Serving OData APIs

• Features Overview
• Mapping of CDS Types
  • Overriding Type Mapping for UUIDs
• OData Annotations
  • Terms and Properties
  • Qualified Annotations
  • Primitives
  • Records
  • Collections
  • References
  • Enumeration Values
  • Annotating Annotations
  • Dynamic Expressions
  • sap: Annotations
  • Differences to ABAP
• OData Vocabularies
  • OASIS Vocabularies
  • SAP Vocabularies
• Data Aggregation
  • Example
  • Transformations
  • Aggregation Methods
  • Custom Aggregates
  • Other Features
• OData V2 Support
  • Enabling OData V2 via Proxy in Node.js Apps
  • Using OData V2 in Java Apps
• Miscellaneous
  • Omitting Elements from APIs
  • Absolute Context URL

Features Overview

OData is an OASIS standard, which essentially enhances plain REST by standardized query options like $select, $expand, $filter, etc. Find a rough overview of the feature coverage in the following table.

• ⁽¹⁾ The elements to be searched are specified with the @cds.search annotation.
• ⁽²⁾ Current limitation: Navigation path identifying the collection can only contain one segment.

Learn more in the Getting Started guide on odata.org. Learn more in the tutorials Take a Deep Dive into OData.

Mapping of CDS Types

The table below lists CDS’s built-in types and their mapping to OData EDM type system.

… with these differences for OData V2:

Overriding Type Mapping for UUIDs

As shown above, CDS UUIDs map to Edm.Guid by default. However, the OData standard puts up restrictive rules for Edm.Guid values - for example, only hyphenated strings are allowed - which can conflict with existing data. Therefore, we allow the default mapping to be overridden as follows:

entity Books {
  key ID : UUID @odata.Type:'Edm.String';
  ...
}

If necessary, you can also add the annotation @odata.MaxLength to override the corresponding property.

OData Annotations

The following sections explain how to add OData annotations to CDS models and how they’re mapped to EDMX outputs.

Terms and Properties

OData defines a strict two-fold key structure composed of @<Vocabulary>.<Term> and all annotations are always specified as a Term with either a primitive value, a record value, or collection values. The properties themselves may, in turn, be primitives, records, or collections.

Example

@Common.Label: 'Customer'
@Common.ValueList: {
  Label: 'Customers',
  CollectionPath: 'Customers'
}
entity Customers { }

This is represented in CSN as follows:

{"definitions":{
  "Customers":{
    "kind": "entity",
    "@Common.Label": "Customer",
    "@Common.ValueList.Label": "Customers",
    "@Common.ValueList.CollectionPath": "Customers"
  }
}}

And would render to EDMX as follows:

<Annotations Target="MyService.Customers">
  <Annotation Term="Common.Label" String="Customer"/>
  <Annotation Term="Common.ValueList">
    <Record Type="Common.ValueListType">
      <PropertyValue Property="Label" String="Customers"/>
      <PropertyValue Property="CollectionPath" String="Customers"/>
    </Record>
  </Annotation>
</Annotations>

The value for @Common.ValueList is flattened to individual key-value pairs in CSN and ‘restructured’ to a record for OData exposure in EDMX.

The rules for this restructuring from CSN sources are:

For each annotated target definition in CSN:

1. Annotations with a single-identifier key are skipped (as OData annotations always have @Vocabulary.Term... key signature).
2. All individual annotations with the same @<Vocabulary.Term> prefix are collected.
3. If there is only one without suffix, → that one is a scalar or array value of an OData term.
4. If there are more with suffix key parts →, it’s a record value for the OData term.

Qualified Annotations

OData foresees qualified annotations, which essentially allow to specify different values for a given property. CDS syntax for annotations was extended to also allow appending OData-style qualifiers after a # sign to an annotation key. However, always only as the last component of a key in the syntax.

For example, this is supported:

@Common.Label: 'Customer'
@Common.Label#Legal: 'Client'
@Common.Label#Healthcare: 'Patient'
@Common.ValueList: {
  Label: 'Customers',
  CollectionPath:'Customers'
}
@Common.ValueList#Legal: {
  Label: 'Clients',
  CollectionPath:'Clients'
}

and would render as follows in CSN:

{
  "@Common.Label": "Customer",
  "@Common.Label#Legal": "Clients",
  "@Common.Label#Healthcare": "Patients",
  "@Common.ValueList.Label": "Customers",
  "@Common.ValueList.CollectionPath": "Customers",
  "@Common.ValueList#Legal.Label": "Clients",
  "@Common.ValueList#Legal.CollectionPath": "Clients",
}

Note that there’s no interpretation and no special handling to these qualifiers in CDS you have to write and apply them 1:1 as your chosen OData vocabularies specify them. Note also that something like:

@Common.ValueList#Legal.Label: "Clients" // unsupported

isn’t supported in CDS syntax → you’d always have to write that as in line three of the previous snippet, which anyways is the recommended and preferred way to reflect OData’s distinction of Terms and Properties.

Primitives

The annotation @Some isn’t a valid term definition. The following example illustrates the rendering of primitive values.

Primitive annotation values, that are, Strings, Numbers, true, false, and null are mapped to corresponding OData annotations as follows:

@Some.Null: null
@Some.Boolean: true
@Some.Integer: 1
@Some.Number: 3.14
@Some.String: 'foo'

<Annotation Term="Some.Null"><Null/></Annotation>
<Annotation Term="Some.Boolean" Bool="true"/>
<Annotation Term="Some.Integer" Int="1"/>
<Annotation Term="Some.Number" Decimal="3.14"/>
<Annotation Term="Some.String" String="foo"/>

Have a look at our sample CAP SFLIGHT, showcasing the usage of OData annotations.

Records

The annotation @Some isn’t a valid term definition. The following example illustrates the rendering of record values.

Record-like source structures are mapped to <Record> nodes in EDMX, with primitive types translated analogously to above:

@Some.Record: {
  Null: null,
  Boolean: true,
  Integer: 1,
  Number: 3.14,
  String: 'foo'
}

<Annotation Term="Some.Record">
  <Record>
    <PropertyValue Property="Null"><Null/></PropertyValue>
    <PropertyValue Property="Boolean" Bool="true"/>
    <PropertyValue Property="Integer" Int="1"/>
    <PropertyValue Property="Number" Decimal="3.14"/>
    <PropertyValue Property="String" String="foo"/>
  </Record>
</Annotation>

Frequently, you need to specify an explicit type for records in OData. Do so by adding a property named Type as the first element of the record. For example:

@UI.Identification: [
  {$Type:'UI.DataField', Value: deliveryId }
]

<Annotation Term="UI.Identification">
  <Collection>
    <Record Type="UI.DataField">
      <PropertyValue Property="Value" Path="deliveryId"/>
    </Record>
  </Collection>
</Annotation>

Have a look at our sample CAP SFLIGHT, showcasing the usage of OData annotations.

Collections

The annotation @Some isn’t a valid term definition. The following example illustrates the rendering of collection values.

Arrays are mapped to <Collection> nodes in EDMX and if primitives showing up as direct elements of the array, these elements are wrapped into individual primitive child nodes of the resulting collection as in. The rules for records and collections being applied recursively:

@Some.Collection: [
  true, 1, 3.14, 'foo',
  { $Type:'UI.DataField', Label:'Whatever', Hidden }
]

<Annotation Term="Some.Collection">
  <Collection>
    <Null/>
    <Bool>true</Bool>
    <Int>1</Int>
    <Decimal>3.14</Decimal>
    <String>foo</String>
    <Record Type="UI.DataField">
      <PropertyValue Property="Label" String="Whatever"/>
      <PropertyValue Property="Hidden" Bool="True"/>
    </Record>
  </Collection>
</Annotation>

References

The annotation @Some isn’t a valid term definition. The following example illustrates the rendering of reference values.

References in cds annotations are mapped to .Path properties or nested <Path> elements respectively:

@Some.Term: My.Reference
@Some.Record: {
  Value: My.Reference
}
@Some.Collection: [
  My.Reference
]

<Annotation Term="Some.Term" Path="My/Reference"/>
<Annotation Term="Some.Record">
  <Record>
    <PropertyValue Property="Value" Path="My/Reference"/>
  </Record>
</Annotation>
<Annotation Term="Some.Collection">
  <Collection>
    <Path>My/Reference</Path>
  </Collection>
</Annotation>

Use a dynamic expression if the generic mapping isn’t able to produce the desired <Path>:

@Some.Term: {$edmJson: {$Path: 'com.sap.foo.EntityContainer.EntityName/FieldName'}}

<Annotation Term="Some.Term">
  <Path>com.sap.foo.EntityContainer.EntityName/FieldName</Path>
</Annotation>

Enumeration Values

Enumeration symbols are mapped to corresponding EnumMember properties in OData. For example:

@Common.FieldControl: #Hidden
@Common.FilterExpressionRestrictions: [{
  Property: deliveryDate,
  AllowedExpressions: #SingleInterval,
}]

<Annotation Term="Common.FieldControl" EnumMember="Common.FieldControlType/Hidden"/>
<Annotation Term="Common.FilterExpressionRestrictions">
  <Collection>
    <Record>
      <PropertyValue Property="Property" PropertyPath="deliveryDate"/>
      <PropertyValue Property="AllowedExpressions" EnumMember="Common.FilterExpressionType/SingleInterval"/>
    </Record>
  </Collection>
</Annotation>

Annotating Annotations

OData has the possibility of annotating annotations. This often occurs in combination with enums like UI.Importance and UI.TextArrangement. CDS has no corresponding language feature. For OData annotations, nesting can be achieved in the following way:

• When a Record is to be annotated, add an additional element to the CDS source structure. The name of this element is the full name of the annotation, including the @, and needs to be quoted in order to pass the compiler.
• When a single value is to be annotated, first turn it into a structure and put the actual value into an artificial property called $value, then add the annotation as a further property.

@UI.LineItem: { ![@UI.TextArrangement]: #TextOnly,
    $value:[
            {$Type: 'UI.DataField',Value: ApplicationName},
            {$Type: 'UI.DataField',Value: Description},
            {$Type: 'UI.DataField',Value: SourceName},
            {$Type: 'UI.DataField',Value: ChangedBy},
            {$Type: 'UI.DataField',Value: ChangedAt}
    ]
 }
@Common.Text: {
  $value: Text, ![@UI.TextArrangement]: #TextOnly
}

As TextArrangement is common, there’s a shortcut for this specific situation:

...
@Common: {
  Text: Text, TextArrangement: #TextOnly
}

In both cases, the resulting EDMX is:

<Annotation Term="UI.LineItem">
  <Collection>
    <Record Type="UI.DataField">
      ...
      <Annotation Term="UI.Importance" EnumMember="UI.ImportanceType/High"/>
    </Record>
  </Collection>
</Annotation>
<Annotation Term="Common.Text" Path="Text">
  <Annotation Term="UI.TextArrangement" EnumMember="UI.TextArrangementType/TextOnly"/>
</Annotation>

Dynamic Expressions

OData supports dynamic expressions in annotations. CDS syntax doesn’t allow to write expressions in annotation values, but for OData annotations you can use the “edm-json inline mechanism” by providing a dynamic expression as defined in the JSON representation of the OData Common Schema Language enclosed in { $edmJson: { ... }}.

Note that here the CDS syntax for string literals with single quotes ('foo') applies, and that paths are not automatically recognized but need to be written as {$Path: 'fieldName'}. The CDS compiler translates the expression into the corresponding XML representation.

For example, the CDS annotation:

@UI.Hidden: {$edmJson: {$If: [{$Eq: [{$Path: 'status'}, 'visible']}, false, true]}}

is translated to:

<Annotation Term="UI.Hidden">
  <If>
    <Eq>
      <Path>status</Path>
      <String>visible</String>
    </Eq>
    <Bool>false</Bool>
    <Bool>true</Bool>
  </If>
</Annotation>

This feature is currently only available on the CAP Node.js stack.

sap: Annotations

In general, backends and SAP Fiori UIs understand or even expect OData V4 annotations. You should use those rather than the OData V2 SAP Extensions.

If necessary, CDS automatically translates OData V4 annotations to OData V2 SAP extensions when invoked with v2 as OData version. This means you shouldn’t have to care for the latter at all.

Nevertheless, in case you need to do so you can add sap:... attribute-style annotations as follows:

  @sap.applicable.path: 'to_eventStatus/EditEnabled'
  action EditEvent(...) returns SomeType;

Which would render to OData EDMX as follows:

  <FunctionImport Name="EditEvent" ...
    sap:applicable-path="to_eventStatus/EditEnabled">
    ...
  </FunctionImport>

The rules are:

• Only strings are supported as values.
• The first dot in @sap. is replaced by a colon :.
• Subsequent dots are replaced by dashes.

Differences to ABAP

In contrast to ABAP CDS, we apply a generic, isomorphic approach where names and positions of annotations are exactly as specified in the OData Vocabularies. This has the following advantages:

• Single source of truth — users only need to consult the official OData specs
• Speed — we don’t need complex case-by-case mapping logic
• No bottlenecks — we always support the full set of OData annotations
• Bidirectional mapping — we can translate CDS to EDMX and vice versa

Not the least, it also saves us lots of efforts as we don’t have to write derivate of all the OData vocabulary specs.

OData Vocabularies

OASIS Vocabularies

SAP Vocabularies

Data Aggregation

Data aggregation in OData V4 is leveraged by the system query option $apply, which defines a pipeline of transformations that is applied to the input set specified by the URI. On the result set of the pipeline the standard system query options come into effect. For data aggregation in OData V2, refer to section Aggregation.

Example

GET /Orders(10)/books?
    $apply=filter(year eq 2000)/
           groupby((author/name),aggregate(price with average as avg))/
    orderby(title)/
    top(3)

This request operates on the books of the order with ID 10. It firstly filters out the books from year 2000 to an intermediate result set. The intermediate result set is grouped by author name and the price is averaged. Finally, the result set is sorted by title and only the top 3 entries are retained.

Transformations

concat

The concat transformation applies additional transformation sequences to the input set and concatenates the result:

GET /Books?$apply=
    filter(author/name eq 'Bram Stroker')/
    concat(
        aggregate($count as totalCount),
        groupby((year), aggregate($count as countPerYear)))

This request filters all books keeping only books by Bram Stroker. From these books concat calculates (1) the total count of books and (2) the count of books per year. The result is heterogeneous.

The concat transformation must be the last of the apply pipeline. If concat is used, then $apply can’t be used in combination with other system query options.

skip, top, and orderby

Beyond the standard transformations specified by OData, CDS Java supports the transformations skip, top, and orderby that allow to sort and paginate an input set:

GET /Order(10)/books?
    $apply=orderby(price desc)/
           top(500)/
           groupBy((author/name),aggregate(price with max as maxPrice))

This query groups the 500 most expensive books by author name and determines the price of the most expensive book per author.

Aggregation Methods

Custom Aggregates

Instead of explicitly using an expression with an aggregation method in the aggregate transformation, the client may use a custom aggregate. A custom aggregate can be considered as a virtual property that aggregates the input set. It’s calculated on server side. How the custom aggregate is calculated the client doesn’t know.

They can only be used for the special case when a default aggregation method can be specified declaratively on server side for a measure.

A custom aggregate is declared in the CDS model as follows:

• The measure must be annotated with an @Aggregation.default annotation that specifies the aggregation method.
• The CDS entity should be annotated with an @Aggregation.CustomAggregate annotation to expose the custom aggregate to the client.

@Aggregation.CustomAggregate#stock : 'Edm.Decimal'
entity Books as projection on bookshop.Books{
  	ID,
  	title,

  	@Aggregation.default: #SUM
  	stock
  };

With this definition, it’s now possible to use the custom aggregate stock in an aggregate transformation:

GET /Books?$apply=aggregate(stock)

which is equivalent to:

GET /Books?$apply=aggregate(stock with sum as stock)

Currencies and Units of Measure

If a property represents a monetary amount, it may have related property that indicates the amount’s currency code. Analogously, a property representing a measured quantity can be related to a unit of measure. To indicate that a property is a currency code or a unit of measure it can be annotated with the Semantics Annotations @Semantics.currencyCode or @Semantics.unitOfMeasure.

@Aggregation.CustomAggregate#amount   : 'Edm.Decimal'
@Aggregation.CustomAggregate#currency : 'Edm.String'
entity Sales {
    key id        : GUID;
        productId : GUID;
        @Semantics.amount.currencyCode: 'currency'
        amount    : Decimal(10,2);
        @Semantics.currencyCode
        currency  : String(3);
}

The CAP Java SDK will expose all properties annotated with the @Semantics.currencyCode or @Semantics.unitOfMeasure as a custom aggregate with the property’s name that returns:

• The property’s value if it’s unique within a group of dimensions
• null otherwise

A custom aggregate for currency code or unit of measure should be also exposed by the @Aggregation.CustomAggregate annotation. Moreover, a property for a monetary amount or a measured quantity should be annotated with @Semantics.amount.currencyCode or @Semantics.quantity.unitOfMeasure to reference the corresponding property that holds the amount’s currency code or the quantity’s unit of measure, rsp..

Other Features

OData V2 Support

While CAP defaults to OData V4, the latest protocol version, some projects need to fallback to OData V2, for example, to keep using existing V2-based UIs.

Enabling OData V2 via Proxy in Node.js Apps

CAP Node.js supports serving the OData V2 protocol through the OData V2 proxy protocol adapter, which translates between the OData V2 and V4 protocols.

If Node.js projects, add the proxy as express.js middleware as follows:

1. Add the proxy package to your project:

    npm add @sap/cds-odata-v2-adapter-proxy
   

2. Add this to a project-local ./srv/server.js:

    const proxy = require('@sap/cds-odata-v2-adapter-proxy')
    const cds = require('@sap/cds')
    cds.on('bootstrap', app => app.use(proxy()))
    module.exports = cds.server
   

3. Access OData V2 services at http://localhost:4004/v2/${path}.
4. Access OData V4 services at http://localhost:4004/${path} (as before).

Example: Read service metadata for CatalogService:

• CDS:

    @path:'/browse'
    service CatalogService { ... }
  

• OData V2: GET http://localhost:4004/v2/browse/$metadata
• OData V4: GET http://localhost:4004/browse/$metadata

Find more detailed instructions at @sap/cds-odata-v2-adapter-proxy.

Using OData V2 in Java Apps

In CAP Java, serving the OData V2 protocol is natively supported by the CDS OData V2 Adapter.

Miscellaneous

Omitting Elements from APIs

Add annotation @cds.api.ignore to suppress unwanted entity fields (for example, foreign-key fields) in APIs exposed from this the CDS model, that is, OData or Open API. For example:

entity Books { ...
  @cds.api.ignore 
  author : Association to Authors;
}

Absolute Context URL

In some scenarios, an absolute context URL is needed. In the Node.js runtime, this can be achieved through configuration cds.odata.contextAbsoluteUrl.

You can use your own URL (including a protocol and a service path), for example:

cds.odata.contextAbsoluteUrl = "https://your.domain.com/yourService"

to customize the annotation as follows:

{
  "@odata.context":"https://your.domain.com/yourService/$metadata#Books(title,author,ID)",
  "value":[
    {"ID": 201,"title": "Wuthering Heights","author": "Emily Brontë"},
    {"ID": 207,"title": "Jane Eyre","author": "Charlotte Brontë"},
    {"ID": 251,"title": "The Raven","author": "Edgar Allen Poe"}
  ]
}

If contextAbsoluteUrl is set to something truthy that doesn’t match http(s)://*, an absolute path is constructed based on the environment of the application on a best effort basis.

Note that we encourage you to stay by the default relative format, if possible, as it’s proxy safe.