Integration SAP Datasphere : Utiliser les donnees ABAP Cloud dans l analytique

Catégorie

Integration

Publié

14. Feb. 2026

Auteur

SAP Datasphere est la solution successeur de SAP Data Warehouse Cloud et constitue le coeur de la SAP Business Data Fabric. Pour les developpeurs ABAP Cloud, cela ouvre la possibilite de mettre a disposition des donnees transactionnelles pour des scenarios analytiques complexes.

Fondamentaux SAP Datasphere

SAP Datasphere reunit le Data Warehousing, le Data Lake, la virtualisation de donnees et la semantique metier dans une plateforme integree. L’architecture est basee sur des Spaces qui fonctionnent comme des zones de donnees isolees.

Concepts cles

Concept	Description
Space	Espace de travail isole pour les equipes
Data Builder	Modelisation des vues et tables
Business Builder	Couche semantique et indicateurs
Data Access Controls	Autorisations au niveau des donnees
Connections	Connexions aux systemes sources

Integration avec ABAP Cloud

Les systemes ABAP Cloud peuvent etre connectes de differentes manieres :

ABAP CDS Extraction : Extraction directe des vues CDS
Services OData : Acces via les interfaces OData
Remote Tables : Acces virtuel aux systemes SAP
Replication Flows : Replication de donnees planifiee

Extraction de vues CDS pour Datasphere

L’extraction de vues CDS est la methode recommandee pour ABAP Cloud vers Datasphere. Les vues doivent etre annotees de maniere specifique.

Vues CDS compatibles extraction

@AbapCatalog.sqlViewName: 'ZSALESORDEREX"
@AbapCatalog.compiler.compareFilter: true
@Analytics.dataCategory: #CUBE
@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.changeDataCapture.automatic: true
define view entity ZI_SalesOrderExtract
  as select from zsalesorder as so
  association [1..1] to ZI_CustomerMaster as _Customer
    on $projection.CustomerId = _Customer.CustomerId
{
      @Analytics.dataExtraction.delta.byElement.name: 'ChangedOn"
      @Analytics.dataExtraction.delta.byElement.maxDelayInSeconds: 1800
  key so.sales_order_id    as SalesOrderId,

      so.customer_id       as CustomerId,
      so.order_date        as OrderDate,
      so.net_amount        as NetAmount,
      so.currency          as Currency,
      so.status            as Status,

      @Semantics.systemDateTime.lastChangedAt: true
      so.changed_on        as ChangedOn,

      @Semantics.systemDateTime.createdAt: true
      so.created_on        as CreatedOn,

      _Customer
}

Annotations importantes pour l’extraction

Annotation	Description
`@Analytics.dataExtraction.enabled`	Active l’extraction
`@Analytics.dataExtraction.delta.changeDataCapture.automatic`	CDC automatique
`@Analytics.dataExtraction.delta.byElement.name`	Champ pour la detection delta
`@Analytics.dataCategory`	Categorisation (#CUBE, #DIMENSION, #FACT)

Vues CDS analytiques pour Datasphere

Pour une analytique optimale, les vues CDS doivent etre modelisees selon le schema en etoile.

Vue Dimension (Client)

@AbapCatalog.viewEnhancementCategory: [#NONE]
@Analytics.dataCategory: #DIMENSION
@Analytics.dataExtraction.enabled: true
@ObjectModel.representativeKey: 'CustomerId"
define view entity ZI_CustomerDimension
  as select from zcustomer
{
  key customer_id   as CustomerId,

      @Semantics.text: true
      customer_name as CustomerName,

      country       as Country,
      region        as Region,
      city          as City,
      customer_type as CustomerType,

      @Semantics.businessDate.createdAt: true
      created_on    as CreatedOn
}

Vue Fact (Chiffres d’affaires)

@AbapCatalog.viewEnhancementCategory: [#NONE]
@Analytics.dataCategory: #FACT
@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.changeDataCapture.automatic: true
define view entity ZI_SalesFactExtract
  as select from zsales_data as sd
  association [1..1] to ZI_CustomerDimension as _Customer
    on $projection.CustomerId = _Customer.CustomerId
  association [1..1] to ZI_ProductDimension as _Product
    on $projection.ProductId = _Product.ProductId
  association [1..1] to ZI_TimeDimension as _Time
    on $projection.SalesDate = _Time.CalendarDate
{
  key sd.sales_id     as SalesId,

      sd.customer_id  as CustomerId,
      sd.product_id   as ProductId,
      sd.sales_date   as SalesDate,

      @Semantics.amount.currencyCode: 'Currency"
      @DefaultAggregation: #SUM
      sd.revenue      as Revenue,

      @DefaultAggregation: #SUM
      sd.quantity     as Quantity,

      @DefaultAggregation: #SUM
      sd.discount     as Discount,

      sd.currency     as Currency,

      @Semantics.systemDateTime.lastChangedAt: true
      sd.changed_at   as ChangedAt,

      _Customer,
      _Product,
      _Time
}

Vue Cube (Agregation)

@AbapCatalog.viewEnhancementCategory: [#NONE]
@Analytics.dataCategory: #CUBE
@Analytics.dataExtraction.enabled: true
define view entity ZI_SalesAnalyticsCube
  as select from ZI_SalesFactExtract as Fact
  association [1..1] to ZI_CustomerDimension as _Customer
    on $projection.CustomerId = _Customer.CustomerId
  association [1..1] to ZI_ProductDimension as _Product
    on $projection.ProductId = _Product.ProductId
{
  key Fact.SalesId,

      -- Dimensions
      @ObjectModel.foreignKey.association: '_Customer"
      Fact.CustomerId,

      @ObjectModel.foreignKey.association: '_Product"
      Fact.ProductId,

      @AnalyticsDetails.query.axis: #ROWS
      Fact.SalesDate,

      -- Indicateurs
      @AnalyticsDetails.query.axis: #COLUMNS
      @Aggregation.default: #SUM
      Fact.Revenue,

      @Aggregation.default: #SUM
      Fact.Quantity,

      @Aggregation.default: #AVG
      Fact.Discount,

      Fact.Currency,

      -- Associations
      _Customer,
      _Product
}

Configurer l’extraction delta

L’extraction delta ne transfere que les donnees modifiees apres le chargement initial. Cela economise des ressources et reduit les latences.

Extraction delta basee sur horodatage

@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.byElement: {
  name: 'ChangedAt',
  maxDelayInSeconds: 3600,
  ignoreDeletions: false
}
define view entity ZI_MaterialWithDelta
  as select from zmaterial
{
  key material_id   as MaterialId,
      material_name as MaterialName,
      base_unit     as BaseUnit,
      material_type as MaterialType,

      @Semantics.systemDateTime.lastChangedAt: true
      changed_at    as ChangedAt,

      deletion_flag as DeletionFlag
}

Change Data Capture (CDC)

Pour une replication proche du temps reel, le CDC automatique peut etre active :

@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.changeDataCapture: {
  automatic: true,
  rowStatus: 'DELTA_STATUS"
}
define view entity ZI_OrderWithCDC
  as select from zorder
  association [0..*] to ZI_OrderItemWithCDC as _Items
    on $projection.OrderId = _Items.OrderId
{
  key order_id       as OrderId,
      customer_id    as CustomerId,
      order_date     as OrderDate,
      order_status   as OrderStatus,

      @Semantics.systemDateTime.lastChangedAt: true
      changed_at     as ChangedAt,

      -- Champ de statut CDC
      delta_status   as DeltaStatus,

      _Items
}

Valeurs du champ CDC

Valeur	Signification
`' '` (vide)	Inchange
`'U'`	Update
`'I'`	Insert
`'D'`	Delete

Scenario delta complexe

@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.changeDataCapture.automatic: true
define view entity ZI_InvoiceExtract
  as select from zinvoice as inv
  inner join zcustomer as cust
    on inv.customer_id = cust.customer_id
{
  key inv.invoice_id   as InvoiceId,
      inv.customer_id  as CustomerId,
      cust.customer_name as CustomerName,
      inv.invoice_date as InvoiceDate,
      inv.total_amount as TotalAmount,
      inv.currency     as Currency,
      inv.payment_status as PaymentStatus,

      -- Champs delta
      @Semantics.systemDateTime.lastChangedAt: true
      case
        when inv.changed_at > cust.changed_at
        then inv.changed_at
        else cust.changed_at
      end as ChangedAt,

      -- Indicateur de suppression calcule
      case
        when inv.deletion_flag = 'X' or cust.deletion_flag = 'X"
        then 'X"
        else '"
      end as DeletionIndicator
}

Autorisations pour l’extraction

L’extraction necessite des autorisations speciales tant dans le systeme ABAP que dans Datasphere.

Autorisations cote ABAP

CLASS zcl_datasphere_auth DEFINITION
  PUBLIC
  CREATE PUBLIC.

  PUBLIC SECTION.
    INTERFACES if_badi_interface.

    METHODS check_extraction_auth
      IMPORTING
        iv_view_name TYPE sxco_cds_object_name
      RETURNING
        VALUE(rv_authorized) TYPE abap_bool.

ENDCLASS.

CLASS zcl_datasphere_auth IMPLEMENTATION.

  METHOD check_extraction_auth.
    " Verification de l'autorisation d'extraction
    DATA: lv_auth_check TYPE abap_bool.

    " Verifier l'objet d'autorisation
    AUTHORITY-CHECK OBJECT 'Z_DS_EXTR"
      ID 'Z_VIEWNAME' FIELD iv_view_name
      ID 'ACTVT' FIELD '03'.  " Afficher

    IF sy-subrc = 0.
      rv_authorized = abap_true.
    ELSE.
      rv_authorized = abap_false.
    ENDIF.
  ENDMETHOD.

ENDCLASS.

Definition de l’objet d’autorisation

" Objet d'autorisation Z_DS_EXTR pour l'extraction Datasphere
" Champs :
" - Z_VIEWNAME : Nom de la vue CDS (CHAR 40)
" - ACTVT : Activite (01=Creer, 02=Modifier, 03=Afficher)

Vues CDS avec controle d’autorisation

@AccessControl.authorizationCheck: #CHECK
@Analytics.dataExtraction.enabled: true
define view entity ZI_ConfidentialSalesData
  as select from zsales_data
{
  key sales_id      as SalesId,
      customer_id   as CustomerId,
      region        as Region,
      revenue       as Revenue,
      profit_margin as ProfitMargin,
      currency      as Currency,
      changed_at    as ChangedAt
}

DCL associee (Data Control Language)

@EndUserText.label: 'Autorisation donnees ventes"
@MappingRole: true
define role ZI_ConfidentialSalesData {
  grant select on ZI_ConfidentialSalesData
  where ( Region ) = aspect pfcg_auth(
    Z_SD_REGN,
    Z_REGION,
    ACTVT = '03"
  );
}

Utilisateur technique pour l’extraction

Pour l’extraction automatique, un utilisateur technique est necessaire :

CLASS zcl_datasphere_extraction_user DEFINITION
  PUBLIC
  FINAL
  CREATE PUBLIC.

  PUBLIC SECTION.
    CLASS-METHODS get_technical_user
      RETURNING
        VALUE(rv_user) TYPE syuname.

    CLASS-METHODS validate_technical_user
      IMPORTING
        iv_user TYPE syuname
      RETURNING
        VALUE(rv_valid) TYPE abap_bool.

ENDCLASS.

CLASS zcl_datasphere_extraction_user IMPLEMENTATION.

  METHOD get_technical_user.
    " Utilisateur technique pour l'extraction Datasphere
    rv_user = 'DATASPHERE_EXTRACT'.
  ENDMETHOD.

  METHOD validate_technical_user.
    " Verification si l'utilisateur existe et est actif
    SELECT SINGLE bname FROM usr02
      WHERE bname = @iv_user
        AND uflag = 0  " Non verrouille
      INTO @DATA(lv_user).

    rv_valid = xsdbool( sy-subrc = 0 ).
  ENDMETHOD.

ENDCLASS.

Live Data Connection vs. Replication

SAP Datasphere offre deux modes d’integration fondamentaux qui couvrent differents cas d’utilisation.

Live Data Connection (Virtualisation)

Avec la Live Data Connection, les donnees ne sont pas copiees mais recuperees directement du systeme source a l’execution.

Avantages :

Donnees toujours actuelles
Pas de consommation de stockage supplementaire
Pas de logique de replication necessaire

Inconvenients :

Latence plus elevee lors des requetes
Charge sur le systeme source
Possibilites de transformation limitees

Cas d’utilisation :

Donnees de base avec faible volume de modifications
Requetes ad-hoc
Prototypage et exploration

Replication (Materialisation)

Avec la replication, les donnees sont physiquement copiees vers Datasphere et y sont persistees.

Avantages :

Haute performance des requetes
Pas de charge sur le systeme source
Transformations complexes possibles
Historisation possible

Inconvenients :

Donnees pas en temps reel
Consommation de stockage supplementaire
Logique de replication necessaire

Cas d’utilisation :

Donnees de mouvements avec volume eleve
Requetes analytiques complexes
Reporting avec cycles de rafraichissement definis

Matrice de decision

Critere	Live Data	Replication
Actualite des donnees	Temps reel	Differe (minutes a heures)
Performance	Moyenne	Elevee
Charge systeme source	Elevee	Faible
Volume de donnees	Petit a moyen	Illimite
Transformations	Limitees	Etendues

Scenario hybride

-- Donnees de base : Live Connection (toujours actuelles)
@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.preferredMode: #VIRTUAL
define view entity ZI_CustomerMasterLive
  as select from zcustomer
{
  key customer_id   as CustomerId,
      customer_name as CustomerName,
      customer_type as CustomerType,
      credit_limit  as CreditLimit,
      currency      as Currency
}

-- Donnees de mouvements : Replication (haute performance)
@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.preferredMode: #REPLICATED
@Analytics.dataExtraction.delta.changeDataCapture.automatic: true
define view entity ZI_SalesHistoryReplicated
  as select from zsales_history
{
  key sales_id     as SalesId,
      customer_id  as CustomerId,
      sales_date   as SalesDate,
      amount       as Amount,
      currency     as Currency,
      changed_at   as ChangedAt
}

Configuration dans Datasphere

Configuration de la connexion

Dans le Data Builder de Datasphere, la connexion au systeme ABAP est configuree :

Connection Type : SAP ABAP
System ID : Identification du systeme (par ex. S4P)
Host : Nom d’hote du systeme ABAP
Client : Mandant
User/Password : Utilisateur technique

Configuration Remote Table

Parametres Remote Table :
- Source Object : ZI_SalesFactExtract
- Load Type : Initial + Delta
- Delta Frequency : Toutes les 15 minutes
- Memory : Large
- Partitioning : Par SalesDate (mensuel)

Definition Data Flow

Pour des transformations complexes, des Data Flows peuvent etre definis :

Source : ZI_SalesFactExtract (Remote Table)
  → Filtre : SalesDate >= '20240101"
  → Agregation : SUM(Revenue) GROUP BY CustomerId, Month
  → Enrichissement : LEFT JOIN CustomerMaster
  → Cible : Sales_Monthly_Summary (Local Table)

Optimisation des performances

Conception de vues CDS pour l’extraction

-- Optimise pour l'extraction
@AbapCatalog.sqlViewAppendage: {
  forSourceViewName: 'ZSALESOPT"
}
@Analytics.dataExtraction.enabled: true
@Analytics.dataExtraction.delta.byElement.name: 'ChangedAt"
define view entity ZI_SalesOptimized
  as select from zsales
{
  key sales_id    as SalesId,

      -- Cle de partitionnement en premier
      @Analytics.dataExtraction.partitionBy: true
      sales_date  as SalesDate,

      -- Champs filtrables indexes
      @Analytics.dataExtraction.filter: #RECOMMENDED
      region      as Region,

      customer_id as CustomerId,
      amount      as Amount,
      currency    as Currency,

      @Semantics.systemDateTime.lastChangedAt: true
      changed_at  as ChangedAt
}

Tailles de lot et parallelisation

CLASS zcl_extraction_config DEFINITION
  PUBLIC
  FINAL
  CREATE PUBLIC.

  PUBLIC SECTION.
    CONSTANTS:
      c_batch_size        TYPE i VALUE 50000,
      c_max_parallel_jobs TYPE i VALUE 4,
      c_delta_interval    TYPE i VALUE 900.  " 15 minutes

    CLASS-METHODS get_extraction_params
      IMPORTING
        iv_view_name TYPE sxco_cds_object_name
      RETURNING
        VALUE(rs_params) TYPE zcl_extraction_config=>ts_params.

    TYPES:
      BEGIN OF ts_params,
        batch_size     TYPE i,
        parallel_jobs  TYPE i,
        delta_interval TYPE i,
        use_compression TYPE abap_bool,
      END OF ts_params.

ENDCLASS.

CLASS zcl_extraction_config IMPLEMENTATION.

  METHOD get_extraction_params.
    " Configuration specifique a la vue
    rs_params = VALUE #(
      batch_size     = c_batch_size
      parallel_jobs  = c_max_parallel_jobs
      delta_interval = c_delta_interval
      use_compression = abap_true
    ).

    " Ajustement pour les grandes vues
    IF iv_view_name CS 'HISTORY' OR iv_view_name CS 'ARCHIVE'.
      rs_params-batch_size = 100000.
      rs_params-parallel_jobs = 8.
    ENDIF.
  ENDMETHOD.

ENDCLASS.

Monitoring et depannage

Surveiller l’extraction

CLASS zcl_extraction_monitor DEFINITION
  PUBLIC
  FINAL
  CREATE PUBLIC.

  PUBLIC SECTION.
    METHODS get_extraction_status
      IMPORTING
        iv_view_name TYPE sxco_cds_object_name
      RETURNING
        VALUE(rs_status) TYPE zcl_extraction_monitor=>ts_status.

    TYPES:
      BEGIN OF ts_status,
        view_name      TYPE sxco_cds_object_name,
        last_run       TYPE timestampl,
        records_total  TYPE i,
        records_delta  TYPE i,
        runtime_ms     TYPE i,
        status         TYPE c LENGTH 10,
        error_message  TYPE string,
      END OF ts_status.

ENDCLASS.

CLASS zcl_extraction_monitor IMPLEMENTATION.

  METHOD get_extraction_status.
    " Exemple de requete de statut
    SELECT SINGLE *
      FROM zextraction_log
      WHERE view_name = @iv_view_name
      ORDER BY timestamp DESCENDING
      INTO @DATA(ls_log).

    IF sy-subrc = 0.
      rs_status = VALUE #(
        view_name      = ls_log-view_name
        last_run       = ls_log-timestamp
        records_total  = ls_log-records_total
        records_delta  = ls_log-records_delta
        runtime_ms     = ls_log-runtime_ms
        status         = ls_log-status
        error_message  = ls_log-error_message
      ).
    ENDIF.
  ENDMETHOD.

ENDCLASS.

Erreurs frequentes

Erreur	Cause	Solution
No extraction enabled	Annotation manquante	Ajouter `@Analytics.dataExtraction.enabled: true`
Delta field not found	Mauvais nom de champ	Verifier le nom du champ dans `byElement.name`
Authorization failed	Autorisation manquante	Autoriser l’utilisateur technique
Timeout during extraction	Volume de donnees trop important	Reduire la taille de lot

Bonnes pratiques

Modelisation en schema en etoile : Separer clairement les vues Fact et Dimension
Toujours activer le delta : Reduit la charge et le volume de transfert
Definir les autorisations de maniere granulaire : N’exposer que les donnees necessaires
Tester les performances : Mesurer les temps d’extraction avant Go-Live
Configurer le monitoring : Detecter rapidement les extractions defaillantes
Maintenir la documentation : Documenter les vues et leur utilisation

Resume

Aspect	Recommandation
Donnees de base	Live Data Connection
Donnees de mouvements	Replication avec delta
Autorisations	Combiner DCL + PFCG
Performance	Partitionnement + Index
Monitoring	Verification reguliere du statut

L’integration d’ABAP Cloud avec SAP Datasphere ouvre de puissantes possibilites analytiques. Avec les bonnes annotations, strategies delta et concepts d’autorisation, un approvisionnement en donnees robuste et performant peut etre realise.

Sujets connexes

Fondamentaux CDS Views - Bases des Core Data Services
Custom Analytical Queries - Requetes analytiques dans ABAP Cloud
CDS Annotations - Toutes les annotations CDS en un coup d’oeil
Nouveautes ABAP SQL - Fonctionnalites SQL actuelles dans ABAP