Consuming Services

This guide is available for Node.js and Java.

Use the toggle in the title bar or press v to switch.

• Introduction
  • Feature Overview
  • Tutorials and Examples
  • Define Scenario
• Get and Import an External Service API
  • From SAP Business Accelerator Hub
  • For a Remote CAP Service
  • Import API Definition
• Local Mocking
  • Add Mock Data
  • Run Local with Mocks
  • Mock Associations
  • Mock Remote Service as OData Service (Node.js)
  • Mock Remote Service as OData Service (Java)
• Execute Queries
  • Execute Queries with Node.js
  • Execute Queries with Java
  • Model Projections
  • Execute Queries on Projections to a Remote Service
  • Building Custom Requests with Node.js
  • Building Custom Requests with Java
• Integrate and Extend
  • Expose Remote Services
  • Expose Remote Services with Associations
  • Mashing up with Remote Services
  • Handle Mashups with Remote Services
  • Limitations and Feature Matrix
• Connect and Deploy
  • Using Destinations
  • Connect to Remote Services Locally
  • Connect to an Application Using the Same XSUAA (Forward Authorization Token)
  • Connect to an Application in Your Kyma Cluster
  • Deployment
  • Destinations and Multitenancy
• Add Qualities
  • Resilience
  • Tracing
• Feature Details
  • Legend
  • Supported Protocols
  • Querying API Features
  • Supported Projection Features
  • Supported Features for Application Defined Destinations

Introduction

If you want to use data from other services or you want to split your application into multiple microservices, you need a connection between those services. We call them remote services. As everything in CAP is a service, remote services are modeled the same way as internal services — using CDS.

CAP supports service consumption with dedicated APIs to import service definitions, query remote services, mash up services, and work locally as much as possible.

Feature Overview

For outbound remote service consumption, the following features are supported:

• OData V4
• OData V2 (Deprecated)
• Querying API
• Projections on remote services

Tutorials and Examples

Example	Description
Capire Bookshop (Fiori)	Example, Node.js, CAP-to-CAP
Example Application (Node.js)	Complete application from the end-to-end Tutorial
Example Application (Java)	Complete application from the end-to-end Tutorial

Define Scenario

Before you start your implementation, you should define your scenario. Answering the following questions gets you started:

• What services (remote/CAP) are involved?
• How do they interact?
• What needs to be displayed on the UI?

You have all your answers and know your scenario, go on reading about external service APIs, getting an API definition from the SAP Business Accelerator Hub or from a CAP project, and importing an API definition to your project.

Sample Scenario from End-to-End Tutorial

User Story

A company wants to ensure that goods are only sourced from suppliers with acceptable risks. There shall be a software system, that allows a clerk to maintain risks for suppliers and their mitigations. The system shall block the supplier used if risks can't be mitigated.

The application is an extension for SAP S/4HANA. It deals with risks and mitigations that are local entities in the application and suppliers that are stored in SAP S/4HANA Cloud. The application helps to reduce risks associated with suppliers by automatically blocking suppliers with a high risk using a remote API Call.

Integrate

The user picks a supplier from the list. That list is coming from the remote system and is exposed by the CAP application. Then the user does a risk assessment. Additional supplier data, like name and blocked status, should be displayed on the UI as well, by integrating the remote supplier service into the local risk service.

Extend

It should be also possible to search for suppliers and show the associated risks by extending the remote supplier service with the local risk service and its risks.

Get and Import an External Service API

To communicate to remote services, CAP needs to know their definitions. Having the definitions in your project allows you to mock them during design time.

These definitions are usually made available by the service provider. As they aren't defined within your application but imported from outside, they're called external service APIs in CAP. Service APIs can be provided in different formats. Currently, EDMX files for OData V2 and V4 are supported.

From SAP Business Accelerator Hub

The SAP Business Accelerator Hub provides many relevant APIs from SAP. You can download API specifications in different formats. If available, use the EDMX format. The EDMX format describes OData interfaces.

To download the Business Partner API (A2X) from SAP S/4HANA Cloud, go to section API Resources, select API Specification, and download the EDMX file.

Get more details in the end-to-end tutorial.

For a Remote CAP Service

We recommend using EDMX as exchange format. Export a service API to EDMX:

Mac/Linux

cds compile srv -s OrdersService -2 edmx > OrdersService.edmx

Windows

cds compile srv -s OrdersService -2 edmx > OrdersService.edmx

Powershell

cds compile srv -s OrdersService -2 edmx -o dest/

You can try it with the orders sample in cap/samples.

By default, CAP works with OData V4 and the EDMX export is in this protocol version as well. The cds compile command offers options for other OData versions and flavors, call cds help compile for more information.

Don't just copy the CDS file for a remote CAP service

Simply copying CDS files from a different application comes with the following issues:

• The effective service API depends on the used protocol.
• CDS files often use includes, which can't be resolved anymore.
• CAP creates unneeded database tables and views for all entities in the file.

Import API Definition

Import the API to your project using cds import.

cds import <input_file> --as cds

<input_file> can be an EDMX (OData V2, OData V4), OpenAPI or AsyncAPI file.

Option	Description
--as cds	The import creates a CDS file (for example API_BUSINESS_PARTNER.cds) instead of a CSN file.

This adds the API in CDS format to the srv/external folder and also copies the input file into that folder.

Further, it adds the API as an external service to your package.json. You use this declaration later to connect to the remote service using a destination.

"cds": {
    "requires": {
        "API_BUSINESS_PARTNER": {
            "kind": "odata-v2",
            "model": "srv/external/API_BUSINESS_PARTNER"
        }
    }
}

Options and flags in .cdsrc.json

Alternatively, you can set the options and flags for cds import in your .cdsrc.json:

{
    "import": {
        "as": "cds",
        "force": true,
        "include_namespaces": "sap,c4c"
    }
}

Now run cds import <filename>

• --as only supports these formats: "csn","cds", and "json"
• --force is applicable only in combination with --as option. By default the --force flag is set to false.

  If set to true, existing CSN/CDS files from previous imports are overwritten.

When importing the specification files, the kind is set according to the following mapping:

Imported Format	Used kind
OData V2	odata-v2
OData V4	odata (alias for odata-v4)
OpenAPI	rest
AsyncAPI	odata

Learn more about type mappings from OData to CDS and vice versa.

TIP

Always use OData V4 (odata) when calling another CAP service.

Limitations

Not all features of OData, OpenAPI, or AsyncAPI are supported in CAP which may lead to the rejection of the imported model by the CDS compiler or may result in a different API when rendered by CAP. Known limitations are cyclic type references and inheritance.

You need to configure remote services in Spring Boot's application.yaml:

srv/src/main/resources/application.yaml

spring:
  config.activate.on-profile: cloud
cds:
  remote.services:
    API_BUSINESS_PARTNER:
      type: "odata-v2"

To work with remote services, add the following dependency to your Maven project:

<dependency>
  <groupId>com.sap.cds</groupId>
  <artifactId>cds-feature-remote-odata</artifactId>
  <scope>runtime</scope>
</dependency>

Learn about all cds.remote.services configuration possibilities.

Local Mocking

When developing your application, you can mock the remote service.

Add Mock Data

As for any other CAP service, you can add mocking data.

The CSV file needs to be added to the srv/external/data folder.

The CSV file needs to be added to the db/data folder.

API_BUSINESS_PARTNER-A_BusinessPartner.csv

BusinessPartner;BusinessPartnerFullName;BusinessPartnerIsBlocked
1004155;Williams Electric Drives;false
1004161;Smith Batteries Ltd;false
1004100;Johnson Automotive Supplies;true

For Java, make sure to add the --with-mocks option to the cds deploy command used to generate the schema.sql in srv/pom.xml. This ensures that tables for the mocked remote entities are created in the database.

Find this source in the end-to-end tutorial

Get more details in the end-to-end tutorial.

Run Local with Mocks

Start your project with the imported service definition.

cds watch

The service is automatically mocked, as you can see in the log output on server start.

...

[cds] - model loaded from 8 file(s):

  ...
  ./srv/external/API_BUSINESS_PARTNER.cds
  ...

[cds] - connect using bindings from: { registry: '~/.cds-services.json' }
[cds] - connect to db > sqlite { database: ':memory:' }
 > filling sap.ui.riskmanagement.Mitigations from ./db/data/sap.ui.riskmanagement-Mitigations.csv
 > filling sap.ui.riskmanagement.Risks from ./db/data/sap.ui.riskmanagement-Risks.csv
 > filling API_BUSINESS_PARTNER.A_BusinessPartner from ./srv/external/data/API_BUSINESS_PARTNER-A_BusinessPartner.csv
/> successfully deployed to sqlite in-memory db

[cds] - serving RiskService { at: '/service/risk', impl: './srv/risk-service.js' }
[cds] - mocking API_BUSINESS_PARTNER { at: '/api-business-partner' }

[cds] - launched in: 1.104s
[cds] - server listening on { url: 'http://localhost:4004' }
[ terminate with ^C ]

mvn spring-boot:run

Mock Associations

You can't get data from associations of a mocked service out of the box.

The associations of imported services lack information how to look up the associated records. This missing relation is expressed with an empty key definition at the end of the association declaration in the CDS model ({ }).

srv/external/API_BUSINESS_PARTNER.cds

entity API_BUSINESS_PARTNER.A_BusinessPartner {
  key BusinessPartner : LargeString;
  BusinessPartnerFullName : LargeString;
  BusinessPartnerType : LargeString;

  ...

  to_BusinessPartnerAddress :
    Association to many API_BUSINESS_PARTNER.A_BusinessPartnerAddress {  };  
};

entity API_BUSINESS_PARTNER.A_BusinessPartnerAddress {
  key BusinessPartner : String(10);
  key AddressID : String(10);

  ...
};

To mock an association, you have to modify the imported file. Before doing any modifications, create a local copy and add it to your source code management system.

cp srv/external/API_BUSINESS_PARTNER.cds srv/external/API_BUSINESS_PARTNER-orig.cds
git add srv/external/API_BUSINESS_PARTNER-orig.cds
...

Import the CDS file again, just using a different name:

cds import ~/Downloads/API_BUSINESS_PARTNER.edmx --keep-namespace \
    --as cds --out srv/external/API_BUSINESS_PARTNER-new.cds

Add an on condition to express the relation:

srv/external/API_BUSINESS_PARTNER-new.cds

entity API_BUSINESS_PARTNER.A_BusinessPartner {
  // ...
  to_BusinessPartnerAddress :
      Association to many API_BUSINESS_PARTNER.A_BusinessPartnerAddress
      on to_BusinessPartnerAddress.BusinessPartner = BusinessPartner;
};

Don't add any keys or remove empty keys, which would change it to a managed association. Added fields aren't known in the service and lead to runtime errors.

Use a 3-way merge tool to take over your modifications, check it and overwrite the previous unmodified file with the newly imported file:

git merge-file API_BUSINESS_PARTNER.cds \
               API_BUSINESS_PARTNER-orig.cds \
               API_BUSINESS_PARTNER-new.cds
mv API_BUSINESS_PARTNER-new.cds API_BUSINESS_PARTNER-orig.cds

To prevent accidental loss of modifications, the cds import --as cds command refuses to overwrite modified files based on a "checksum" that is included in the file.

Mock Remote Service as OData Service (Node.js)

As shown previously you can run one process including a mocked external service. However, this mock doesn't behave like a real external service. The communication happens in-process and doesn't use HTTP or OData. For a more realistic testing, let the mocked service run in a separate process.

First install the required packages:

npm add @sap-cloud-sdk/http-client@3.x @sap-cloud-sdk/connectivity@3.x @sap-cloud-sdk/resilience@3.x

Then start the CAP application with the mocked remote service only:

cds mock API_BUSINESS_PARTNER

If the startup is completed, run cds watch in the same project from a different terminal:

cds watch

CAP tracks locally running services. The mocked service API_BUSINESS_PARTNER is registered in file ~/.cds-services.json. cds watch searches for running services in that file and connects to them.

Node.js only supports OData V4 protocol and so does the mocked service. There might still be some differences to the real remote service if it uses a different protocol, but it's much closer to it than using only one instance. In the console output, you can also easily see how the communication between the two processes happens.

Mock Remote Service as OData Service (Java)

You configure CAP to do OData and HTTP requests for a mocked service instead of doing it in-process. Configure a new Spring Boot profile (for example mocked):

srv/src/main/resources/application.yaml

spring:
  config.activate.on-profile: mocked
cds:
  application.services:
  - name: API_BUSINESS_PARTNER-mocked
    model: API_BUSINESS_PARTNER
    serve.path: API_BUSINESS_PARTNER
  remote.services:
    API_BUSINESS_PARTNER:
      destination:
        name: "s4-business-partner-api-mocked"

The profile exposes the mocked service as OData service and defines a destination to access the service. The destination just points to the CAP application itself. You need to implement some Java code for this:

DestinationConfiguration.java

@EventListener
void applicationReady(ApplicationReadyEvent ready) {
  int port = Integer.valueOf(environment.getProperty("local.server.port"));
  DefaultHttpDestination mockDestination = DefaultHttpDestination
      .builder("http://localhost:" + port)
      .name("s4-business-partner-api-mocked").build();

  DefaultDestinationLoader loader = new DefaultDestinationLoader();
  loader.registerDestination(mockDestination);
  DestinationAccessor.prependDestinationLoader(loader);
}

Now, you just need to run the application with the new profile:

mvn spring-boot:run -Dspring-boot.run.profiles=default,mocked

When sending a request to your CAP application, for example the Suppliers entity, it is transformed to the request for the mocked remote service and requested from itself as a OData request. Therefore, you'll see two HTTP requests in your CAP application's log.

For example:

http://localhost:8080/service/risk/Suppliers

2021-09-21 15:18:44.870 DEBUG 34645 — [nio-8080-exec-1] o.s.web.servlet.DispatcherServlet        : GET "/service/risk/Suppliers", parameters={}
...
2021-09-21 15:18:45.292 DEBUG 34645 — [nio-8080-exec-2] o.s.web.servlet.DispatcherServlet        : GET "/API_BUSINESS_PARTNER/A_BusinessPartner?$select=BusinessPartner,BusinessPartnerFullName,BusinessPartnerIsBlocked&$top=1000&$skip=0&$orderby=BusinessPartner%20asc&sap-language=de&sap-valid-at=2021-09-21T13:18:45.211722Z", parameters={masked}
...
2021-09-21 15:18:45.474 DEBUG 34645 — [nio-8080-exec-2] o.s.web.servlet.DispatcherServlet        : Completed 200 OK
2021-09-21 15:18:45.519 DEBUG 34645 — [nio-8080-exec-1] o.s.web.servlet.DispatcherServlet        : Completed 200 OK

Try out the example application.

Execute Queries

You can send requests to remote services using CAP's powerful querying API.

Execute Queries with Node.js

Connect to the service before sending a request, as usual in CAP:

const bupa = await cds.connect.to('API_BUSINESS_PARTNER');

Then execute your queries using the Querying API:

const { A_BusinessPartner } = bupa.entities;
const result = await bupa.run(SELECT(A_BusinessPartner).limit(100));

We recommend limiting the result set and avoid the download of large data sets in a single request. You can limit the result as in the example: .limit(100).

Many features of the querying API are supported for OData services. For example, you can resolve associations like this:

const { A_BusinessPartner } = bupa.entities;
const result = await bupa.run(SELECT.from(A_BusinessPartner, bp => {
    bp('BusinessPartner'),
    bp.to_BusinessPartnerAddress(addresses => {
      addresses('*')
    })
  }).limit(100));

Learn more about querying API examples.

Learn more about supported querying API features.

Execute Queries with Java

You can use dependency injection to get access to the remote service:

@Autowired
@Qualifier(ApiBusinessPartner_.CDS_NAME)
CqnService bupa;

Then execute your queries using the Querying API:

CqnSelect select = Select.from(ABusinessPartner_.class).limit(100);
List<ABusinessPartner> businessPartner = bupa.run(select).listOf(ABusinessPartner.class);

Learn more about querying API examples.

Learn more about supported querying API features.

Model Projections

External service definitions, like generated CDS or CSN files during import, can be used as any other CDS definition, but they don't generate database tables and views unless they are mocked.

It's best practice to use your own "interface" to the external service and define the relevant fields in a projection in your namespace. Your implementation is then independent of the remote service implementation and you request only the information that you require.

using {  API_BUSINESS_PARTNER as bupa } from '../srv/external/API_BUSINESS_PARTNER';

entity Suppliers as projection on bupa.A_BusinessPartner {
  key BusinessPartner as ID,
  BusinessPartnerFullName as fullName,
  BusinessPartnerIsBlocked as isBlocked,
}

As the example shows, you can use field aliases as well.

Learn more about supported features for projections.

Execute Queries on Projections to a Remote Service

Connect to the service before sending a request, as usual in CAP:

const bupa = await cds.connect.to('API_BUSINESS_PARTNER');

Then execute your queries:

const suppliers = await bupa.run(SELECT(Suppliers).where({ID}));

CAP resolves projections and does the required mapping, similar to databases.

A brief explanation, based on the previous query, what CAP does:

• Resolves the Suppliers projection to the external service interface API_BUSINESS_PARTNER.A_Business_Partner.
• The where condition for field ID will be mapped to the BusinessPartner field of A_BusinessPartner.
• The result is mapped back to the Suppliers projection, so that values for the BusinessPartner field are mapped back to ID.

This makes it convenient to work with external services.

Building Custom Requests with Node.js

If you can't use the querying API, you can craft your own HTTP requests using send:

bupa.send({
  method: 'PATCH',
  path: A_BusinessPartner,
  data: {
    BusinessPartner: 1004155,
    BusinessPartnerIsBlocked: true
  }
})

Learn more about the send API.

Building Custom Requests with Java

For Java, you can use the HttpClient API to implement your custom requests. The API is enhanced by the SAP Cloud SDK to support destinations.

Learn more about using the HttpClient Accessor.

Learn more about using destinations.

Integrate and Extend

By creating projections on remote service entities and using associations, you can create services that combine data from your local service and remote services.

What you need to do depends on the scenarios and how your remote services should be integrated into, as well as extended by your local services.

Expose Remote Services

To expose a remote service entity, you add a projection on it to your CAP service:

using {  API_BUSINESS_PARTNER as bupa } from '../srv/external/API_BUSINESS_PARTNER';

extend service RiskService with {
  entity BusinessPartners as projection on bupa.A_BusinessPartner;
}

CAP automatically tries to delegate queries to database entities, which don't exist as you're pointing to an external service. That behavior would produce an error like this:

<error xmlns="https://docs.oasis-open.org/odata/ns/metadata">
<code>500</code>
<message>SQLITE_ERROR: no such table: RiskService_BusinessPartners in: SELECT BusinessPartner, Customer, Supplier, AcademicTitle, AuthorizationGroup, BusinessPartnerCategory, BusinessPartnerFullName, BusinessPartnerGrouping, BusinessPartnerName, BusinessPartnerUUID, CorrespondenceLanguage, CreatedByUser, CreationDate, (...)  FROM RiskService_BusinessPartner ALIAS_1 ORDER BY BusinessPartner COLLATE NOCASE ASC LIMIT 11</message>
</error>

To avoid this error, you need to handle projections. Write a handler function to delegate a query to the remote service and run the incoming query on the external service.

Node.js

module.exports = cds.service.impl(async function() {
  const bupa = await cds.connect.to('API_BUSINESS_PARTNER');

  this.on('READ', 'BusinessPartners', req => {
      return bupa.run(req.query);
  });
});

Java

@Component
@ServiceName(RiskService_.CDS_NAME)
public class RiskServiceHandler implements EventHandler {
  @Autowired
  @Qualifier(ApiBusinessPartner_.CDS_NAME)
  CqnService bupa;

  @On(entity = BusinessPartners.CDS_NAME)
  Result readSuppliers(CdsReadEventContext context) {
    return bupa.run(context.getCqn());
  }
}

For Node.js, get more details in the end-to-end tutorial.

WARNING

If you receive 404 errors, check if the request contains fields that don't exist in the service and start with the name of an association. cds import adds an empty keys declaration ({ }) to each association. Without this declaration, foreign keys for associations are generated in the runtime model, that don't exist in the real service. To solve this problem, you need to reimport the external service definition using cds import.

This works when accessing the entity directly. Additional work is required to support navigation and expands from or to a remote entity.

Instead of exposing the remote service's entity unchanged, you can model your own projection. For example, you can define a subset of fields and change their names.

TIP

CAP does the magic that maps the incoming query, according to your projections, to the remote service and maps back the result.

using { API_BUSINESS_PARTNER as bupa } from '../srv/external/API_BUSINESS_PARTNER';

extend service RiskService with {
  entity Suppliers as projection on bupa.A_BusinessPartner {
    key BusinessPartner as ID,
    BusinessPartnerFullName as fullName,
    BusinessPartnerIsBlocked as isBlocked
  }
}

module.exports = cds.service.impl(async function() {
  const bupa = await cds.connect.to('API_BUSINESS_PARTNER');

  this.on('READ', 'Suppliers', req => {
      return bupa.run(req.query);
  });
});

Learn more about queries on projections to remote services.

Expose Remote Services with Associations

It's possible to expose associations of a remote service entity. You can adjust the projection for the association target and change the name of the association:

using { API_BUSINESS_PARTNER as bupa } from '../srv/external/API_BUSINESS_PARTNER';

extend service RiskService with {
  entity Suppliers as projection on bupa.A_BusinessPartner {
    key BusinessPartner as ID,
    BusinessPartnerFullName as fullName,
    BusinessPartnerIsBlocked as isBlocked,
    to_BusinessPartnerAddress as addresses: redirected to SupplierAddresses
  }

  entity SupplierAddresses as projection on bupa.A_BusinessPartnerAddress {
    BusinessPartner as bupaID,
    AddressID as ID,
    CityName as city,
    StreetName as street,
    County as county
  }
}

As long as the association is only resolved using expands (for example .../risk/Suppliers?$expand=addresses), a handler for the source entity is sufficient:

this.on('READ', 'Suppliers', req => {
    return bupa.run(req.query);
});

If you need to resolve the association using navigation or request it independently from the source entity, add a handler for the target entity as well:

this.on('READ', 'SupplierAddresses', req => {
    return bupa.run(req.query);
});

As usual, you can put two handlers into one handler matching both entities:

this.on('READ', ['Suppliers', 'SupplierAddresses'], req => {
    return bupa.run(req.query);
});

Mashing up with Remote Services

You can combine local and remote services using associations. These associations need manual handling, because of their different data sources.

Integrate Remote into Local Services

Use managed associations from local entities to remote entities:

@path: 'service/risk'
service RiskService {
  entity Risks : managed {
    key ID      : UUID  @(Core.Computed : true);
    title       : String(100);
    prio        : String(5);
    supplier    : Association to Suppliers;
  }

  entity Suppliers as projection on BusinessPartner.A_BusinessPartner {
    key BusinessPartner as ID,
    BusinessPartnerFullName as fullName,
    BusinessPartnerIsBlocked as isBlocked,
  };
}

Extend a Remote by a Local Service

You can augment a projection with a new association, if the required fields for the on condition are present in the remote service. The use of managed associations isn't possible, because this requires to create new fields in the remote service.

entity Suppliers as projection on bupa.A_BusinessPartner {
  key BusinessPartner as ID,
  BusinessPartnerFullName as fullName,
  BusinessPartnerIsBlocked as isBlocked,
  risks : Association to many Risks on risks.supplier.ID = ID,
};

Handle Mashups with Remote Services

Depending on how the service is accessed, you need to support direct requests, navigation, or expands. CAP resolves those three request types only for service entities that are served from the database. When crossing the boundary between database and remote sourced entities, you need to take care of those requests.

The list of required implementations for mashups explains the different combinations.

Handle Expands Across Local and Remote Entities

Expands add data from associated entities to the response. For example, for a risk, you want to display the suppliers name instead of just the technical ID. But this property is part of the (remote) supplier and not part of the (local) risk.

Get more details in the end-to-end tutorial.

To handle expands, you need to add a handler for the main entity:

1. Check if a relevant $expand column is present.
2. Remove the $expand column from the request.
3. Get the data for the request.
4. Execute a new request for the expand.
5. Add the expand data to the returned data from the request.

Example of a CQN request with an expand:

{
  "from": { "ref": [ "RiskService.Suppliers" ] },
  "columns": [
    { "ref": [ "ID" ] },
    { "ref": [ "fullName" ] },
    { "ref": [ "isBlocked" ] },
    { "ref": [ "risks" ] },
    { "expand": [
      { "ref": [ "ID" ] },
      { "ref": [ "title" ] },
      { "ref": [ "descr" ] },
      { "ref": [ "supplier_ID" ] }
    ] }
  ]
}

See an example how to handle expands in Node.js.

See an example how to handle expands in Java.

Expands across local and remote can cause stability and performance issues. For a list of items, you need to collect all IDs and send it to the database or the remote system. This can become long and may exceed the limits of a URL string in case of OData. Do you really need expands for a list of items?

GET /service/risk/Risks?$expand=supplier

Or is it sufficient for single items?

GET /service/risk/Risks(545A3CF9-84CF-46C8-93DC-E29F0F2BC6BE)/?$expand=supplier

Keep performance in mind

Consider to reject expands if it's requested on a list of items.

Handle Navigations Across Local and Remote Entities

Navigations allow to address items via an association from a different entity:

GET /service/risks/Risks(20466922-7d57-4e76-b14c-e53fd97dcb11)/supplier

The CQN consists of a from condition with 2 values for ref. The first ref selects the record of the source entity of the navigation. The second ref selects the name of the association, to navigate to the target entity.

{
  "from": {
    "ref": [ {
      "id": "RiskService.Risks",
      "where": [
        { "ref": [ "ID" ] },
        "=",
        { "val": "20466922-7d57-4e76-b14c-e53fd97dcb11" }
      ]},
      "supplier"
    ]
  },
  "columns": [
    { "ref": [ "ID" ] },
    { "ref": [ "fullName" ] },
    { "ref": [ "isBlocked" ] }
  ],
  "one": true
}

To handle navigations, you need to check in your code if the from.ref object contains 2 elements. Be aware, that for navigations the handler of the target entity is called.

If the association's on condition equals the key of the source entity, you can directly select the target entity using the key's value. You find the value in the where block of the first from.ref entry.

Otherwise, you need to select the source item using that where block and take the required fields for the associations on condition from that result.

See an example how to handle navigations in Node.js.

See an example how to handle navigations in Java.

Limitations and Feature Matrix

Required Implementations for Mashups

You need additional logic, if remote entities are in the game. The following table shows what is required. "Local" is a database entity or a projection on a database entity.

Request	Example	Implementation
Local (including navigations and expands)	/service/risks/Risks	Handled by CAP
Local: Expand remote	/service/risks/Risks?$expand=supplier	Delegate query w/o expand to local service and implement expand.
Local: Navigate to remote	/service/risks(...)/supplier	Implement navigation and delegate query target to remote service.
Remote (including navigations and expands to the same remote service)	/service/risks/Suppliers	Delegate query to remote service
Remote: Expand local	/service/risks/Suppliers?$expand=risks	Delegate query w/o expand to remote service and implement expand.
Remote: Navigate to local	/service/Suppliers(...)/risks	Implement navigation, delegate query for target to local service

Transient Access vs. Replication

This chapter shows only techniques for transient access.

The following matrix can help you to find the best approach for your scenario:

Feature	Transient Access	Replication
Filtering on local or remote fields 1	Possible	Possible
Filtering on local and remote fields 2	Not possible	Possible
Relationship: Uni-/Bidirectional associations	Possible	Possible
Relationship: Flatten	Not possible	Possible
Evaluate user permissions in remote system	Possible	Requires workarounds 3
Data freshness	Live data	Outdated until replicated
Performance	Degraded 4	Best

¹ It's not required to filter both, on local and remote fields, in the same request.
² It's required to filter both, on local and remote fields, in the same request.
³ Because replicated data is accessed, the user permission checks of the remote system aren't evaluated.
⁴ Depends on the connectivity and performance of the remote system.

Connect and Deploy

Using Destinations

Destinations contain the necessary information to connect to a remote system. They're basically an advanced URL, that can carry additional metadata like, for example, the authentication information.

You can choose to use SAP BTP destinations or application defined destinations.

Use SAP BTP Destinations

CAP leverages the destination capabilities of the SAP Cloud SDK.

Create Destinations on SAP BTP

Create a destination using one or more of the following options.

• Register a system in your global account: You can check here how to Register an SAP System in your SAP BTP global account and which systems are supported for registration. Once the system is registered and assigned to your subaccount, you can create a service instance. A destination is automatically created along with the service instance.

• Connect to an on-premise system: With SAP BTP Cloud Connector, you can create a connection from your cloud application to an on-premise system.

• Manually create destinations: You can create destinations manually in your SAP BTP subaccount. See section destinations in the SAP BTP documentation.

• Create a destination to your application: If you need a destination to your application, for example, to call it from a different application, then you can automatically create it in the MTA deployment.

Use Destinations with Node.js

In your package.json, a configuration for the API_BUSINESS_PARTNER looks like this:

"cds": {
  "requires": {
    "API_BUSINESS_PARTNER": {
      "kind": "odata",
      "model": "srv/external/API_BUSINESS_PARTNER"
    }
  }
}

If you've imported the external service definition using cds import, an entry for the service in the package.json has been created already. Here you specify the name of the destination in the credentials block.

In many cases, you also need to specify the path prefix to the service, which is added to the destination's URL. For services listed on the SAP Business Accelerator Hub, you can find the path in the linked service documentation.

Since you don't want to use the destination for local testing, but only for production, you can profile it by wrapping it into a [production] block:

"cds": {
  "requires": {
    "API_BUSINESS_PARTNER": {
      "kind": "odata",
      "model": "srv/external/API_BUSINESS_PARTNER",
      "[production]": {
        "credentials": {
          "destination": "S4HANA",
          "path": "/sap/opu/odata/sap/API_BUSINESS_PARTNER"
        }
      }
    }
  }
}

Additionally, you can provide destination options inside a destinationOptions object:

"cds": {
  "requires": {
    "API_BUSINESS_PARTNER": {
      /* ... */
      "[production]": {
        "credentials": {
          /* ... */
        },
        "destinationOptions": {
          "selectionStrategy": "alwaysSubscriber",
          "useCache": true
        }
      }
    }
  }
}

The selectionStrategy property controls how a destination is resolved.

The useCache option controls whether the SAP Cloud SDK caches the destination. It's enabled by default but can be disabled by explicitly setting it to false. Read Destination Cache to learn more about how the cache works.

If you want to configure additional headers for the HTTP request to the system behind the destination, for example an Application Interface Register (AIR) header, you can specify such headers in the destination definition itself using the property URL.headers.<header-key>.

Use Destinations with Java

Destinations are configured in Spring Boot's application.yaml file:

srv/src/main/resources/application.yaml

cds:
  remote.services:
    API_BUSINESS_PARTNER:
      type: "odata-v2"
      destination:
        name: "cpapp-bupa"
      http:
        suffix: "/sap/opu/odata/sap"

Learn more about configuring destinations for Java.

Use Application Defined Destinations

If you don't want to use SAP BTP destinations, you can also define destinations, which means the URL, authentication type, and additional configuration properties, in your application configuration or code.

Application defined destinations support a subset of properties and authentication types of the SAP BTP destination service.

Configure Application Defined Destinations in Node.js

You specify the destination properties in credentials instead of putting the name of a destination there.

This is an example of a destination using basic authentication:

"cds": {
  "requires": {
    "REVIEWS": {
      "kind": "odata",
      "model": "srv/external/REVIEWS",
      "[production]": {
        "credentials": {
          "url": "https://reviews.ondemand.com/reviews",
          "authentication": "BasicAuthentication",
          "username": "<set from code or env>",
          "password": "<set from code or env>",
          "headers": {
            "my-header": "header value"
          },
          "queries": {
            "my-url-param": "url param value"
          }
        }
      }
    }
  }
}

Supported destination properties.

WARNING

You shouldn't put any sensitive information here.

Instead, set the properties in the bootstrap code of your CAP application:

const cds = require("@sap/cds");

if (cds.env.requires?.credentials?.authentication === "BasicAuthentication") {
  const credentials = /* read your credentials */
  cds.env.requires.credentials.username = credentials.username;
  cds.env.requires.credentials.password = credentials.password;
}

You might also want to set some values in the application deployment. This can be done using env variables. For this example, the env variable for the URL would be cds_requires_REVIEWS_credentials_destination_url.

This variable can be parameterized in the manifest.yml for a cf push based deployment:

manifest.yml

applications:
- name: reviews
  ...
  env:
    cds_requires_REVIEWS_credentials_url: ((reviews_url))

cf push --var reviews_url=https://reviews.ondemand.com/reviews

The same can be done using mtaext file for MTA deployment.

If the URL of the target service is also part of the MTA deployment, you can automatically receive it as shown in this example:

mta.yaml

 - name: reviews
   provides:
    - name: reviews-api
      properties:
        reviews-url: ${default-url}
 - name: bookshop
   requires:
    ...
    - name: reviews-api
   properties:
     cds_requires_REVIEWS_credentials_url: ~{reviews-api/reviews-url}

.env

cds_requires_REVIEWS_credentials_url=http://localhost:4008/reviews

WARNING

For the configuration path, you must use the underscore ("_") character as delimiter. CAP supports dot (".") as well, but Cloud Foundry won't recognize variables using dots. Your service name mustn't contain underscores.

Implement Application Defined Destinations in Node.js

There is no API to create a destination in Node.js programmatically. However, you can change the properties of a remote service before connecting to it, as shown in the previous example.

Configure Application Defined Destinations in Java

Destinations are configured in Spring Boot's application.yaml file.

srv/src/main/resources/application.yaml

cds:
  remote.services:
    REVIEWS:
      type: "odata-v4"
      destination:
        properties:
          url: https://reviews.ondemand.com/reviews
          authentication: TokenForwarding
      http:
        headers:
          my-header: "header value"
        queries:
          my-url-param: "url param value"

Learn more about supported destination properties.

Implement Application Defined Destinations in Java

You can use the APIs offered by SAP Cloud SDK to create destinations programmatically. The destination can be used by its name the same way as destinations on the SAP BTP destination service.

srv/src/main/resources/application.yaml

cds:
  remote.services:
    REVIEWS:
      type: "odata-v2"
      destination:
        name: "reviews-destination"

Learn more about programmatic destination registration. See examples for different authentication types.

Connect to Remote Services Locally

If you use SAP BTP destinations, you can access them locally using CAP's hybrid testing capabilities with the following procedure:

Bind to Remote Destinations

Your local application needs access to an XSUAA and Destination service instance in the same subaccount where the destination is:

1. Login to your Cloud Foundry org and space

2. Create an XSUAA service instance and service key:

   cf create-service xsuaa application cpapp-xsuaa
   cf create-service-key cpapp-xsuaa cpapp-xsuaa-key

3. Create a Destination service instance and service key:

   cf create-service destination lite cpapp-destination
   cf create-service-key cpapp-destination cpapp-destination-key

4. Bind to XSUAA and Destination service:

   cds bind -2 cpapp-xsuaa,cpapp-destination

   Learn more about cds bind.

Run a Node.js Application with a Destination

Add the destination for the remote service to the hybrid profile in the .cdsrc-private.json file:

{
  "requires": {
    "[hybrid]": {
      "auth": {
        /* ... */
      },
      "destinations": {
        /* ... */
      },
      "API_BUSINESS_PARTNER": {
        "credentials": {
          "path": "/sap/opu/odata/sap/API_BUSINESS_PARTNER",
          "destination": "cpapp-bupa"
        }
      }
    }
  }
}

Run your application with the Destination service:

cds watch --profile hybrid

TIP

If you are developing in the Business Application Studio and want to connect to an on-premise system, you will need to do so via Business Application Studio's built-in proxy, for which you need to add configuration in an .env file. See Connecting to External Systems From the Business Application Studio for more details.

Run a Java Application with a Destination

Add a new profile hybrid to your application.yaml file that configures the destination for the remote service.

srv/src/main/resources/application.yaml

spring:
  config.activate.on-profile: hybrid
  sql.init.schema-locations:
  - "classpath:schema-nomocks.sql"
cds:
  remote.services:
  - name: API_BUSINESS_PARTNER
    type: "odata-v2"
    destination:
      name: "cpapp-bupa"
    http:
      suffix: "/sap/opu/odata/sap"

Run your application with the Destination service:

cds bind --exec -- mvn spring-boot:run \
  -Dspring-boot.run.profiles=default,hybrid

Learn more about cds bind --exec.

TIP

If you are developing in the Business Application Studio and want to connect to an on-premise system, you will need to do so via Business Application Studio's built-in proxy, for which you need to add configuration to your destination environment variable. See Reach On-Premise Service from the SAP Business Application Studio for more details.

Connect to an Application Using the Same XSUAA (Forward Authorization Token)

If your application consists of microservices and you use one (or more) as a remote service as described in this guide, you can leverage the same XSUAA instance. In that case, you don't need an SAP BTP destination at all.

Assuming that your microservices use the same XSUAA instance, you can just forward the authorization token. The URL of the remote service can be injected into the application in the MTA or Cloud Foundry deployment using application defined destinations.

Forward Authorization Token with Node.js

To enable the token forwarding, set the forwardAuthToken option to true in your application defined destination:

{
  "requires": {
    "kind": "odata",
    "model": "./srv/external/OrdersService",
    "credentials": {
      "url": "<set via env var in deployment>",
      "forwardAuthToken": true
    }
  }
}

Forward Authorization Token with Java

For Java, you set the authentication type to TOKEN_FORWARDING for the destination.

You can implement it in your code:

urlFromConfig = ...; // read from config
DefaultHttpDestination mockDestination = DefaultHttpDestination
    .builder(urlFromConfig)
    .name("order-service")
    .authenticationType(AuthenticationType.TOKEN_FORWARDING)
    .build();

Or declare the destination in your application.yaml file:

srv/src/main/resources/application.yaml

cds:
  remote.services:
    order-service:
      type: "odata-v4"
      destination:
        properties:
          url: "<set via env var in deployment>"
          authentication: TokenForwarding

Alternatively to setting the authentication type, you can set the property forwardAuthToken to true.

Connect to an Application in Your Kyma Cluster

The Istio service mesh provides secure communication between the services in your service mesh. You can access a service in your applications' namespace by just reaching out to http://<service-name> or using the full hostname http://<service-name>.<namespace>.svc.cluster.local. Istio sends the requests through an mTLS tunnel.

With Istio, you can further secure the communication by configuring authentication and authorization for your services

Deployment

Your micro service needs bindings to the XSUAA and Destination service to access destinations on SAP BTP. If you want to access an on-premise service using Cloud Connector, then you need a binding to the Connectivity service as well.

Learn more about deploying CAP applications.Learn more about deploying an application using the end-to-end tutorial.

Add Required Services to MTA Deployments

The MTA-based deployment is described in the deployment guide. You can follow this guide and make some additional adjustments to the generated mta.yml file.

cds add xsuaa,destination,connectivity --for production

Learn what this does in the background...

1. Adds XSUAA, Destination, and Connectivity services to your mta.yaml:
   mta.yml

   - name: cpapp-uaa
     type: org.cloudfoundry.managed-service
     parameters:
       service: xsuaa
       service-plan: application
       path: ./xs-security.json
   
   - name: cpapp-destination
     type: org.cloudfoundry.managed-service
     parameters:
       service: destination
       service-plan: lite
   
   # Required for on-premise connectivity only
   - name: cpapp-connectivity
     type: org.cloudfoundry.managed-service
     parameters:
       service: connectivity
       service-plan: lite

2. Requires the services for your server in the mta.yaml:
   mta.yaml

   - name: cpapp-srv
     ...
     requires:
       ...
       - name: cpapp-uaa
       - name: cpapp-destination
       - name: cpapp-connectivity # Required for on-premise connectivity only

Build your application:

mbt build -t gen --mtar mta.tar

Now you can deploy it to Cloud Foundry:

cf deploy gen/mta.tar

Connectivity Service Credentials on Kyma

The secret of the connectivity service on Kyma needs to be modified for the Cloud SDK to connect to on-premise destinations.

Support for Connectivity Service Secret in JavaSupport for Connectivity Service Secret in Node.js

Destinations and Multitenancy

With the destination service, you can access destinations in your provider account, the account your application is running in, and destinations in the subscriber accounts of your multitenant-aware application.

Use Destinations from Subscriber Account

Customers want to see business partners from, for example, their SAP S/4 HANA system.

As provider, you need to define a name for a destination, which enables access to systems of the subscriber of your application. In addition, your multitenant application or service needs to have a dependency to the destination service. For destinations in an on-premise system, the connectivity service must be bound.

The subscriber needs to create a destination with that name in their subscriber account, for example, pointing to their SAP S/4HANA system.

Destination Resolution

The destination is read from the tenant of the request's JWT (authorization) token. If no JWT token is present, the destination is read from the tenant of the application's XSUAA binding.

The destination is read from the tenant of the request's JWT (authorization) token. If no JWT token is present or the destination isn't found, the destination is read from the tenant of the application's XSUAA binding.

JWT token vs. XSUAA binding

Using the tenant of the request's JWT token means reading from the subscriber subaccount for a multitenant application. The tenant of the application's XSUAA binding points to the destination of the provider subaccount, the account where the application is deployed to.

You can change the destination lookup behavior as follows:

"cds": {
  "requires": {
    "SERVICE_FOR_PROVIDER": {
      /* ... */
      "credentials": {
        /* ... */
      },
      "destinationOptions": {
        "selectionStrategy": "alwaysProvider",
        "jwt": null
      }
    }
  }
}

Setting the selectionStrategy property for the destination options to alwaysProvider, you can ensure that the destination is always read from your provider subaccount. With that you ensure that a subscriber cannot overwrite your destination.

Set the destination option jwt to null, if you don't want to pass the request's JWT to SAP Cloud SDK. Passing the request's JWT to SAP Cloud SDK has implications on, amongst others, the effective defaults for selection strategy and isolation level. In rare cases, these defaults are not suitable, for example when the request to the upstream server does not depend on the current user. Please see Authentication and JSON Web Token (JWT) Retrieval for more details.

For Java use the property retrievalStrategy in the destination configuration, to ensure that the destination is always read from your provider subaccount:

cds:
  remote.services:
    service-for-provider:
      type: "odata-v4"
      destination:
        retrievalStrategy: "AlwaysProvider"

Read more in the full reference of all supported retrieval strategy values. Please note that the value must be provided in pascal case, for example: AlwaysProvider.

Add Qualities

Resilience

There are two ways to make your outbound communications resilient:

1. Run your application in a service mesh (for example, Istio, Linkerd, etc.). For example, Kyma is provided as service mesh.
2. Implement resilience in your application.

Refer to the documentation for the service mesh of your choice for instructions. No code changes should be required.

To build resilience into your application, there are libraries to help you implement functions, like doing retries, circuit breakers or implementing fallbacks.

You can use the resilience features provided by the SAP Cloud SDK with CAP Java. You need to wrap your remote calls with a call of ResilienceDecorator.executeSupplier and a resilience configuration (ResilienceConfiguration). Additionally, you can provide a fallback function.

ResilienceConfiguration config = ResilienceConfiguration.of(AdminServiceAddressHandler.class)
  .timeLimiterConfiguration(TimeLimiterConfiguration.of(Duration.ofSeconds(10)));

context.setResult(ResilienceDecorator.executeSupplier(() ->  {
  // ..to access the S/4 system in a resilient way..
  logger.info("Delegating GET Addresses to S/4 service");
  return bupa.run(select);
}, config, (t) -> {
  // ..falling back to the already replicated addresses in our own database
  logger.warn("Falling back to already replicated Addresses");
  return db.run(select);
}));

See the full example

There's no resilience library provided out of the box for CAP Node.js. However, you can use packages provided by the Node.js community. Usually, they provide a function to wrap your code that adds the resilience logic.

Resilience in Kyma

Kyma clusters run an Istio service mesh. Istio allows to configure resilience for the network destinations of your service mesh.

Tracing

CAP adds headers for request correlation to its outbound requests that allows logging and tracing across micro services.

Learn more about request correlation in Node.js.Learn more about request correlation in Java.

Feature Details

Legend

Tag	Explanation
✓	supported
✗	not supported

Supported Protocols

Protocol	Java	Node.js
odata-v2	✓	✓
odata-v4	✓	✓
rest	✗	✓

TIP

The Node.js runtime supports odata as an alias for odata-v4 as well.

Querying API Features

Feature	Java	Node.js
READ	✓	✓
INSERT/UPDATE/DELETE	✓	✓
Actions	✓	✓
columns	✓	✓
where	✓	✓
orderby	✓	✓
limit (top & skip)	✓	✓
$apply (groupedby, ...)	✗	✗
$search (OData v4)	✓	✓
search (SAP OData v2 extension)	✓	✓

Supported Projection Features

Feature	Java	Node.js
Resolve projections to remote services	✓	✓
Resolve multiple levels of projections to remote services	✓	✓
Aliases for fields	✓	✓
excluding	✓	✓
Resolve associations (within the same remote service)	✓	✓
Redirected associations	✓	✓
Flatten associations	✗	✗
where conditions	✗	✗
order by	✗	✗
Infix filter for associations	✗	✗
Model Associations with mixins	✓	✓

Supported Features for Application Defined Destinations

The following properties and authentication types are supported for application defined destinations:

Properties

These destination properties are fully supported by both, the Java and the Node.js runtime.

TIP

This list specifies the properties for application defined destinations.

Properties	Description
url
authentication	Authentication type
username	User name for BasicAuthentication
password	Password for BasicAuthentication
headers	Map of HTTP headers
queries	Map of URL parameters
forwardAuthToken	Forward auth token

Destination Type in SAP Cloud SDK for JavaScriptHttpDestination Type in SAP Cloud SDK for Java

Authentication Types

Authentication Types	Java	Node.js
NoAuthentication	✓	✓
BasicAuthentication	✓	✓
TokenForwarding	✓	✗ Use forwardAuthToken
OAuth2ClientCredentials	code only	✗
UserTokenAuthentication	code only	✗