capire

Appearance

Providing Services ​

This guide introduces how to define and implement services, leveraging generic implementations provided by the CAP runtimes, complemented by domain-specific custom logic.

Intro: Core Concepts ​

The following sections give a brief overview of CAP's core concepts.

Service-Centric Paradigm ​

A CAP application commonly provides services defined in CDS models and served by the CAP runtimes. Every active thing in CAP is a service. They embody the behavioral aspects of a domain in terms of exposed entities, actions, and events.

This graphic is explained in the accompanying text.

Ubiquitous Events ​

At runtime, everything happening is in response to events. CAP features a ubiquitous notion of events, which represent both, requests coming in through synchronous APIs, as well as asynchronous event messages, blurring the line between both worlds.

This graphic shows that consumers send events to services and that there are hooks, so that event handlers can react on those events.

Event Handlers ​

Service providers basically react on events in event handlers, plugged in to respective hooks provided by the core service runtimes.

Service Definitions ​

Services as APIs ​

In its most basic form, a service definition simply declares the data entities and operations it serves. For example:

cds
service BookshopService {

  entity Books {
    key ID : UUID;
    title  : String;
    author : Association to Authors;
  }

  entity Authors {
    key ID : UUID;
    name   : String;
    books  : Association to many Books on books.author = $self;
  }

  action submitOrder (book : Books:ID, quantity : Integer);

}

This definition effectively defines the API served by BookshopService.

This graphic is explained in the accompanying text.

Simple service definitions like that are all we need to run full-fledged servers out of the box, served by CAP's generic runtimes, without any implementation coding required.

Services as Facades ​

In contrast to the all-in-one definition above, services usually expose views, aka projections, on underlying domain model entities:

cds
using { sap.capire.bookshop as my } from '../db/schema';
service BookshopService {
  entity Books as projection on my.Books;
  entity Authors as projection on my.Authors;
  action submitOrder (book : Books:ID, quantity : Integer);
}

This way, services become facades to encapsulated domain data, exposing different aspects tailored to respective use cases.

This graphic is explained in the accompanying text.

Denormalized Views ​

Instead of exposing access to underlying data in a 1:1 fashion, services frequently expose denormalized views, tailored to specific use cases.

For example, the following service definition, undiscloses information about maintainers from end users and also marks the entities as @readonly:

cds
using { sap.capire.bookshop as my } from '../db/schema';

/** For serving end users */
service CatalogService @(path:'/browse') {

  /** For displaying lists of Books */
  @readonly entity ListOfBooks as projection on Books
  excluding { descr };

  /** For display in details pages */
  @readonly entity Books as projection on my.Books { *,
    author.name as author
  } excluding { createdBy, modifiedBy };

}

Learn more about CQL the language used for projections.See also: Prefer Single-Purposed Services!Find above sources in cap/samples.

Auto-Exposed Entities ​

Annotate entities with @cds.autoexpose to automatically include them in services containing entities with Association referencing to them. For example, this is commonly done for code list entities in order to serve Value Lists dropdowns on UIs:

cds
service Zoo {
  entity Foo { //...
    code : Association to SomeCodeList;
  }
}
@cds.autoexpose entity SomeCodeList {...}

Learn more about Auto-Exposed Entities in the CDS reference docs.

Redirected Associations ​

When exposing related entities, associations are automatically redirected. This ensures that clients can navigate between projected entities as expected. For example:

cds
service AdminService {
  entity Books as projection on my.Books;
  entity Authors as projection on my.Authors;
  //> AdminService.Authors.books refers to AdminService.Books
}

Learn more about Redirected Associations in the CDS reference docs.

Generic Providers ​

The CAP runtimes for Node.js and Java provide a wealth of generic implementations, which serve most requests automatically, with out-of-the-box solutions to recurring tasks such as search, pagination, or input validation — the majority of this guide focuses on these generic features.

In effect, a service definition as introduced above is all we need to run a full-fledged server out of the box. The need for coding reduces to real custom logic specific to a project's domain → section Custom Logic picks that up.

Serving CRUD Requests ​

The CAP runtimes for Node.js and Java provide generic handlers, which automatically serve all CRUD requests to entities for CDS-modelled services on top of a default primary database.

This comprises read and write operations like that:

  • GET /Books/201 → reading single data entities
  • GET /Books?... → reading data entity sets with advanced query options
  • POST /Books {....} → creating new data entities
  • PUT/PATCH /Books/201 {...} → updating data entities
  • DELETE /Books/201 → deleting data entities

No filtering and sorting for virtual elements

CAP runtimes delegate filtering and sorting to the database. Therefore filtering and sorting is not available for virtual elements.

Deep Reads and Writes ​

CDS and the runtimes have advanced support for modeling and serving document-oriented data. The runtimes provide generic handlers for serving deeply nested document structures out of the box as documented in here.

Deep READ ​

You can read deeply nested documents by expanding along associations or compositions. For example, like this in OData:

http
GET .../Orders?$expand=header($expand=items)
js
SELECT.from ('Orders', o => {
  o.ID, o.title, o.header (h => {
    h.ID, h.status, h.items('*')
  })
})

Learn more about cds.ql

Both would return an array of nested structures as follows:

js
[{
    ID:1, title: 'first order', header: { // to-one
      ID:2, status: 'open', items: [{     // to-many
        ID:3, description: 'first order item'
      },{
        ID:4, description: 'second order item'
      }]
    }
  },
  ...
]

Deep INSERT ​

Create a parent entity along with child entities in a single operation, for example, like that:

http
POST .../Orders {
  ID:1, title: 'new order', header: { // to-one
    ID:2, status: 'open', items: [{   // to-many
      ID:3, description: 'child of child entity'
    },{
      ID:4, description: 'another child of child entity'
    }]
  }
}

Note that Associations and Compositions are handled differently in (deep) inserts and updates:

  • Compositions → runtime deeply creates or updates entries in target entities
  • Associations → runtime fills in foreign keys to existing target entries

For example, the following request would create a new Book with a reference to an existing Author, with {ID:12} being the foreign key value filled in for association author:

http
POST .../Books {
  ID:121, title: 'Jane Eyre', author: {ID:12}
}

Deep UPDATE ​

Deep UPDATE of the deeply nested documents look very similar to deep INSERT:

http
PUT .../Orders/1 {
  title: 'changed title of existing order', header: {
    ID:2, items: [{
      ID:3, description: 'modified child of child entity'
    },{
      ID:5, description: 'new child of child entity'
    }]
  }]
}

Depending on existing data, child entities will be created, updated, or deleted as follows:

  • entries existing on the database, but not in the payload, are deleted → for example, ID:4
  • entries existing on the database, and in the payload are updated → for example, ID:3
  • entries not existing on the database are created → for example, ID:5

PUT vs PATCH — Omitted fields get reset to default values or null in case of PUT requests; they are left untouched for PATCH requests.

Omitted compositions have no effect, whether during PATCH or during PUT. That is, to delete all children, the payload must specify null or [], respectively, for the to-one or to-many composition.

Deep DELETE ​

Deleting a root of a composition hierarchy results in a cascaded delete of all nested children.

sql
DELETE .../Orders/1  -- would also delete all headers and items

Auto-Generated Keys ​

On CREATE operations, key elements of type UUID are filled in automatically. In addition, on deep inserts and upserts, respective foreign keys of newly created nested objects are filled in accordingly.

For example, given a model like that:

cds
entity Orders {
  key ID : UUID;
  title  : String;
  Items  : Composition of many OrderItems on Items.order = $self;
}
entity OrderItems {
  key order : Association to Orders;
  key pos  : Integer;
  descr: String;
}

When creating a new Order with nested OrderItems like that:

js
POST .../Orders {
  title: 'Order #1', Items: [
    { pos:1, descr: 'Item #1' },
    { pos:2, descr: 'Item #2' }
  ]
}

CAP runtimes will automatically fill in Orders.ID with a new uuid, as well as the nested OrderItems.order.ID referring to the parent.

Searching Data ​

CAP runtimes provide out-of-the-box support for advanced search of a given text in all textual elements of an entity including nested entities along composition hierarchies.

A typical search request looks like that:

js
GET .../Books?$search=Heights

That would basically search for occurrences of "Heights" in all text fields of Books, that is, in title and descr using database-specific contains operations (for example, using like '%Heights%' in standard SQL).

By default search is limited to the elements of type String of an entity that aren't calculated or virtual. Yet, sometimes you may want to deviate from this default and specify a different set of searchable elements, or to extend the search to associated entities. Use the @cds.search annotation to do so. The general usage is:

cds
@cds.search: {
    element1,         // included
    element2 : true,  // included
    element3 : false, // excluded
    assoc1,           // extend to searchable elements in target entity
    assoc2.elementA   // extend to a specific element in target entity
}
entity E { }

Learn more about the syntax of annotations.

Including Fields ​

cds
@cds.search: { title }
entity Books { ... }

Searches the title element only.

Extend Search to Associated Entities ​

Node.js: Only w/ streamlined database services

For Node.js projects, this feature is only available with the streamlined @cap-js/ database services (default with @sap/cds >= 8)

cds
@cds.search: { author }
entity Books { ... }

@cds.search: { biography: false }
entity Authors { ... }

Searches all elements of the Books entity, as well as all searchable elements of the associated Authors entity. Which elements of the associated entity are searchable is determined by the @cds.search annotation on the associated entity. So, from Authors, all elements of type String are searched but biography is excluded.

Extend to Individual Elements in Associated Entities ​
cds
@cds.search: { author.name }
entity Books { ... }

Searches only in the element name of the associated Authors entity.

Excluding Fields ​

cds
@cds.search: { isbn: false }
entity Books { ... }

Searches all elements of type String excluding the element isbn, which leaves the title and descr elements to be searched.

TIP

You can explicitly annotate calculated elements to make them searchable, even though they aren't searchable by default. The virtual elements won't be searchable even if they're explicitly annotated.

Prerequisite: For CAP Java, you need to run in HEX optimization mode on SAP HANA Cloud and enable cds.sql.hana.search.fuzzy

Fuzzy search is a fault-tolerant search feature of SAP HANA Cloud, which returns records even if the search term contains additional characters, is missing characters, or has typographical errors.

You can configure the fuzziness in the range [0.0, 1.0]. The value 1.0 enforces exact search.

  • Java: cds.sql.hana.search.fuzzinessThreshold
  • Node.js:cds.hana.fuzzy

Override the fuzziness for elements, using the @Search.fuzzinessThreshold annotation:

cds
entity Books {
   @Search.fuzzinessThreshold: 0.7
   title : String;
}

The relevance of a search match depends on the weight of the element causing the match. By default, all searchable elements have equal weight. To adjust the weight of an element, use the @Search.ranking annotation. Allowed values are HIGH, MEDIUM (default), and LOW:

cds
entity Books {
   @Search.ranking: HIGH
   title         : String;

   @Search.ranking: LOW
   publisherName : String;
}

Wildcards in search terms

When using wildcards in search terms, an exact pattern search is performed. Supported wildcards are '*' matching zero or more characters and '?' matching a single character. You can escape wildcards using '\'.

Pagination & Sorting ​

Implicit Pagination ​

By default, the generic handlers for READ requests automatically truncate result sets to a size of 1,000 records max. If there are more entries available, a link is added to the response allowing clients to fetch the next page of records.

The OData response body for truncated result sets contains a nextLink as follows:

http
GET .../Books
>{
  value: [
    {... first record ...},
    {... second record ...},
    ...
  ],
  @odata.nextLink: "Books?$skiptoken=1000"
}

To retrieve the next page of records from the server, the client would use this nextLink in a follow-up request, like so:

http
GET .../Books?$skiptoken=1000

On firing this query, you get the second set of 1,000 records with a link to the next page, and so on, until the last page is returned, with the response not containing a nextLink.

WARNING

Per OData specification for Server Side Paging, the value of the nextLink returned by the server must not be interpreted or changed by the clients.

Reliable Pagination ​

Note: This feature is available only for OData V4 endpoints.

Using a numeric skip token based on the values of $skip and $top can result in duplicate or missing rows if the entity set is modified between the calls. Reliable Pagination avoids this inconsistency by generating a skip token based on the values of the last row of a page.

The reliable pagination is available with following limitations:

  • Results of functions or arithmetic expressions can't be used in the $orderby option (explicit ordering).
  • The elements used in the $orderby of the request must be of simple type.
  • All elements used in $orderby must also be included in the $select option, if it's set.
  • Complex concatenations of result sets aren't supported.

WARNING

Don't use reliable pagination if an entity set is sorted by elements that contain sensitive information, the skip token could reveal the values of these elements.

The feature can be enabled with the following configuration options set to true:

  • Java: cds.query.limit.reliablePaging.enabled
  • Node.js: cds.query.limit.reliablePaging

Paging Limits ​

You can configure default and maximum page size limits in your project configuration as follows:

json
"cds": {
  "query": {
    "limit": {
      "default": 20, //> no default
      "max": 100     //> default 1000
    }
  }
}
  • The maximum limit defines the maximum number of items that can get retrieved, regardless of $top.
  • The default limit defines the number of items that are retrieved if no $top was specified.
Annotation @cds.query.limit ​

You can override the defaults by applying the @cds.query.limit annotation on the service or entity level, as follows:

cds
@cds.query.limit: { default?, max? } | Number

The limit definitions for CatalogService and AdminService in the following example are equivalent.

cds
@cds.query.limit.default: 20
@cds.query.limit.max: 100
service CatalogService {
  // ...
}
@cds.query.limit: { default: 20, max: 100 }
service AdminService {
  // ...
}

@cds.query.limit can be used as shorthand if no default limit needs to be specified at the same level.

cds
@cds.query.limit: 100
service CatalogService {
  entity Books as projection on my.Books;     //> pages at 100
  @cds.query.limit: 20
  entity Authors as projection on my.Authors; //> pages at 20
}
service AdminService {
  entity Books as projection on my.Books;     //> pages at 1000 (default)
}
Precedence ​

The closest limit applies, that means, an entity-level limit overrides that of its service, and a service-level limit overrides the global setting. The value 0 disables the respective limit at the respective level.

cds
@cds.query.limit.default: 20
service CatalogService {
  @cds.query.limit.max: 100
  entity Books as projection on my.Books;     //> default = 20 (from CatalogService), max = 100
  @cds.query.limit: 0
  entity Authors as projection on my.Authors; //> no default, max = 1,000 (from environment)
}

Implicit Sorting ​

Paging requires implied sorting, otherwise records might be skipped accidentally when reading follow-up pages. By default the entity's primary key is used as a sort criterion.

For example, given a service definition like this:

cds
service CatalogService {
  entity Books as projection on my.Books;
}

The SQL query executed in response to incoming requests to Books will be enhanced with an additional order-by clause as follows:

sql
SELECT ... from my_Books
ORDER BY ID; -- default: order by the entity's primary key

If the request specifies a sort order, for example, GET .../Books?$orderby=author, both are applied as follows:

sql
SELECT ... from my_Books ORDER BY
  author,     -- request-specific order has precedence
  ID;         -- default order still applied in addition

We can also define a default order when serving books as follows:

cds
service CatalogService {
  entity Books as projection on my.Books order by title asc;
}

Now, the resulting order by clauses are as follows for GET .../Books:

sql
SELECT ... from my_Books ORDER BY
  title asc,  -- from entity definition
  ID;         -- default order still applied in addition

... and for GET .../Books?$orderby=author:

sql
SELECT ... from my_Books ORDER BY
  author,     -- request-specific order has precedence
  title asc,  -- from entity definition
  ID;         -- default order still applied in addition

Concurrency Control ​

CAP runtimes support different ways to avoid lost-update situations as documented in the following.

Use optimistic locking to detect concurrent modification of data across requests. The implementation relies on ETags.

Use pessimistic locking to protect data from concurrent modification by concurrent transactions. CAP leverages database locks for pessimistic locking.

Conflict Detection Using ETags ​

The CAP runtimes support optimistic concurrency control and caching techniques using ETags. An ETag identifies a specific version of a resource found at a URL.

Enable ETags by adding the @odata.etag annotation to an element to be used to calculate an ETag value as follows:

cds
using { managed } from '@sap/cds/common';
entity Foo : managed {...}
annotate Foo with { modifiedAt @odata.etag }

The value of an ETag element should uniquely change with each update per row. The modifiedAt element from the pre-defined managed aspect is a good candidate, as this is automatically updated. You could also use update counters or UUIDs, which are recalculated on each update.

You use ETags when updating, deleting, or invoking the action bound to an entity by using the ETag value in an If-Match or If-None-Match header. The following examples represent typical requests and responses:

http
POST Employees { ID:111, name:'Name' }
> 201 Created {'@odata.etag': 'W/"2000-01-01T01:10:10.100Z"',...}
//> Got new ETag to be used for subsequent requests...
http
GET Employees/111
If-None-Match: "2000-01-01T01:10:10.100Z"
> 304 Not Modified // Record was not changed
http
GET Employees/111
If-Match: "2000-01-01T01:10:10.100Z"
> 412 Precondition Failed // Record was changed by another user
http
UPDATE Employees/111
If-Match: "2000-01-01T01:10:10.100Z"
> 200 Ok {'@odata.etag': 'W/"2000-02-02T02:20:20.200Z"',...}
//> Got new ETag to be used for subsequent requests...
http
UPDATE Employees/111
If-Match: "2000-02-02T02:20:20.200Z"
> 412 Precondition Failed // Record was modified by another user
http
DELETE Employees/111
If-Match: "2000-02-02T02:20:20.200Z"
> 412 Precondition Failed // Record was modified by another user

If the ETag validation detects a conflict, the request typically needs to be retried by the client. Hence, optimistic concurrency should be used if conflicts occur rarely.

Pessimistic Locking ​

Pessimistic locking allows you to lock the selected records so that other transactions are blocked from changing the records in any way.

Use exclusive locks when reading entity data with the intention to update it in the same transaction and you want to prevent the data to be read or updated in a concurrent transaction.

Use shared locks if you only need to prevent the entity data to be updated in a concurrent transaction, but don't want to block concurrent read operations.

The records are locked until the end of the transaction by commit or rollback statement.

Learn more about using the SELECT ... FOR UPDATE statement in the Node.js runtime.

Learn more about using the Select.lock() method in the Java runtime.

WARNING

Pessimistic locking is not supported by SQLite. H2 supports exclusive locks only.

Input Validation ​

CAP runtimes automatically validate user input, controlled by the following annotations.

@readonly ​

Elements annotated with @readonly, as well as calculated elements, are protected against write operations. That is, if a CREATE or UPDATE operation specifies values for such fields, these values are silently ignored.

By default virtual elements are also calculated.

TIP

The same applies for fields with the OData Annotations @FieldControl.ReadOnly (static), @Core.Computed, or @Core.Immutable (the latter only on UPDATEs).

@mandatory ​

Elements marked with @mandatory are checked for nonempty input: null and (trimmed) empty strings are rejected.

cds
service Sue {
  entity Books {
    key ID : UUID;
    title  : String @mandatory;
  }
}

In addition to server-side input validation as introduced above, this adds a corresponding @FieldControl annotation to the EDMX so that OData / Fiori clients would enforce a valid entry, thereby avoiding unnecessary request roundtrips:

xml
<Annotations Target="Sue.Books/title">
  <Annotation Term="Common.FieldControl" EnumMember="Common.FieldControlType/Mandatory"/>
</Annotations>

@assert .unique ​

Annotate an entity with @assert.unique.<constraintName>, specifying one or more element combinations to enforce uniqueness checks on all CREATE and UPDATE operations. For example:

cds
@assert.unique: {
  locale: [ parent, locale ],
  timeslice: [ parent, validFrom ],
}
entity LocalizedTemporalData {
  key record_ID : UUID; // technical primary key
  parent    : Association to Data;
  locale    : String;
  validFrom : Date;  validTo : Date;
}

This annotation is applicable to entities, which result in tables in SQL databases only.

The value of the annotation is an array of paths referring to elements in the entity. These elements may be of a scalar type, structs, or managed associations. Individual foreign keys or unmanaged associations are not supported.

If structured elements are specified, the unique constraint will contain all columns stemming from it. If the path points to a managed association, the unique constraint will contain all foreign key columns stemming from it.

TIP

You don't need to specify @assert.unique constraints for the primary key elements of an entity as these are automatically secured by a SQL PRIMARY KEY constraint.

@assert .target ​

Annotate a managed to-one association of a CDS model entity definition with the @assert.target annotation to check whether the target entity referenced by the association (the reference's target) exists. In other words, use this annotation to check whether a non-null foreign key input in a table has a corresponding primary key in the associated/referenced target table.

You can check whether multiple targets exist in the same transaction. For example, in the Books entity, you could annotate one or more managed to-one associations with the @assert.target annotation. However, it is assumed that dependent values were inserted before the current transaction. For example, in a deep create scenario, when creating a book, checking whether an associated author exists that was created as part of the same deep create transaction isn't supported, in this case, you will get an error.

The @assert.target check constraint is meant to validate user input and not to ensure referential integrity. Therefore only CREATE, and UPDATE events are supported (DELETE events are not supported). To ensure that every non-null foreign key in a table has a corresponding primary key in the associated/referenced target table (ensure referential integrity), the @assert.integrity constraint must be used instead.

If the reference's target doesn't exist, an HTTP response (error message) is provided to HTTP client applications and logged to stdout in debug mode. The HTTP response body's content adheres to the standard OData specification for an error response body.

Example ​

Add @assert.target annotation to the service definition as previously mentioned:

cds
entity Books {
  key ID : UUID;
  title  : String;
  author : Association to Authors @assert.target;
}

entity Authors {
  key ID : UUID;
  name   : String;
  books  : Association to many Books on books.author = $self;
}

HTTP Request — assume that an author with the ID "796e274a-c3de-4584-9de2-3ffd7d42d646" doesn't exist in the database

http
POST Books HTTP/1.1
Accept: application/json;odata.metadata=minimal
Prefer: return=minimal
Content-Type: application/json;charset=UTF-8

{"author_ID": "796e274a-c3de-4584-9de2-3ffd7d42d646"}

HTTP Response

http
HTTP/1.1 400 Bad Request
odata-version: 4.0
content-type: application/json;odata.metadata=minimal

{"error": {
  "@Common.numericSeverity": 4,
  "code": "400",
  "message": "Value doesn't exist",
  "target": "author_ID"
}}

TIP

In contrast to the @assert.integrity constraint, whose check is performed on the underlying database layer, the @assert.target check constraint is performed on the application service layer before the custom application handlers are called.

WARNING

Cross-service checks are not supported. It is expected that the associated entities are defined in the same service.

WARNING

The @assert.target check constraint relies on database locks to ensure accurate results in concurrent scenarios. However, locking is a database-specific feature, and some databases don't permit to lock certain kinds of objects. On SAP HANA, for example, views with joins or unions can't be locked. Do not use @assert.target on such artifacts/entities.

@assert .format ​

Allows you to specify a regular expression string (in ECMA 262 format in CAP Node.js and java.util.regex.Pattern format in CAP Java) that all string input must match.

cds
entity Foo {
  bar : String @assert.format: '[a-z]ear';
}

@assert .range ​

Allows you to specify [ min, max ] ranges for elements with ordinal types — that is, numeric or date/time types. For enum elements, true can be specified to restrict all input to the defined enum values.

cds
entity Foo {
  bar : Integer  @assert.range: [ 0, 3 ];
  boo : Decimal  @assert.range: [ 2.1, 10.25 ];
  car : DateTime @assert.range: ['2018-10-31', '2019-01-15'];
  zoo : String   @assert.range enum { high; medium; low; };
}

... with open intervals ​

By default, specified [min,max] ranges are interpreted as closed intervals, that means, the performed checks are min ≤ input ≤ max. You can also specify open intervals by wrapping the min and/or max values into parenthesis like that:

cds
@assert.range: [(0),100]    // 0 < input ≤ 100
@assert.range: [0,(100)]    // 0 ≤ input < 100
@assert.range: [(0),(100)]  // 0 < input < 100

In addition, you can use an underscore _ to represent Infinity like that:

cds
@assert.range: [(0),_]  // positive numbers only, _ means +Infinity here
@assert.range: [_,(0)]  // negative number only, _ means -Infinity here

Basically values wrapped in parentheses (x) can be read as excluding x for min or max. Note that the underscore _ doesn't have to be wrapped into parenthesis, as by definition no number can be equal to Infinity .

Support in latest runtimes

Support for open intervals and infinity has been added to CAP Node.js, i.e. @sap/cds version 8.5. Support in CAP Java is not yet available but will follow soon.

@assert .notNull ​

Annotate a property with @assert.notNull: false to have it ignored during the generic not null check, for example if your persistence fills it automatically.

cds
entity Foo {
  bar : String not null @assert.notNull: false;
}

Custom Logic ​

As most standard tasks and use cases are covered by generic service providers, the need to add service implementation code is greatly reduced and minified, and hence the quantity of individual boilerplate coding.

The remaining cases that need custom handlers, reduce to real custom logic, specific to your domain and application, such as:

  • Domain-specific programmatic Validations
  • Augmenting result sets, for example to add computed fields for frontends
  • Programmatic Authorization Enforcements
  • Triggering follow-up actions, for example calling other services or emitting outbound events in response to inbound events
  • And more... In general, all the things not (yet) covered by generic handlers

In Node.js, the easiest way to add custom implementations for services is through equally named .js files placed next to a service definition's .cds file:

sh
./srv
  - cat-service.cds  # service definitions
  - cat-service.js   # service implementation
...

Learn more about providing service implementations in Node.js.

In Java, you'd assign EventHandler classes using dependency injection as follows:

Java
@Component
@ServiceName("org.acme.Foo")
public class FooServiceImpl implements EventHandler {...}

Learn more about Event Handler classes in Java.

Custom Event Handlers ​

Within your custom implementations, you can register event handlers like that:

js
module.exports = function (){
  this.on ('submitOrder', (req)=>{...}) //> custom actions
  this.on ('CREATE',`Books`, (req)=>{...})
  this.before ('UPDATE',`*`, (req)=>{...})
  this.after ('READ',`Books`, (books)=>{...})
}
Java
@Component
@ServiceName("BookshopService")
public class BookshopServiceImpl implements EventHandler {
  @On(event="submitOrder") public void onSubmitOrder (EventContext req) {...}
  @On(event="CREATE", entity="Books") public void onCreateBooks (EventContext req) {...}
  @Before(event="UPDATE", entity="*") public void onUpdate (EventContext req) {...}
  @After(event="READ", entity="Books") public void onReadBooks (EventContext req) {...}
}

Learn more about adding event handlers in Node.js.

Learn more about adding event handlers in Java.

Hooks: on, before, after ​

In essence, event handlers are functions/method registered to be called when a certain event occurs, with the event being a custom operation, like submitOrder, or a CRUD operation on a certain entity, like READ Books; in general following this scheme:

  • <hook:on|before|after> , <event> , [<entity>] → handler function

CAP allows to plug in event handlers to these different hooks, that is phases during processing a certain event:

  • on handlers run instead of the generic/default handlers.
  • before handlers run before the on handlers
  • after handlers run after the on handlers, and get the result set as input

on handlers form an interceptor stack: the topmost handler getting called by the framework. The implementation of this handler is in control whether to delegate to default handlers down the stack or not.

before and after handlers are listeners: all registered listeners are invoked in parallel. If one vetoes / throws an error the request fails.

Within Event Handlers ​

Event handlers all get a uniform Request/Event Message context object as their primary argument, which, among others, provides access to the following information:

  • The event name — that is, a CRUD method name, or a custom-defined one
  • The target entity, if any
  • The query in CQN format, for CRUD requests
  • The data payload
  • The user, if identified/authenticated
  • The tenant using your SaaS application, if enabled

Learn more about implementing event handlers in Node.js.Learn more about implementing event handlers in Java.

Actions & Functions ​

In addition to common CRUD operations, you can declare domain-specific custom operations as shown below. These custom operations always need custom implementations in corresponding events handlers.

You can define actions and functions in CDS models like that:

cds
service Sue {
  // unbound actions & functions
  function sum (x:Integer, y:Integer) returns Integer;
  function stock (id : Foo:ID) returns Integer;
  action add (x:Integer, to: Integer) returns Integer;

  // bound actions & functions
  entity Foo { key ID:Integer } actions {
    function getStock() returns Integer;
    action order (x:Integer) returns Integer;
    //bound to the collection and not a specific instance of Foo
    action customCreate (in: many $self, x: String) returns Foo;
  }
}

Learn more about modeling actions and functions in CDS.

The differentiation between Actions and Functions as well as bound and unbound stems from the OData specifications, and in essence is as follows:

  • Actions modify data in the server
  • Functions retrieve data
  • Unbound actions/functions are like plain unbound functions in JavaScript.
  • Bound actions/functions always receive the bound entity's primary key as implicit first argument, similar to this pointers in Java or JavaScript. The exception are bound actions to collections, which are bound against the collection and not a specific instance of the entity. An example use case are custom create actions for the SAP Fiori elements UI.

Implementing Actions / Functions ​

In general, implement actions or functions like that:

js
module.exports = function Sue(){
  this.on('sum', ({data:{x,y}}) => x+y)
  this.on('add', ({data:{x,to}}) => stocks[to] += x)
  this.on('stock', ({data:{id}}) => stocks[id])
  this.on('getStock','Foo', ({params:[id]}) => stocks[id])
  this.on('order','Foo', ({params:[id],data:{x}}) => stocks[id] -= x)
}

Event handlers for actions or functions are very similar to those for CRUD events, with the name of the action/function replacing the name of the CRUD operations. No entity is specific for unbound actions/functions.

Method-style Implementations in Node.js, you can alternatively implement actions and functions using conventional JavaScript methods with subclasses of cds.Service:

js
module.exports = class Sue extends cds.Service {
  sum(x,y) { return x+y }
  add(x,to) { return stocks[to] += x }
  stock(id) { return stocks[id] }
  getStock(Foo,id) { return stocks[id] }
  order(Foo,id,x) { return stocks[id] -= x }
}

Calling Actions / Functions ​

HTTP Requests to call the actions/function declared above look like that:

js
GET .../sue/sum(x=1,y=2)              // unbound function
GET .../sue/stock(id=2)               // unbound function
POST .../sue/add {"x":1,"to":2}       // unbound action
GET .../sue/Foo(2)/Sue.getStock()     // bound function
POST .../sue/Foo(2)/Sue.order {"x":1} // bound action

Note: You always need to add the () for functions, even if no arguments are required. The OData standard specifies that bound actions/functions need to be prefixed with the service's name. In the previous example, entity Foo has a bound action order. That action must be called via /Foo(2)/Sue.order instead of simply /Foo(2)/order. For convenience, however, the Node.js runtime also allows calling bound actions/functions without prefixing them with the service name.


Programmatic usage via generic APIs would look like this for Node.js:

js
  const srv = await cds.connect.to('Sue')
  // unbound actions/functions
  await srv.send('sum',{x:1,y:2})
  await srv.send('add',{x:11,to:2})
  await srv.send('stock',{id:2})
  // bound actions/functions
  await srv.send('getStock','Foo',{id:2})
  //for passing the params property, use this syntax
  await srv.send({ event: 'order', entity: 'Foo', data: {x:3}, params: [2]})

Note: Always pass the target entity name as second argument for bound actions/functions.


Programmatic usage via typed API — Node.js automatically equips generated service instances with specific methods matching the definitions of actions/functions found in the services' model. This allows convenient usage like that:

js
  const srv = await cds.connect.to(Sue)
  // unbound actions/functions
  srv.sum(1,2)
  srv.add(11,2)
  srv.stock(2)
  // bound actions/functions
  srv.getStock('Foo',2)
  srv.order('Foo',2,3)

Note: Even with that typed APIs, always pass the target entity name as second argument for bound actions/functions.



Serving Media Data ​

CAP provides out-of-the-box support for serving media and other binary data.

Annotating Media Elements ​

You can use the following annotations in the service model to indicate that an element in an entity contains media data.

@Core.MediaType : Indicates that the element contains media data (directly or using a redirect). The value of this annotation is either a string with the contained MIME type (as shown in the first example), or is a path to the element that contains the MIME type (as shown in the second example).

@Core.IsMediaType : Indicates that the element contains a MIME type. The @Core.MediaType annotation of another element can reference this element.

@Core.IsURL @Core.MediaType : Indicates that the element contains a URL pointing to the media data (redirect scenario).

@Core.ContentDisposition.Filename : Indicates that the element is expected to be displayed as an attachment, that is downloaded and saved locally. The value of this annotation is a path to the element that contains the Filename (as shown in the fourth example ).

@Core.ContentDisposition.Type : Can be used to instruct the browser to display the element inline, even if @Core.ContentDisposition.Filename is specified, by setting to inline (see the fifth example). If omitted, the behavior is @Core.ContentDisposition.Type: 'attachment'.

Learn more how to enable stream support in SAP Fiori elements.

The following examples show these annotations in action:

  1. Media data is stored in a database with a fixed media type image/png:
cds
entity Books { //...
  image : LargeBinary @Core.MediaType: 'image/png';
}
  1. Media data is stored in a database with a variable media type:
cds
entity Books { //...
  image : LargeBinary @Core.MediaType: imageType;
  imageType : String  @Core.IsMediaType;
}
  1. Media data is stored in an external repository:
cds
entity Books { //...
  imageUrl  : String @Core.IsURL @Core.MediaType: imageType;
  imageType : String @Core.IsMediaType;
}
  1. Content disposition data is stored in a database with a variable disposition:
cds
entity Authors { //...
  image : LargeBinary @Core.MediaType: imageType @Core.ContentDisposition.Filename: fileName;
  fileName : String;
}
  1. The image shall have the suggested file name but be displayed inline nevertheless:
cds
entity Authors { //...
  image : LargeBinary @Core.MediaType: imageType @Core.ContentDisposition.Filename: fileName @Core.ContentDisposition.Type: 'inline';
  fileName : String;
}

Learn more about the syntax of annotations.

WARNING

In case you rename the properties holding the media type or content disposition information in a projection, you need to update the annotation's value as well.

Reading Media Resources ​

Read media data using GET requests of the form /Entity(<ID>)/mediaProperty:

cds
GET ../Books(201)/image
> Content-Type: application/octet-stream

The response's Content-Type header is typically application/octet-stream.

Although allowed by RFC 2231, Node.js does not support line breaks in HTTP headers. Hence, make sure you remove any line breaks from your @Core.IsMediaType content.

Read media data with @Core.ContentDisposition.Filename in the model:

cds
GET ../Authors(201)/image
> Content-Disposition: 'attachment; filename="foo.jpg"'

The media data is streamed automatically.

Learn more about returning a custom streaming object (Node.js - beta).

Creating a Media Resource ​

As a first step, create an entity without media data using a POST request to the entity. After creating the entity, you can insert a media property using the PUT method. The MIME type is passed in the Content-Type header. Here are some sample requests:

cds
POST ../Books
Content-Type: application/json
{ <JSON> }
cds
PUT ../Books(201)/image
Content-Type: image/png
<MEDIA>

The media data is streamed automatically.

Updating Media Resources ​

The media data for an entity can be updated using the PUT method:

cds
PUT ../Books(201)/image
Content-Type: image/png
<MEDIA>

The media data is streamed automatically.

Deleting Media Resources ​

One option is to delete the complete entity, including all media data:

cds
DELETE ../Books(201)

Alternatively, you can delete a media data element individually:

cds
DELETE ../Books(201)/image

Using External Resources ​

The following are requests and responses for the entity containing redirected media data from the third example, "Media data is stored in an external repository".

This format is used by OData-Version: 4.0. To be changed in OData-Version: 4.01.

cds
GET: ../Books(201)
>{ ...
    image@odata.mediaReadLink: "http://other-server/image.jpeg",
    image@odata.mediaContentType: "image/jpeg",
    imageType: "image/jpeg"
}

Conventions & Limitations ​

General Conventions ​

  • Binary data in payloads must be a Base64 encoded string.
  • Binary data in URLs must have the format binary'<url-safe base64 encoded>'. For example:
http
GET $filter=ID eq binary'Q0FQIE5vZGUuanM='

Node.js Runtime Conventions and Limitations ​

  • The usage of binary data in some advanced constructs like the $apply query option and /any() might be limited.
  • On SQLite, binary strings are stored as plain strings, whereas a buffer is stored as binary data. As a result, if in a CDS query, a binary string is used to query data stored as binary, this wouldn't work.
  • Please note, that SQLite doesn't support streaming. That means, that LargeBinary fields are read as a whole (not in chunks) and stored in memory, which can impact performance.
  • SAP HANA Database Client for Node.js (HDB) and SAP HANA Client for Node.js (@sap/hana-client) packages handle binary data differently. For example, HDB automatically converts binary strings into binary data, whereas SAP HANA Client doesn't.
  • In the Node.js Runtime, all binary strings are converted into binary data according to SAP HANA property types. To disable this default behavior, you can set the environment variable cds.hana.base64_to_buffer: false.

Best Practices ​

Single-Purposed Services ​

We strongly recommend designing your services for single use cases. Services in CAP are cheap, so there's no need to save on them.

DON'T: Single Services Exposing All Entities 1:1 ​

The anti-pattern to that are single services exposing all underlying entities in your app in a 1:1 fashion. While that may save you some thoughts in the beginning, it's likely that it will result in lots of headaches in the long run:

  • They open huge entry doors to your clients with only few restrictions
  • Individual use-cases aren't reflected in your API design
  • You have to add numerous checks on a per-request basis...
  • Which have to reflect on the actual use cases in complex and expensive evaluations

DO: One Service Per Use Case ​

For example, let's assume that we have a domain model defining Books and Authors more or less as above, and then we add Orders. We could define the following services:

cds
using { my.domain as my } from './db/schema';
cds
/** Serves end users browsing books and place orders */
service CatalogService {
  @readonly entity Books as select from my.Books {
    ID, title, author.name as author
  };
  @requires: 'authenticated-user'
  @insertonly entity Orders as projection on my.Orders;
}
cds
/** Serves registered users managing their account and their orders */
@requires: 'authenticated-user'
service UsersService {
  @restrict: [{ grant: 'READ', where: 'buyer = $user' }] // limit to own ones
  @readonly entity Orders as projection on my.Orders;
  action cancelOrder ( ID:Orders.ID, reason:String );
}
cds
/** Serves administrators managing everything */
@requires: 'authenticated-user'
service AdminService {
  entity Books   as projection on my.Books;
  entity Authors as projection on my.Authors;
  entity Orders  as projection on my.Orders;
}

These services serve different use cases and are tailored for each. Note, for example, that we intentionally don't expose the Authors entity to end users.

Late-Cut Microservices ​

Compared to Microservices, CAP services are 'Nano'. As shown in the previous sections, you should design your application as a set of loosely coupled, single-purposed services, which can all be served embedded in a single-server process at first (that is, a monolith).

Yet, given such loosely coupled services, and enabled by CAP's uniform way to define and consume services, you can decide later on to separate, deploy, and run your services as separate microservices, even without changing your models or code.

This flexibility allows you to, again, focus on solving your domain problem first, and avoid the efforts and costs of premature microservice design and DevOps overhead, at least in the early phases of development.