schema) subject to the following rules:

• The combination of function name, binding parameter type, and unordered set of non-binding parameter names MUST be unique within a schema.
• The combination of function name, binding parameter type, and ordered set of parameter types MUST be unique within a schema.
• All bound functions with the same function name and binding parameter type within a schema MUST specify the same return type.

Unbound functions support overloading subject to the following rules:

• The combination of function name and unordered set of parameter names MUST be unique within a schema.
• The combination of function name and ordered set of parameter types MUST be unique within a schema.
• All unbound functions with the same function name within a schema MUST specify the same return type.

An unbound function MAY have the same name as a bound function.

Note that type definitions can be used to disambiguate overloads for both bound and unbound functions, even if they specify the same underlying type.

12.5 Bound or Unbound Action or Function Overloads

An action or function overload MAY indicate that it is bound. If not explicitly indicated, it is unbound.

Bound actions or functions are invoked on resources matching the type of the binding parameter. The binding parameter can be of any type, and it MAY be nullable.

Unbound actions are invoked from the entity container through an action import.

Unbound functions are invoked as static functions within a common expression (see OData-URL, section 5.1.1), or from the entity container through a function import.

12.6 Entity Set Path

Bound actions and functions that return an entity or a collection of entities MAY specify an entity set path if the entity set of the returned entities depends on the entity set of the binding parameter value.

The entity set path consists of a series of segments joined together with forward slashes.

The first segment of the entity set path MUST be the name of the binding parameter. The remaining segments of the entity set path MUST represent navigation segments or type casts.

A navigation segment names the simple identifier of the navigation property to be traversed. A type-cast segment names the qualified name of the entity type that should be returned from the type cast.

12.7 Composable Function

A function MAY indicate that it is composable. If not explicitly indicated, it is not composable.

A composable function can be invoked with additional path segments or key predicates appended to the resource path that identifies the composable function, and with system query options as appropriate for the type returned by the composable function.

12.8 Return Type

The return type of an action or function overload MAY be any type in scope, or a collection of any type in scope.

The facets MaxLength, Precision, Scale, and SRID can be used as appropriate to specify value restrictions of the return type, as well as the Unicode facet for 4.01 and greater payloads.

For a single-valued return type the facets apply to the returned value. For a collection-valued return type the facets apply to the items in the returned collection.

12.9 Parameter

An action or function overload MAY specify parameters.

A bound action or function overload MUST specify at least one parameter; the first parameter is its binding parameter. The order of parameters MUST NOT change unless the schema version changes.

Each parameter MUST have a name that is a simple identifier. The parameter name MUST be unique within the action or function overload.

The parameter MUST specify a type. It MAY be any type in scope, or a collection of any type in scope.

The facets MaxLength, Precision, Scale, or SRID can be used as appropriate to specify value restrictions of the parameter, as well as the Unicode facet for 4.01 and greater payloads.

For single-valued parameters the facets apply to the parameter value. If the parameter value is a collection, the facets apply to the items in the collection.

"TopSellingProducts": [
  {
    "$Kind": "Function",
    "$Parameter": [
      {
        "$Name": "Year",
        "$Nullable": true,
        "$Type": "Edm.Decimal",
        "$Precision": 4,
        "$Scale": 0
      }
    ],
    "$ReturnType": {
      "$Collection": true,
      "$Type": "self.Product"
    }
  }
]

13 Entity Container

Each metadata document used to describe an OData service MUST define exactly one entity container.

The entity container’s name is a simple identifier that MUST be unique within its schema.

Entity containers define the entity sets, singletons, function and action imports exposed by the service.

Entity set, singleton, action import, and function import names MUST be unique within an entity container.

An entity set allows access to entity type instances. Simple entity models frequently have one entity set per entity type.

"Products": {
  "$Collection": true,
  "$Type": "self.Product"
},
"Categories": {
  "$Collection": true,
  "$Type": "self.Category"
}

Other entity models may expose multiple entity sets per type.

"StandardCustomers": {
  "$Collection": true,
  "$Type": "self.Customer",
  "$NavigationPropertyBinding": {
    "Orders": "Orders"
  }
},
"PreferredCustomers": {
  "$Collection": true,
  "$Type": "self.Customer",
  "$NavigationPropertyBinding": {
    "Orders": "Orders"
  }
},
"Orders": {
  "$Collection": true,
  "$Type": "self.Order"
}

There are separate entity sets for standard customers and preferred customers, but only one entity set for orders. The entity sets for standard customers and preferred customers both have navigation property bindings to the orders entity set, but the orders entity set does not have a navigation property binding for the Customer navigation property, since it could lead to either set of customers.

An entity set can expose instances of the specified entity type as well as any entity type inherited from the specified entity type.

A singleton allows addressing a single entity directly from the entity container without having to know its key, and without requiring an entity set.

A function import or an action import is used to expose a function or action defined in an entity model as a top level resource.

"DemoService": {
  "$Kind": "EntityContainer",
  "Products": {
    "$Collection": true,
    "$Type": "self.Product",
    "$NavigationPropertyBinding": {
      "Category": "Categories",
      "Supplier": "Suppliers"
    },
    "@UI.DisplayName": "Product Catalog"
  },
  "Categories": {
    "$Collection": true,
    "$Type": "self.Category",
    "$NavigationPropertyBinding": {
      "Products": "Products"
    }
  },
  "Suppliers": {
    "$Collection": true,
    "$Type": "self.Supplier",
    "$NavigationPropertyBinding": {
      "Products": "Products"
    },
    "@UI.DisplayName": "Supplier Directory"
  },
  "MainSupplier": {
    "$Type": "self.Supplier"
  },
  "LeaveRequestApproval": {
    "$Action": "self.Approval"
  },
  "ProductsByRating": {
    "$EntitySet": "Products",
    "$Function": "self.ProductsByRating"
  }
}

13.1 Extending an Entity Container

An entity container MAY specify that it extends another entity container in scope. All children of the “base” entity container are added to the “extending” entity container.

If the “extending” entity container defines an entity set with the same name as defined in any of its “base” containers, then the entity set’s type MUST specify an entity type derived from the entity type specified for the identically named entity set in the “base” container. The same holds for singletons. Action imports and function imports cannot be redefined, nor can the “extending” container define a child with the same name as a child of a different kind in a “base” container.

Note: services should not introduce cycles by extending entity containers. Clients should be prepared to process cycles introduced by extending entity containers.

"Extending": {
  "$Kind": "EntityContainer",
  "$Extends": "Some.Other.Schema.Base",
   …
}

13.2 Entity Set

Entity sets are top-level collection-valued resources.

An entity set is identified by its name, a simple identifier that MUST be unique within its entity container.

An entity set MUST specify a type that MUST be an entity type in scope.

An entity set MUST contain only instances of its specified entity type or its subtypes. The entity type MAY be abstract but MUST have a key defined.

An entity set MAY indicate whether it is included in the service document. If not explicitly indicated, it is included.

Entity sets that cannot be queried without specifying additional query options SHOULD NOT be included in the service document.

13.3 Singleton

Singletons are top-level single-valued resources.

A singleton is identified by its name, a simple identifier that MUST be unique within its entity container.

A singleton MUST specify a type that MUST be an entity type in scope.

A singleton MUST reference an instance its entity type.

13.4 Navigation Property Binding

If the entity type of an entity set or singleton declares navigation properties, a navigation property binding allows describing which entity set or singleton will contain the related entities.

An entity set or a singleton SHOULD contain a navigation property binding for each non-containment navigation property that can be reached from the entity type through a sequence of type casts, complex properties, or containment navigation properties.

If omitted, clients MUST assume that the target entity set or singleton can vary per related entity.

13.4.1 Navigation Property Path Binding

A navigation property binding MUST specify a path to a navigation property of the entity set’s or singleton’s declared entity type, or a navigation property reached through a chain of type casts, complex properties, or containment navigation properties. If the navigation property is defined on a subtype, the path MUST contain the qualified name of the subtype, followed by a forward slash, followed by the navigation property name. If the navigation property is defined on a complex type used in the definition of the entity set’s entity type, the path MUST contain a forward-slash separated list of complex property names and qualified type names that describe the path leading to the navigation property.

The path can traverse one or more containment navigation properties, but the last navigation property segment MUST be a non-containment navigation property and there MUST NOT be any non-containment navigation properties prior to the final navigation property segment.

If the path traverses collection-valued complex properties or collection-valued containment navigation properties, the binding applies to all items of these collections.

If the path contains a recursive sub-path (i.e. a path leading back to the same structured type, the binding applies recursively to any positive number of cycles through that sub-path.

OData 4.01 services MAY have a type-cast segment as the last path segment, allowing to bind instances of different sub-types to different targets.

The same navigation property path MUST NOT be specified in more than one navigation property binding; navigation property bindings are only used when all related entities are known to come from a single entity set. Note that it is possible to have navigation property bindings for paths that differ only in a type-cast segment, allowing to bind instances of different sub-types to different targets. If paths differ only in type-cast segments, the most specific path applies.

13.4.2 Binding Target

A navigation property binding MUST specify a target via a simple identifier or target path. It specifies the entity set, singleton, or containment navigation property that contains the related entities.

If the target is a simple identifier, it MUST resolve to an entity set or singleton defined in the same entity container.

If the target is a target path, it MUST resolve to an entity set, singleton, or direct or indirect containment navigation property of a singleton in scope. The path can traverse single-valued containment navigation properties or single-valued complex properties before ending in a containment navigation property, and there MUST NOT be any non-containment navigation properties prior to the final segment.

"Categories": {
  "$Collection": true,
  "$Type": "self.Category",
  "$NavigationPropertyBinding": {
    "Products": "SomeSet"
  }
}

"Categories": {
  "$Collection": true,
  "$Type": "self.Category",
  "$NavigationPropertyBinding": {
    "Products": "SomeModel.SomeContainer/SomeSet"
  }
}

"Categories": {
  "$Collection": true,
  "$Type": "self.Category",
  "$NavigationPropertyBinding": {
    "Products/Supplier": "Suppliers"
  }
}

13.5 Action Import

Action imports sets are top-level resources that are never included in the service document.

An action import is identified by its name, a simple identifier that MUST be unique within its entity container.

An action import MUST specify the name of an unbound action in scope.

If the imported action returns an entity or a collection of entities, a simple identifier or target path value MAY be specified to identify the entity set that contains the returned entities. If a simple identifier is specified, it MUST resolve to an entity set defined in the same entity container. If a target path is specified, it MUST resolve to an entity set in scope.

13.6 Function Import

Function imports sets are top-level resources.

A function import is identified by its name, a simple identifier that MUST be unique within its entity container.

A function import MUST specify the name of an unbound function in scope. All unbound overloads of the imported function can be invoked from the entity container.

If the imported function returns an entity or a collection of entities, a simple identifier or target path value MAY be specified to identify the entity set that contains the returned entities. If a simple identifier is specified, it MUST resolve to an entity set defined in the same entity container. If a target path is specified, it MUST resolve to an entity set in scope.

A function import for a parameterless function MAY indicate whether it is included in the service document. If not explicitly indicated, it is not included.

14 Vocabulary and Annotation

Vocabularies and annotations provide the ability to annotate metadata as well as instance data, and define a powerful extensibility point for OData. An annotation applies a term to a model element and defines how to calculate a value for the applied term.

Metadata annotations are terms applied to model elements. Behaviors or constraints described by a metadata annotation must be consistent with the annotated model element. Such annotations define additional behaviors or constraints on the model element, such as a service, entity type, property, function, action, or parameter. For example, a metadata annotation may define ranges of valid values for a particular property. Metadata annotations are applied in CSDL documents describing or referencing an entity model.

Instance annotations are terms applied to a particular instance within an OData payload, such as described in OData-JSON. An instance annotation can be used to define additional information associated with a particular result, entity, property, or error. For example, whether a property is read-only for a particular instance. Where the same annotation is defined at both the metadata and instance level, the instance-level annotation overrides the annotation specified at the metadata level. Annotations that apply across instances should be specified as metadata annotations.

A vocabulary is a schema containing a set of terms where each term is a named metadata extension. Anyone can define a vocabulary (a set of terms) that is scenario-specific or company-specific; more commonly used terms can be published as shared vocabularies such as the OData Core vocabulary OData-VocCore.

A term can be used to:

• Extend model elements and type instances with additional information. Type instances are instances of a primitive type, including type definitions and enumeration types, of an entity type, or of a complex type.
• Map instances of annotated structured types to an interface defined by the term type; i.e. annotations allow viewing instances of a structured type as instances of a differently structured type specified by the applied term.

A service SHOULD NOT require a client to interpret annotations. Clients SHOULD ignore invalid or unknown terms and silently treat unexpected or invalid values (including invalid type, invalid literal expression, invalid targets, etc.) as an unknown value for the term. Unknown or invalid annotations should never result in an error, as long as the payload remains well-formed.

"Product": {
  "$Kind": "EntityType",
  "$Key": [
    "ID"
  ],
  "ID": {
    "$Type": "Edm.Int32"
  },
  "Name": {
    "$Nullable": true
  },
  "Description": {
    "$Nullable": true
  },
  "@UI.DisplayName": {
    "$Path": "Name"
  },
  "@SearchVocabulary.SearchResult": {
    "Title": {
      "$Path": "Name"
    },
    "Abstract": {
      "$Path": "Description"
    },
    "Url": {
      "$Apply": [
        "Products(",
        {
          "$Apply": [
            {
              "$Path": "ID"
            },
            ")"
          ],
          "$Function": "odata.concat"
        }
      ],
      "$Function": "odata.concat"
    }
  }
}

14.1 Term

A term allows annotating a model element or OData resource representation with additional data.

The term’s name is a simple identifier that MUST be unique within its schema.

The term’s type MUST be a type in scope, or a collection of a type in scope.

14.1.1 Specialized Term

A term MAY specialize another term in scope by specifying it as its base term.

When applying a specialized term, the base term MUST also be applied with the same qualifier, and so on until a term without a base term is reached.

14.1.2 Applicability

Applicability specifies a list of model elements to which the term MAY be applied. If no list is supplied, the term MAY be applied to any model element. The list of model elements MAY be extended in future versions of the vocabulary. As the intended usage may evolve over time, clients SHOULD be prepared for any term to be applied to any model element and SHOULD be prepared to handle unknown values within the list of model constructs. Applicability is expressed using the following symbolic values:

"IsURL": {
  "$Kind": "Term",
  "$Type": "Core.Tag",
  "$DefaultValue": true,
  "$AppliesTo": [
    "Property",
    "Term"
  ],
  "@Core.Description": "Properties and terms annotated with this term MUST contain a valid URL",
  "@Core.RequiresType": "Edm.String"
}

14.2 Annotation

An annotation applies a term to a model element and defines how to calculate a value for the term application. Both term and model element MUST be in scope. Section 14.1.2 specifies which model elements MAY be annotated with a term.

The value of an annotation is specified as an annotation expression, which is either a constant expression representing a constant value, or a dynamic expression. The most common construct for assigning an annotation value is a path expression that refers to a property of the same or a related structured type.

"AmountInReportingCurrency": {
  "$Nullable": true,
  "$Type": "Edm.Decimal",
  "$Scale": 0,
  "@Measures.ISOCurrency": "USD",
  "@Measures.ISOCurrency@Core.Description": "The parent company's currency"
},
"AmountInTransactionCurrency": {
  "$Nullable": true,
  "$Type": "Edm.Decimal",
  "$Scale": 0,
  "@Measures.ISOCurrency": {
    "$Path": "Currency"
  }
},
"Currency": {
  "$Nullable": true,
  "$MaxLength": 3
}

If an entity type or complex type is annotated with a term that itself has a structured type, an instance of the annotated type may be viewed as an “instance” of the term, and the qualified term name may be used as a term-cast segment in path expressions.

Structured types “inherit” annotations from their direct or indirect base types. If both the type and one of its base types is annotated with the same term and qualifier, the annotation on the type completely replaces the annotation on the base type; structured or collection-valued annotation values are not merged. Similarly, properties of a structured type inherit annotations from identically named properties of a base type.

It is up to the definition of a term to specify whether and how annotations with this term propagate to places where the annotated model element is used, and whether they can be overridden. E.g. a “Label” annotation for a UI can propagate from a type definition to all properties using that type definition and may be overridden at each property with a more specific label, whereas an annotation marking a type definition as containing a phone number will propagate to all using properties but may not be overridden.

14.2.1 Qualifier

A term can be applied multiple times to the same model element by providing a qualifier to distinguish the annotations. The qualifier is a simple identifier.

The combination of target model element, term, and qualifier uniquely identifies an annotation.

"@UI.DisplayName#Tablet": {
  "$Path": "FirstName"
}

14.2.2 Target

The target of an annotation is the model element the term is applied to.

The target of an annotation MAY be specified indirectly by “nesting” the annotation within the model element. Whether and how this is possible is described per model element in this specification.

The target of an annotation MAY also be specified directly; this allows defining an annotation in a different schema than the targeted model element.

This external targeting is only possible for model elements that are uniquely identified within their parent, and all their ancestor elements are uniquely identified within their parent.

These are the direct children of a schema with a unique name (i.e. except actions and functions whose overloads to not possess a natural identifier), and all direct children of an entity container.

MySchema.MyAction(MySchema.MyBindingType) 
MySchema.MyAction(Collection(MySchema.BindingType)) 
MySchema.MyAction()

MySchema.MyAction

MySchema.MyEntityContainer/MyActionImport

MySchema.MyEntityType/@MyVocabulary.MyTerm 
MySchema.MyEntityType/@MyVocabulary.MyTerm#MyQualifier

MySchema.MyComplexType

MySchema.MyEntityContainer

MySchema.MyEntityContainer/MyEntitySet

MySchema.MyEntityType

MySchema.MyEnumType

MySchema.MyEnumType/MyMember

MySchema.MyFunction(MySchema.MyBindingParamType, 
  First.NonBinding.ParamType) 
MySchema.MyFunction(First.NonBinding.ParamType, 
  Second.NonBinding.ParamType)

MySchema.MyFunction

MySchema.MyEntityContainer/MyFunctionImport

MySchema.MyEntityContainer/MyEntitySet 
  /MyNavigationProperty 
MySchema.MyEntityContainer/MyEntitySet 
  /MySchema.MyEntityType/MyNavProperty 
MySchema.MyEntityContainer/MyEntitySet 
  /MyComplexProperty/MyNavProperty 
MySchema.MyEntityContainer/MySingleton 
  /MyComplexProperty/MyNavProperty

MySchema.MyEntityType/MyNavigationProperty 
MySchema.MyComplexType/MyNavigationProperty

MySchema.MyEntityContainer/MyFunctionImport/MyParameter

MySchema.MyFunction/MyParameter

MySchema.MyEntityContainer/MyEntitySet 
 /MyProperty 
MySchema.MyEntityContainer/MyEntitySet 
 /MySchema.MyEntityType/MyProperty 
MySchema.MyEntityContainer/MyEntitySet 
  /MyComplexProperty/MyProperty

MySchema.MyEntityType/MyProperty 
MySchema.MyComplexType/MyProperty

MySchema.MyEntityContainer/MyFunctionImport/$ReturnType

MySchema.MyFunction/$ReturnType 
MySchema.MyFunction(MySchema.MyBindingParamType, 
  First.NonBinding.ParamType)/$ReturnType

MySchema.MyEntityContainer/MySingleton

MySchema.MyTerm

MySchema.MyTypeDefinition

All qualified names used in a target expression MUST be in scope.

External targeting is possible for properties and navigation properties of singletons or entities in a particular entity set. These annotations override annotations on the properties or navigation properties targeted via the declaring structured type.

14.3 Constant Expression

Constant expressions allow assigning a constant value to an applied term.

14.3.1 Binary

"@UI.Thumbnail": "T0RhdGE"

14.3.2 Boolean

"@UI.ReadOnly": true

14.3.3 Date

"@vCard.birthDay": "2000-01-01"

14.3.4 DateTimeOffset

"@UI.LastUpdated": "2000-01-01T16:00:00.000Z"

14.3.5 Decimal

"@UI.Width": 3.14

"@UI.Width": "3.14"

14.3.6 Duration

"@task.duration": "P7D"

14.3.7 Enumeration Member

"@self.HasPattern": "1"

"@self.HasPattern": "Red"

"@self.HasPattern": "17"

"@self.HasPattern": "Red,Striped"

14.3.8 Floating-Point Number

"@UI.FloatWidth": 3.14

"@UI.FloatWidth": "INF"

14.3.9 Guid

"@UI.Id": "21EC2020-3AEA-1069-A2DD-08002B30309D"

14.3.10 Integer

"@An.Int": 42

"@A.Very.Long.Int": "9007199254740992"

14.3.11 String

"@UI.DisplayName": "Product Catalog"

14.3.12 Time of Day

"@UI.EndTime": "21:45:00"

14.4 Dynamic Expression

Dynamic expressions allow assigning a calculated value to an applied term.

14.4.1 Path Expressions

Path expressions allow assigning a value to an applied term or term component. There are two kinds of path expressions:

• A model path is evaluated on the entity model of a service and resolves to a model element. Model paths are used within Annotation Path, Model Element Path, Navigation Property Path, and Property Path expressions. They allow assigning values to terms or term properties of the built-in types Edm.AnnotationPath, Edm.ModelElementPath, Edm.NavigationPropertyPath, Edm.PropertyPath, and Edm.AnyPropertyPath.
• An instance path is evaluated on a type instance and its nested or related type instances and resolves to the instance or collection of instances identified by the path. Instance paths are used within Value Path expressions. They allow assigning values to terms or term properties of model-defined types or of built-in types other than the Edm.*Path types.

14.4.1.1 Path Syntax

Model paths and instance paths share a common syntax which is derived from the path expression syntax of URLs, see OData-URL.

A path MUST be composed of zero or more path segments joined together by forward slashes (/).

Paths starting with a forward slash (/) are absolute paths, and the first path segment MUST be the qualified name of a model element, e.g. an entity container. The remaining path after the second forward slash is interpreted relative to that model element.

/My.Schema.MyEntityContainer/MyEntitySet

Paths not starting with a forward slash are interpreted relative to the annotation target, following the rules specified in section “Path Evaluation”.

Address/City

If a path segment is a qualified name, it represents a type cast, and the segment MUST be the name of a type in scope. If the type or instance identified by the preceding path part cannot be cast to the specified type, the path expression evaluates to the null value.

…/self.Manager/…

If a path segment starts with an at (@) character, it represents a term cast. The at (@) character MUST be followed by a qualified name that MAY be followed by a hash (#) character and a simple identifier. The qualified name preceding the hash character MUST resolve to a term that is in scope, the simple identifier following the hash sign is interpreted as a qualifier for the term. If the model element or instance identified by the preceding path part has not been annotated with that term (and if present, with that qualifier), the term cast evaluates to the null value. Four special terms are implicitly “annotated” for media entities and stream properties:

• odata.mediaEditLink
• odata.mediaReadLink
• odata.mediaContentType
• odata.mediaEtag

…/@Capabilities.SortRestrictions/…

If a path segment is a simple identifier, it MUST be the name of a child model element of the model element identified by the preceding path part, or a structural or navigation property of the instance identified by the preceding path part. A sequence of navigation segments can traverse multiple CSDL documents. The document containing the path expression only needs to reference the next traversed document to bring the navigation target type into scope, and each traversed document in turn needs to reference only its next document.

A model path MAY contain any number of segments representing collection-valued structural or navigation properties. The result of the expression is the model element reached via this path.

…/Orders/Items/Product/…

An instance path MUST NOT contain more than one segment representing a collection-valued construct, e.g. an entity set or a collection-valued navigation property that is not followed by a key predicate, or a collection-valued structural property that is not followed by an index segment. The result of the expression is the collection of instances resulting from applying any remaining path segments that operate on a single-valued expression to each instance in the collection-valued segment.

An instance path MAY terminate in a $count segment if the previous segment is collection-valued, in which case the path evaluates to the number of items in the collection identified by the preceding segment.

…/Addresses/Street

…/Addresses/$count

A model path MAY contain path segments starting with a navigation property, then followed by an at (@) character, then followed by the qualified name of a term in scope, and optionally followed by a hash (#) character and a simple identifier which is interpreted as a qualifier for the term. If the navigation property has not been annotated with that term (and if present, with that qualifier), the path segment evaluates to the null value. This allows addressing annotations on the navigation property itself; annotations on the entity type specified by the navigation property are addressed via a term-cast segment.

…/Items@Core.Description

…/Items/@Core.Description

An instance path MAY contain path segments starting with an entity set or a collection-valued navigation property, then followed by a key predicate using parentheses-style convention, see OData-URL. The key values are either primitive literals or instance paths. If the key value is a relative instance path, it is interpreted according to the same rule below as the instance path it is part of, not relative to the instance identified by the preceding path part.

/self.container/SettingsCollection('FeatureXxx')/IsAvailable

/self.container/Products(ID=ProductID)/Name

An instance path MAY contain an index segment immediately following a path segment representing an ordered collection-valued structural property. The index is zero-based and MUST be an integer literal. Negative integers count from the end of the collection, with -1 representing the last item in the collection. Remaining path segments are evaluated relative to the identified item of the collection.

Addresses/1

Addresses/-1/Street

14.4.1.2 Path Evaluation

Annotations MAY be embedded within their target, or specified separately, e.g. as part of a different schema, and specify a path to their target model element. The latter situation is referred to as targeting in the remainder of this section.

The host of an annotation is the model element targeted by the annotation, unless that target is another annotation or a model element (collection, record or property value) directly or indirectly embedded within another annotation, in which case the host is the host of that other annotation.

If the value of an annotation is expressed dynamically with a path expression, the path evaluation rules for this expression depend upon the model element by which the annotation is hosted.

For annotations hosted by an entity container, the path is evaluated starting at the entity container, i.e. an empty path resolves to the entity container, and non-empty paths MUST start with a segment identifying a container child (entity set, function import, action import, or singleton). The subsequent segments follow the rules for path expressions targeting the corresponding child element.

For annotations hosted by an entity set or a singleton, the path is evaluated starting at the entity set or singleton, i.e. an empty path resolves to the entity set or singleton, and non-empty paths MUST follow the rules for annotations targeting the declared entity type of the entity set or singleton.

For annotations hosted by an entity type or complex type, the path is evaluated starting at the type, i.e. an empty path resolves to the type, and the first segment of a non-empty path MUST be a structural or navigation property of the type, a type cast, or a term cast.

For annotations hosted by an action, action import, function, function import, parameter, or return type, the first segment of the path MUST be the name of a parameter of the action or function or $ReturnType.

For annotations hosted by a structural or navigation property, the path evaluation rules additionally depend upon how the annotation target is specified, as follows:

1. If the annotation is directly or indirectly embedded within the hosting property, the path is evaluated starting at the directly enclosing type of the hosting property. This allows e.g. specifying the value of an annotation on one property to be calculated from values of other properties of the same enclosing type. An empty path resolves to the enclosing type, and non-empty paths MUST follow the rules for annotations targeting the directly enclosing type.

2. If the annotation uses targeting and the target expression starts with an entity container, or the annotation is directly or indirectly embedded within such an annotation, the path is evaluated starting at the declared type of the hosting property. An empty path resolves to the declared type of the property, and non-empty paths MUST follow the rules for annotations targeting the declared type of the property. If the type is primitive, the first segment of a non-empty path MUST be a type cast or a term cast.

3. If the annotation uses targeting and the target expression does not start with an entity container, or the annotation is directly or indirectly embedded within such an annotation, the path is evaluated starting at the outermost entity type or complex type named in the target expression. This allows e.g. specifying the value of an annotation on one property to be calculated from values of other properties of the outermost type. An empty path resolves to the outermost type, and the first segment of a non-empty path MUST be a structural or navigation property of the outermost type, a type cast, or a term cast.

"self": {
  "A": {
    "$Kind": "EntityType",
    "A1": {
      "$Type": "Edm.Boolean"
    },
    "A2": {
      "$Type": "self.B"
      "@Core.Description@Core.IsLanguageDependent": {
        "$Path": "A1"
      },
      "@Core.Description": "…"
    }
  },
  "B": {
    "$Kind": "ComplexType",
    "B1": {
      "$Type": "Edm.Boolean"
    }
  },

  "Container": {
    "$Kind": "EntityContainer",
    "SetA": {
      "$Collection": true,
      "$Type": "self.A"
    }
  },
  "$Annotations": {
    "self.Container/SetA/A2": {
      "@Core.Description#viaSet@Core.IsLanguageDependent": {
        "$Path": "B1"
      },
      "@Core.Description#viaSet": "…"
    },
    "self.Container/SetA/A2/@Core.Description#viaSet": {
      "@Core.IsLanguageDependent": {
        "$Path": "B1"
      }
    },

    "self.A/A2": {
      "@Core.Description#external@Core.IsLanguageDependent": {
        "$Path": "A1"
      },
      "@Core.Description#external": "…"
    },
    "self.A/A2/@Core.Description": {
      "@Core.IsLanguageDependent": {
        "$Path": "A1"
      }
    }
  }
}

14.4.1.3 Annotation Path

The annotation path expression provides a value for terms or term properties that specify the built-in types Edm.AnnotationPath or Edm.ModelElementPath. Its argument is a model path with the following restriction:

• A non-null path MUST resolve to an annotation.

The value of the annotation path expression is the path to the annotation, not its instance value. This is useful for terms that reuse or refer to other terms.

A term or term property of type Edm.AnnotationPath can be annotated with the term Validation.AllowedTerms (see OData-VocValidation) if its intended value is an annotation path that ends in a term cast with one of the listed terms.

"@UI.ReferenceFacet": "Product/Supplier/@UI.LineItem",
"@UI.CollectionFacet#Contacts": [
  "Supplier/@Communication.Contact",
  "Customer/@Communication.Contact"
]

14.4.1.4 Model Element Path

The model element path expression provides a value for terms or term properties that specify the built-in type Edm.ModelElementPath. Its argument is a model path.

The value of the model element path expression is the path itself, not the instance(s) identified by the path.

"@org.example.MyFavoriteModelElement": "/self.someAction"

14.4.1.5 Navigation Property Path

The navigation property path expression provides a value for terms or term properties that specify the built-in types Edm.NavigationPropertyPath, Edm.AnyPropertyPath, or Edm.ModelElementPath. Its argument is a model path with the following restriction:

• A non-null path MUST resolve to a model element whose type is an entity type, or a collection of entity types, e.g. a navigation property.

The value of the navigation property path expression is the path itself, not the entity or collection of entities identified by the path.

"@UI.HyperLink": "Supplier",
"@Capabilities.UpdateRestrictions": {
  "NonUpdatableNavigationProperties": [
    "Supplier",
    "Category"
  ]
}

14.4.1.6 Property Path

The property path expression provides a value for terms or term properties that specify one of the built-in types Edm.PropertyPath, Edm.AnyPropertyPath, or Edm.ModelElementPath. Its argument is a model path with the following restriction:

• A non-null path MUST resolve to a model element whose type is a primitive or complex type, an enumeration type, a type definition, or a collection of one of these types.

The value of the property path expression is the path itself, not the value of the structural property or annotation identified by the path.

"@UI.RefreshOnChangeOf": "ChangedAt",
"@Capabilities.UpdateRestrictions": {
  "NonUpdatableProperties": [
    "CreatedAt",
    "ChangedAt"
  ]
}

14.4.1.7 Value Path

The value path expression allows assigning a value by traversing an object graph. It can be used in annotations that target entity containers, entity sets, entity types, complex types, navigation properties of structured types, and structural properties of structured types. Its argument is an instance path.

The value of the path expression is the instance or collection of instances identified by the path.

"@UI.DisplayName": {
  "$Path": "FirstName"
},
"@UI.DisplayName#second": {
  "$Path": "@vCard.Address#work/FullName"
}

14.4.2 Comparison and Logical Operators

Annotations MAY use the following logical and comparison expressions which evaluate to a Boolean value. These expressions MAY be combined and they MAY be used anywhere instead of a Boolean expression.

The And and Or operators require two operand expressions that evaluate to Boolean values. The Not operator requires a single operand expression that evaluates to a Boolean value. For details on null handling for comparison operators see OData-URL.

The other comparison operators require two operand expressions that evaluate to comparable values.

{
  "$And": [
    {
      "$Path": "IsMale"
    },
    {
      "$Path": "IsMarried"
    }
  ]
},
{
  "$Or": [
    {
      "$Path": "IsMale"
    },
    {
      "$Path": "IsMarried"
    }
  ]
},
{
  "$Not": {
    "$Path": "IsMale"
  }
},
{
  "$Eq": [
    null,
    {
      "$Path": "IsMale"
    }
  ]
},
{
  "$Ne": [
    null,
    {
      "$Path": "IsMale"
    }
  ]
},
{
  "$Gt": [
    {
      "$Path": "Price"
    },
    20
  ]
},
{
  "$Ge": [
    {
      "$Path": "Price"
    },
    10
  ]
},
{
  "$Lt": [
    {
      "$Path": "Price"
    },
    20
  ]
},
{
  "$Le": [
    {
      "$Path": "Price"
    },
    100
  ]
},
{
  "$Has": [
    {
      "$Path": "Fabric"
    },
    "Red"
  ]
},
{
  "$In": [
    {
      "$Path": "Size"
    },
    [
      "XS",
      "S"
    ]
  ]
}

14.4.3 Arithmetic Operators

Annotations MAY use the following arithmetic expressions which evaluate to a numeric value. These expressions MAY be combined, and they MAY be used anywhere instead of a numeric expression of the appropriate type. The semantics and evaluation rules for each arithmetic expression is identical to the corresponding arithmetic operator defined in OData-URL.

The Neg operator requires a single operand expression that evaluates to a numeric value. The other arithmetic operators require two operand expressions that evaluate to numeric values.

{
  "$Add": [
    {
      "$Path": "StartDate"
    },
    {
      "$Path": "Duration"
    }
  ]
},
{
  "$Sub": [
    {
      "$Path": "Revenue"
    },
    {
      "$Path": "Cost"
    }
  ]
},
{
  "$Neg": {
    "$Path": "Height"
  }
},
{
  "$Mul": [
    {
      "$Path": "NetPrice"
    },
    {
      "$Path": "TaxRate"
    }
  ]
},
{
  "$Div": [
    {
      "$Path": "Quantity"
    },
    {
      "$Path": "QuantityPerParcel"
    }
  ]
},
{
  "$DivBy": [
    {
      "$Path": "Quantity"
    },
    {
      "$Path": "QuantityPerParcel"
    }
  ]
},
{
  "$Mod": [
    {
      "$Path": "Quantity"
    },
    {
      "$Path": "QuantityPerParcel"
    }
  ]
}

14.4.4 Apply Client-Side Functions

The apply expression enables a value to be obtained by applying a client-side function. The apply expression MAY have operand expressions. The operand expressions are used as parameters to the client-side function.

OData defines the following functions. Services MAY support additional functions that MUST be qualified with a namespace other than odata. Function names qualified with odata are reserved for this specification and its future versions.

14.4.4.1 Canonical Functions

All canonical functions defined in OData-URL can be used as client-side functions, qualified with the namespace odata. The semantics of these client-side functions is identical to their counterpart function defined in OData-URL.

For example, the odata.concat client-side function takes two expressions as arguments. Each argument MUST evaluate to a primitive or enumeration type. It returns a value of type Edm.String that is the concatenation of the literal representations of the results of the argument expressions. Values of primitive types other than Edm.String are represented according to the appropriate alternative in the primitiveValue rule of OData-ABNF, i.e. Edm.Binary as binaryValue, Edm.Boolean as booleanValue etc.

"@UI.DisplayName": {
  "$Apply": [
    "Product: ",
    {
      "$Apply": [
        {
          "$Path": "ProductName"
        },
        {
          "$Apply": [
            " (",
            {
              "$Apply": [
                {
                  "$Path": "Available/Quantity"
                },
                {
                  "$Apply": [
                    " ",
                    {
                      "$Apply": [
                        {
                          "$Path": "Available/Unit"
                        },
                        " available)"
                      ],
                      "$Function": "odata.concat"
                    }
                  ],
                  "$Function": "odata.concat"
                }
              ],
              "$Function": "odata.concat"
            }
          ],
          "$Function": "odata.concat"
        }
      ],
      "$Function": "odata.concat"
    }
  ],
  "$Function": "odata.concat"
}

ProductName is of type String, Quantity in complex type Available is of type Decimal, and Unit in Available is of type enumeration, so the result of the Path expression is represented as the member name of the enumeration value.

14.4.4.2 Function odata.fillUriTemplate

The odata.fillUriTemplate client-side function takes two or more expressions as arguments and returns a value of type Edm.String.

The first argument MUST be of type Edm.String and specifies a URI template according to RFC6570, the other arguments MUST be labeled element expressions. Each labeled element expression specifies the template parameter name as its name and evaluates to the template parameter value.

RFC6570 defines three kinds of template parameters: simple values, lists of values, and key-value maps.

Simple values are represented as labeled element expressions that evaluate to a single primitive value. The literal representation of this value according to OData-ABNF is used to fill the corresponding template parameter.

Lists of values are represented as labeled element expressions that evaluate to a collection of primitive values.

Key-value maps are represented as labeled element expressions that evaluate to a collection of complex types with two properties that are used in lexicographic order. The first property is used as key, the second property as value.

{
  "$Apply": [
    "http://host/someAPI/Actors/{actorName}/CV",
    {
      "$LabeledElement": {
        "$Path": "Actor/Name"
      },
      "$Name": "self.actorName"
    }
  ],
  "$Function": "odata.fillUriTemplate"
}

14.4.4.3 Function odata.matchesPattern

The odata.matchesPattern client-side function takes two string expressions as arguments and returns a Boolean value. It is the counterpart of the identically named URL function OData-URL, section 5.1.1.7.1.

The function returns true if the second expression evaluates to an ECMAScript (JavaScript) regular expression and the result of the first argument expression matches that regular expression, using syntax and semantics of ECMAScript regular expressions.

{
  "$Apply": [
    {
      "$Path": "FirstName"
    },
    "^[^b-d]+$"
  ],
  "$Function": "odata.matchesPattern"
}

14.4.4.4 Function odata.uriEncode

The odata.uriEncode client-side function takes one argument of primitive type and returns the URL-encoded OData literal that can be used as a key value in OData URLs or in the query part of OData URLs.

Note: string literals are surrounded by single quotes as required by the parentheses-style key syntax.

{
  "$Apply": [
    "http://host/service/Genres({genreName})",
    {
      "$LabeledElement": {
        "$Apply": [
          {
            "$Path": "NameOfMovieGenre"
          }
        ],
        "$Function": "odata.uriEncode"
      },
      "$Name": "self.genreName"
    }
  ],
  "$Function": "odata.fillUriTemplate"
}

14.4.5 Cast

The cast expression casts the value obtained from its single child expression to the specified type. The cast expression follows the same rules as the cast canonical function defined in OData-URL.

"@UI.Threshold": {
  "$Cast": {
    "$Path": "Average"
  },
  "$Type": "Edm.Decimal"
}

14.4.6 Collection

The collection expression enables a value to be obtained from zero or more item expressions. The value calculated by the collection expression is the collection of the values calculated by each of the item expressions. The values of the child expressions MUST all be type compatible.

"@seo.SeoTerms": [
  "Product",
  "Supplier",
  "Customer"
]

14.4.7 If-Then-Else

The if-then-else expression enables a value to be obtained by evaluating a condition expression. It MUST contain exactly three child expressions. There is one exception to this rule: if and only if the if-then-else expression is an item of a collection expression, the third child expression MAY be omitted, reducing it to an if-then expression. This can be used to conditionally add an element to a collection.

The first child expression is the condition and MUST evaluate to a Boolean result, e.g. the comparison and logical operators can be used.

The second and third child expressions are evaluated conditionally. The result MUST be type compatible with the type expected by the surrounding expression.

If the first expression evaluates to true, the second expression MUST be evaluated and its value MUST be returned as the result of the if-then-else expression. If the first expression evaluates to false and a third child element is present, it MUST be evaluated and its value MUST be returned as the result of the if-then-else expression. If no third expression is present, nothing is added to the surrounding collection.

"@person.Gender": {
  "$If": [
    {
      "$Path": "IsFemale"
    },
    "Female",
    "Male"
  ]
}

14.4.8 Is-Of

The is-of expression checks whether the value obtained from its single child expression is compatible with the specified type. It returns true if the child expression returns a type that is compatible with the specified type, and false otherwise.

"@Self.IsPreferredCustomer": {
  "$IsOf": {
    "$Path": "Customer"
  },
  "$Type": "self.PreferredCustomer"
}

14.4.9 Labeled Element

The labeled element expression assigns a name to its single child expression. The value of the child expression can then be reused elsewhere with a labeled element reference expression.

A labeled element expression MUST contain exactly one child expression. The value of the child expression is also the value of the labeled element expression.

A labeled element expression MUST provide a simple identifier value as its name that MUST be unique within the schema containing the expression.

"@UI.DisplayName": {
  "$LabeledElement": {
    "$Path": "FirstName"
  },
  "$Name": "CustomerFirstName"
}

14.4.10 Labeled Element Reference

The labeled element reference expression MUST specify the qualified name of a labeled element expression in scope and returns the value of the identified labeled element expression as its value.

"@UI.DisplayName": {
  "$LabeledElementReference": "self.CustomerFirstName"
}

14.4.11 Null

The null expression indicates the absence of a value. The null expression MAY be annotated.

"@UI.DisplayName": null,

"@UI.Address": {
  "$Null": null,
  "@self.Reason": "Private"
}