Production Architectural Blueprint

SAP CDS Virtual Elements & SADL Calculation Architecture Guide

Domain: RAP Query Layer & Procedural Calculation Exits Standard: IF_SADL_EXIT_CALC_ELEMENT_READ Clean Core Tier: Tier-1 On-Stack

1. The Limits of SQL in Core Data Services

Core Data Services (CDS) push business logic down to the SAP HANA database engine with extraordinary efficiency. However, certain enterprise business calculations cannot be expressed purely in declarative SQL:

• Procedural algorithms (complex pricing condition determination, custom tax matrix lookups).
• Calling released ABAP APIs during read execution (e.g. factory calendar conversions, currency conversions with custom fallback rules).
• Dynamic Fiori Elements criticality evaluation (computing colored pill badges based on multi-entity heuristics).

CDS Virtual Elements solve this architectural challenge by allowing developers to define non-persisted, transient fields in a CDS Projection View that are calculated on-the-fly by an ABAP class via the SADL (Service Adaptation and Description Layer) framework.

2. Architectural Decision Matrix: Virtual Elements vs Unmanaged Query vs Pure SQL

Evaluation Vector	Pure SQL Expression (CDS)	CDS Virtual Element (SADL Exit)	RAP Custom Query Provider
Execution Location	SAP HANA Database Engine	ABAP Application Server (Post-SQL)	ABAP Application Server (Full Query)
Procedural ABAP Logic	No (Declarative SQL only)	Yes (Full ABAP Cloud language)	Yes (Custom paging/sorting/filtering)
Pagination ($top / $skip)	Native HANA database pushdown	Native HANA database pushdown	Manually handled in ABAP code
Sorting & Filtering ($filter)	Directly filterable in SQL	Requires custom SADL filter transform	Parsed manually via EML request
Standard Recommendation	Default whenever possible	Recommended for calculated UI fields	Reserved for remote APIs / non-HANA data

3. The SADL Calculation Lifecycle

When an OData client or Fiori Elements app requests a page of 50 records from a projection view with a virtual element:

1. Discovery: SADL inspects the CDS metadata and identifies @ObjectModel.virtualElement: true.
2. Pre-Fetch Info: SADL invokes if_sadl_exit_calc_element_read~get_calculation_info on the calculation class. The method informs SADL which underlying database fields are needed for calculation.
3. SQL Execution: SADL generates the optimized SQL SELECT ... FROM view LIMIT 50 OFFSET 0 including the requested dependency fields.
4. Vectorized Calculation: SADL passes the returned 50-row internal table into if_sadl_exit_calc_element_read~calculate. The ABAP method loops across the table in-memory and sets the virtual element values.
5. Response Serialization: SADL returns the enriched JSON payload to the client.

4. Implementing the Clean Core SADL Calculation Class

CLASS zcl_calc_sales_criticality DEFINITION
  PUBLIC
  FINAL
  CREATE PUBLIC.

  PUBLIC SECTION.
    INTERFACES if_sadl_exit_calc_element_read.
ENDCLASS.

CLASS zcl_calc_sales_criticality IMPLEMENTATION.

  METHOD if_sadl_exit_calc_element_read~get_calculation_info.
    " Crucial: Prevent N+1 select queries by requesting prerequisite fields
    INSERT 'DELIVERYSTATUS' INTO TABLE et_requested_orig_elements.
  ENDMETHOD.

  METHOD if_sadl_exit_calc_element_read~calculate.
    DATA(lt_original_data) = it_original_data.

    LOOP AT lt_original_data ASSIGNING FIELD-SYMBOL(<row>).
      ASSIGN COMPONENT 'DELIVERYSTATUS' OF STRUCTURE <row> TO FIELD-SYMBOL(<status>).
      ASSIGN COMPONENT 'DELIVERYCRITICALITY' OF STRUCTURE <row> TO FIELD-SYMBOL(<crit>).
      
      IF <status> IS ASSIGNED AND <crit> IS ASSIGNED.
        CASE <status>.
          WHEN 'C'. <crit> = 3. " Green (Delivered)
          WHEN 'B'. <crit> = 2. " Yellow (Partially Delivered)
          WHEN 'A'. <crit> = 1. " Red (Blocked / Delinquent)
          WHEN OTHERS. <crit> = 0. " Grey
        ENDCASE.
      ENDIF.
    ENDLOOP.

    ct_calculated_data = CORRESPONDING #( lt_original_data ).
  ENDMETHOD.

ENDCLASS.

5. Critical Anti-Patterns to Avoid

• Anti-Pattern 1: Missing Calculation Info: Forgetting to register fields in get_calculation_info causes calculate to receive empty or unassigned fields when the Fiori UI does not explicitly include that column in the table view.
• Anti-Pattern 2: Unvectorized Processing: Treating calculate as a single-record processor rather than a bulk internal table processor severely degrades throughput.
• Anti-Pattern 3: Statefulness in Calculation Class: The calculation class should remain stateless. Do not store intermediate transaction state in instance attributes across requests.