Skip to content

Latest commit

 

History

History

abap

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 

README.md

ABAP Cloud / RAP implementation

The same business object as the CAP service in ../cap, implemented on-stack with the ABAP RESTful Application Programming Model.

Both implement the identical rule set. The message numbers are the same (ZPR_MSG 001..012), the approval matrix is the same, the risk weights are the same - so an incident ticket quotes the same code no matter which implementation raised it, and the two test suites guard each other against one-sided changes.

What is in here

src/
├── dbtables/   7 database tables (DDL source)
├── cds/        interface views, projection views, metadata extensions,
│               abstract entities for the action parameters
├── behavior/   behaviour definitions (base + projection)
├── classes/    rule classes with ABAP Unit tests, behaviour pools,
│               virtual element calculator
└── services/   service definitions

The parts worth reading first

File Why
classes/zcl_pr_approval_policy.clas.abap The approval matrix, the state machine and the submit validations. No SELECT, no RAP artefact - pure rules.
classes/zcl_pr_approval_policy.clas.testclasses.abap 35 ABAP Unit tests for those rules. Runs without a database.
classes/zcl_pr_risk_scoring.clas.abap The weighted supplier risk score, plus 16 unit tests next to it.
behavior/zi_pr_requisition.bdef.asbdef Managed, draft enabled BO: determinations, validations, actions, side effects, feature control.
classes/zbp_i_pr_requisition.clas.locals_imp.abap The behaviour pool. Reads, calls the rule classes, writes - no business logic of its own.
cds/zi_pr_supplier.ddls.asddls The clean core piece: supplier master data is joined from the released I_Supplier, never copied.

Design decisions you can ask me about

Rules live in plain classes, not in the behaviour pool. A handler method needs a RAP runtime, a transaction and a database to be executed. A class method needs none of that, which is why 51 of the tests in this package run in under a second and are part of every transport check.

The header determination sits on the item. Total value, risk class and required approval level belong to the header, but only an item change can make them wrong. A determination on the header would never fire when only an item changed - so the determination that maintains them sits where the trigger is and updates the parent.

Feature control is a convenience, never a security boundary. Every rule that greys out a button is evaluated a second time inside the action. The action is the one that counts.

Supplier master data is extended, not replicated. ZPR_SUP_RISK holds only the five attributes S/4HANA has no field for. Name, country and the purchasing block come from I_Supplier on every read, so a block set in the backend is effective here immediately.

Getting it into a system

The CDS views, behaviour definitions, classes, metadata extensions and service definitions are stored in their ADT source form, which is exactly what abapGit serialises - .abapgit.xml is set up for a PREFIX folder logic on /src/.

One thing to know before importing: the subfolders under src/ group the objects by kind so the package reads well on GitHub. abapGit's PREFIX folder logic maps folders to sub-packages, not to object kinds. So either create the matching sub-packages (ZPR_PROCUREMENT_CDS, ZPR_PROCUREMENT_CLASSES, ...) or flatten src/ into a single folder before pulling. Readability won here because this package is read far more often than it is imported.

Two things are not in the repository and are created by ADT rather than by hand:

  1. Draft tables ZPR_REQ_HDR_D, ZPR_REQ_ITM_D, ZPR_APPROVAL_D - generated from the persistent tables by the quick fix Create draft table.
  2. Service bindings - created on the service definitions ZUI_PR_Requisition and ZUI_PR_Supplier as OData V4 / UI, then published.

You also need the message class ZPR_MSG with numbers 001 to 012 and the authorisation objects ZPR_APPR (field ZPR_LEVEL) and ZPR_ADMIN.

The customizing tables ZPR_RATING and ZPR_CNTRYRSK expect the same content as the CAP fixtures in ../cap/db/data - that is what keeps the two stacks comparable.

Honest scope

This package is written for review, not for a one-click import. It has been written against the ABAP Cloud syntax of the 2023+ releases and follows the Clean ABAP style guide, but it has not been activated in a system as part of this repository - there is no ABAP system in the build pipeline. The CAP implementation in ../cap is the part that is executed and tested on every commit; the value of this package is that it shows the same problem solved on-stack, and that the two solutions can be read side by side.

What the CAP side has and this one does not

The customer facing part of the demo was built on the side-by-side stack, so these exist only there:

  • the approver inbox (MyApprovalTasks),
  • the audit trail per requisition (RequisitionEvents),
  • automatic adjustment of the approval path when a supplier's risk changes (ADR 0008),
  • purchase order creation and supplier sync through the S/4HANA OData APIs (ADR 0009),
  • notifications to approvers and requesters (ADR 0010),
  • the approval matrix maintained in Excel instead of in code (ADR 0011).

On-stack, the last point would not be an integration at all: RAP writes the purchase order with EML against the released purchase order business object in the same system. That is exactly the trade-off ADR 0001 is about.

The rules those features rest on are framework free on the CAP side, so a port would move reassessApprovalPath() into ZCL_PR_APPROVAL_POLICY unchanged and rewrite only the plumbing around it.