Change Tracking

The feature tracks the changes of all modifying operations executed via CQN statements, which are indirectly triggered by the protocol adapters or directly by a custom code. Changes made through the native SQL, JDBC, or other means that bypass the CAP Java runtime or that are forwarded to the remote services aren't tracked.

Enabling Change Tracking

To use the change tracking feature, you need to add a dependency to cds-feature-change-tracking in the pom.xml file of your service:

<dependency>
  <groupId>com.sap.cds</groupId>
    <artifactId>cds-feature-change-tracking</artifactId>
    <scope>runtime</scope>
</dependency>

• Your POM must also include the goal to resolve the CDS model delivered from the feature. See Reference the New CDS Model in an Existing CAP Java Project.

• If you use SAP Fiori elements as your UI framework and intend to use the built-in UI, update your SAP UI5 version to 1.121.2 or higher.

Annotating Entities

To capture changes for an entity, you need to extend it with a technical aspect and annotate it with the annotation @changelog that declares the elements whose changes are to be logged.

Given the following entity that represents a book on the domain level:

namespace model;

entity Books {
    key ID: UUID;
    title: String;
    stock: Integer;
}

And the corresponding service definition with the projection of the entity:

namespace srv;

using {model} from '../db/schema'; // Our domain model

service Bookshop {
    entity Books as projection on model.Books;
}

Include the change log model that is provided by this feature:

using {sap.changelog as changelog} from 'com.sap.cds/change-tracking';

Extend the domain entity with the aspect changelog.changeTracked like this:

extend model.Books with changelog.changeTracked;

This aspect adds the association changes that lets you consume the change log both programmatically via CQN statements and in the UI. This implies that every projection of the entity Books has this association and the changes will be visible in all of them.

Annotate elements of the entity that you want to track with the @changelog annotation:

annotate Bookshop.Books {
  title @changelog;
  stock @changelog;
};

Your complete service definition should look like this:

namespace srv;

using {sap.changelog as changelog} from 'com.sap.cds/change-tracking';
using {model} from '../db/schema';

// The domain entity extended with change tracking aspect.
extend model.Books with changelog.changeTracked;

service Bookshop {
    entity Books as projection on model.Books;
}

// Projection is annotated to indicate which elements are change tracked.
annotate Bookshop.Books {
  title @changelog;
  stock @changelog;
};

Personal data is ignored

Elements with personal data, that is, elements that are annotated with @PersonalData and hence subject to audit logging, are ignored by the change tracking.

The level where you annotate your elements with the annotation @changelog is very important. If you annotate the elements on the domain level, every change made through every projection of the entity is tracked. If you annotate the elements on the service level, only the changes made through that projection are tracked.

In case of the books example above, the changes made through the service entity Bookshop.Books are tracked, but the changes made on the domain entity are omitted. That can be beneficial if you have a service that is used for data replication or mass changes where change tracking can be a very expensive operation, and you do not want to generate changes from such operations.

Change tracking also works with the entities that have compositions and tracks the changes made to the items of the compositions.

For example, if you have an entity that represents the order with a composition that represents the items of the order, you can annotate the elements of both and track the changes made through the order and the items in a deep update.

entity OrderItems {
  key ID: UUID;
  [...]
  quantity: Integer @changelog;
}

entity Orders {
  key ID: UUID;
  customerName: String @changelog;
  [...]
  items: Composition of many OrderItems;
}

Identifiers for Changes

You can store some elements of the entity together with the changes in the change log to produce a user-friendly identifier. You define this identifier by annotating the entity with the @changelog annotation and including the elements that you want to store together with the changed value:

annotate Bookshop.Book with @changelog: [
  title
];

This identifier can contain the elements of the entity or values of to-one associations that are reachable via path. For example, for a book you can store an author name if you have an association from the book to the author.

When you define the identifier for an entity, keep in mind that the projections of the annotated entity will inherit the annotation @changelog. If you change the structure of the projection, for example, exclude or rename the elements that are used in the identifier, you must annotate the projection again to provide updated element names in the identifier.

The best candidates for identifier elements are the elements that are insert-only or that don't change often.

Stored as-is

The values of the identifier are stored together with the change log as-is. They are not translated and some data types might not be formatted per user locale or some requirements, for example, different units of measurement or currencies. You should consider this when you decide what to include in the identifier.

Identifiers for Associated Entities

When your entity has an association to an other entity, you might want to log the changes in their relationship.

Given the Orders entity with an association to a Customer instead of the element with customer name:

entity Orders {
  key ID: UUID;
  customer: Association to Customer;
  [...]
}

If you annotate such an association with @changelog, by default, the change log stores the value of the associated entity key. If you want, you can store some human-readable identifier instead. You define this by annotating the association with an own identifier:

annotate Orders {
  customer @changelog: [ customer.name ]
}

Elements from the @changelog annotation value must always be prefixed by the association name. The same caveats as for the identifiers for the entities apply here.

If you annotate a composition with an identifier, the change log will contain an entry with the identifier's value. Additionally, it will include change log entries for all annotated elements of the composition's target entity.

Validation required

If the target of the association is missing, for example, when an entity is updated with the ID for a customer that does not exists, the changelog entry will not be created. You need to validate such cases in the custom code or use annotations, for example, @assert.target.

This feature can also be used for to-many compositions, when you don't need to track the deep changes, but still want to track the additions and removals in the composition.

With association identifiers you also must consider the changes in your entities structure along the projections. In case your target entity is exposed using different projections with removed or renamed elements, you also need to adjust the identifier accordingly in the source entity.

Displaying Changes

The changes of the entity are exposed as an association changes that you can use to display the change log in the UI. By default, the entity Changes is auto-exposed, but it won't be writable via OData requests.

If you want to display the change log together with the overview of your entity, you need to add the facet to the object page that displays the changes:

annotate Bookshop.Books with @(
  UI : { ...
    Facets : [ ...
       {
          $Type               : 'UI.ReferenceFacet',
          ID                  : 'ChangeHistoryFacet',
          Label               : '{i18n>ChangeHistory}',
          Target              : 'changes/@UI.PresentationVariant',
          ![@UI.PartOfPreview]: false
        } ...
   ] ...
  } ...);

If you want to have a common UI for all changes, you need to expose the change log as a projection and define your own presentation for it as the changes are exposed only as part of the change-tracked entity. This projection must be read-only and shouldn't be writable via OData requests.

The change log is extended with the texts for your entities from the @title annotation and the element. Otherwise, the change log contains only the technical names of the entities and the elements. Titles are translated, if they're annotated as translatable. See Externalizing Texts Bundles for more information.

How Changes are Stored

The namespace sap.changelog defines an entity Changes that reflects each change, so the changes are stored in a flat table for all entities together.

Each entry in the Changes entity contains the following information:

• A marker that represents the nature of the change: addition, modification, or deletion.
• The qualified name of the entity that was changed and the qualified name of the root entity. They depend on the projection that was used to change the entity and reflect the root and a target of the modifying operation. For flat entities, they're the same.
• The attribute of the target projection that was changed.
• The new and old values as strings.
• The user who made the change and the timestamp of the change.
• The data type of the changed attribute.
• The technical path from the root entity to the tracked target entity.

Detection of Changes

The change tracking intercepts the modifying CQL statements (Insert, Upsert, Update, and Delete) and requires additional READ events to retrieve the old and the new image of the entity.

These two images are compared and differences are stored in the change log. The nature of the change is determined by comparing the old and new values of the entity: data that weren't present in the old values are considered as added whereas data that aren't present in the new values are considered as deleted. Elements that are present in both old and new values but have different values are considered as modified. Each change detected by the change tracking feature is stored in the change log as a separate entry.

In the case of the deeply structured documents, for example, entities with the compositions, the change tracking feature detects the changes across the complete document and stores them in the change log with the metadata reflecting the structure of the change.

For example, given the order and item model from above, if you change values for the tracked elements with the deep update, for example, the customer name in the order and the quantity of the item, the change log contains two entries: one for the order and one for the item. The change log entry for the item will also reflect that the root of the change is an order.

Prefer deep updates for change tracked entities

If you change the values of the OrderItems entity directly via an OData request or a CQL statement, the change log contains only one entry for the item and won't be associated with an order.

Reacting on Changes

You can write an event handler to observe the change log entries. Keep in mind, that the change log entries are created for each statement and this event will not be bound to any kind of transaction or a batch operation.

import cds.gen.sap.changelog.Changes;

@Component
@ServiceName("ChangeTrackingService$Default")
public class ChangeTrackingHandler implements EventHandler {
	
  @After(event = "createChanges")
  void afterCreate(EventContext context) {
    Result result = (Result) context.get("result");
    result.listOf(Changes.class).forEach(c -> {
      // Do something with the change log entry
	});
  }
}

You can query the change log entries via CQN statements, as usual.

Things to Consider when Using Change Tracking

• Consider the storage costs of the change log. The change log can grow very fast and can consume a lot of space in case of frequent changes. You should consider the retention policy of the change log as it won't be deleted when you delete the entities.
• Consider the performance impact. Change tracking needs to execute additional reads during updates to retrieve and compare updated values. This can slow down the update operations and can be very expensive in the case of updates that affect a lot of entities.
• Consider the ways your entities are changed. You might want to track the changes only on the service projection level that are used for the user interaction and not on the domain level (for instance during data replication).
• If you want to expose the complete change log to the user, you need to consider the security implications of this. If your entities have complex access rules, you need to consider how to extend these rules to the change log.