SAP RAP Draft Architecture & Concurrency Lock Management

1. The 3-Layer Storage Architecture of Draft

When an entity is declared with draft; in its Behavior Definition (BDEF), the SAP RAP engine coordinates three storage tiers:

2. Draft Lifecycle Operations

The RAP framework provides built-in draft actions that govern transition between active and draft states:

• Edit: Reads the active record, copies it into the draft table, sets draftinprocessbyuser, and takes an exclusive enqueue lock.
• Activate: Executes standard validations and determinations in the CHECK_BEFORE_SAVE phase. If valid, writes draft state to active persistence and purges the draft row.
• Discard: Drops the draft row completely, reverts any pending changes, and immediately releases the exclusive enqueue lock.
• Resume: Re-associates a user's browser session with their existing draft after session timeout.
• Prepare: Triggers validations on draft data before explicit activation to surface UI field errors.

3. Optimistic Concurrency & Total ETag

To prevent the "Lost Update" problem where User B overwrites User A's uncommitted changes, RAP mandates two levels of ETag definitions:

managed implementation in class zbp_r_order_root unique;
strict ( 2 );
with draft;

define behavior for ZR_ORDER_ROOT alias Order
persistent table zsd_order_a
draft table zsd_order_d
lock master total etag LastChangedAt  // Checks concurrent edits on active record
etag master LocalLastChangedAt        // Checks concurrent edits on draft buffer
{
  draft action Edit;
  draft action Activate optimized;
  draft action Discard;
  draft action Resume;
  draft determine action Prepare;

  field ( readonly, numbering : managed ) OrderUUID;
  field ( readonly ) OrderID, CreatedAt, CreatedBy, LastChangedAt;

  create;
  update;
  delete;
}

4. Enterprise Housekeeping: The Draft Accumulation Problem

In high-volume systems, users frequently abandon browser sessions or close laptop lids without clicking "Save" or "Cancel". This creates two major problems:

1. Orphan Locks: The active record remains locked with the message "Object is locked by user X in another draft session".
2. Draft Table Bloat: Thousands of abandoned draft rows accumulate in Z*_D tables, slowing down indexing and CDS joins.

5. Production EML Cleanup Implementation

The following is the contract-compliant method for discarding stale draft records programmatically:

" =====================================================================
" Method: Purge Stale Drafts via EML
" Language Version: ABAP Cloud (Language Version 5)
" =====================================================================
METHOD purge_expired_drafts.
  DATA(lv_cutoff) = utclong_add( val = utclong_current( ) hour = -72 ).

  " 1. Query expired drafts via administrative metadata
  SELECT OrderUUID
    FROM zsd_order_d
    WHERE draftlastchangedatetime < @lv_cutoff
    INTO TABLE @DATA(lt_expired).

  CHECK lt_expired IS NOT INITIAL.

  " 2. Construct Discard EML payload
  DATA: lt_discard TYPE TABLE FOR ACTION IMPORT zr_order_root~Discard.
  lt_discard = VALUE #(
    FOR ls IN lt_expired (
      %tky-%is_draft  = if_abap_behv=>mk-on
      %tky-OrderUUID  = ls-OrderUUID
    )
  ).

  " 3. Execute framework discard action
  MODIFY ENTITIES OF zr_order_root
    ENTITY Order
    EXECUTE Discard FROM lt_discard
    FAILED DATA(lt_failed)
    REPORTED DATA(lt_reported).

  " 4. Release locks
  COMMIT ENTITIES.
ENDMETHOD.