Instance updates attribute of different calling instance

[former_member224107]

The issue is happening within the context of Badi development for SAP BPC.

In the method of the Badi where we can develop our code, we use an exising

standard SAP class (which is CL_UJO_WRITE_BACK) to save data to the

database. We create an instance of that class, and call a method passing some

parameters. Within that method an attribute of the class is updated. Somehow, it

also updates the attribute of the calling instance.

I will explain this again schematically :

Class X , Instance Xa

Instance Xa holds attribute y

Instance Xa calls method which calls a Badi.

In the Badi method, Instance Xb is created of

the same class. Attribute y is updated in method of

Instance Xb. Somehow, it also updates attribute y

of Instance Xa.

Is this the normal behaviour of Abap objects ?

[former_member186338]

Sorry, but please explain what are you doing in general!

What is the badi type? What do you want to achieve?

[gregor_dieckmann]

A static attribute is shared in all instances, an instance attribute is (as the name indicates) if bound to the instance.

[former_member224107]

It is a protected instance attribute (ds_wb_param) of this class. This value of this attribute is changed

upon calling the method IF_UJO_WRITE_BACK~WRITE_BACK, by passing this as a parameter.

I am using this class to perform a write back to the BPC model without checking workstatus and

without updating the audit - for which I use that method and pass the parameters. However , by

doing this, it also changes the attribute ds_wb_param in the SAP standard program of the calling instance.

[former_member186338]

Once again, what do you want to achieve? What is the business logic behind? What is the badi type? Custom logic or?

[former_member224107]

Hi Vadim,

this is a complicated business logic which is already live for a period of time. I cannot explain this on this forum,

and it is not really relevant for this issue. The problem arises now because the question has come that the

saved data need to be audited, and because of this issue the data is not saved in the audit log. It is an ABAP

technical issue, that's why I posted it in the ABAP forum.

[former_member224107]

I am using the write back badi. Because I am calling the write back method of that class, an attribute of the

standard program is updated (and as a consequence the audit log is not updated, I was able to see this

in debugging). I assume that a Badi should never update class attributes of the standard

SAP program. I now wonder whether this could be a bug in the kernel.

[matt]

Try what I've suggested in my answer, then get back to us.

[horst_keller · Accepted solution]

Works as documented.

https://help.sap.com/doc/abapdocu_752_index_htm/7.52/en-US/abapprivate.htm?file=abapprivate.htm

The smallest encapsulation unit of ABAP Objects is the class, not the object.

[hubert_heitzer]

Yes, unfortunately this is a normal behaviour of ABAP Objects.

A private attribute of an instance of class X can be changed by every other Instance of class X.

[SuhaSaha]

A private attribute of an instance of class X can be changed by every other Instance of class X.

It is not true, unless:

• the attribute is static, or
• the instances of the class X are the same, i.e., class X is a singleton!

Otherwise it would violate basic OO principles.

[hubert_heitzer]

Hi Suhas,

I am also disappointed but try this:

REPORT  z_demo.

CLASS lcl_oo_violation_demo DEFINITION.
  PUBLIC SECTION.
    METHODS:
      constructor,
      violate_basic_oo_principles IMPORTING ir_oo_violation TYPE REF TO lcl_oo_violation_demo,
      get_state                   RETURNING VALUE(rv_state) TYPE        string
      .
  PRIVATE SECTION.
    DATA:
      mv_state TYPE string
    .
ENDCLASS.

CLASS lcl_oo_violation_demo IMPLEMENTATION.
  METHOD constructor.
    mv_state = 'initial'.
  ENDMETHOD.

  METHOD violate_basic_oo_principles.
    ir_oo_violation->mv_state = 'changed by someone'.
  ENDMETHOD.

  METHOD get_state.
    rv_state = mv_state.
  ENDMETHOD.
ENDCLASS.


START-OF-SELECTION.

  DATA:
    lr_source TYPE REF TO lcl_oo_violation_demo,
    lr_target TYPE REF TO lcl_oo_violation_demo
    .

  CREATE OBJECT lr_source.
  CREATE OBJECT lr_target.

  WRITE: / lr_target->get_state( ).
  lr_source->violate_basic_oo_principles( lr_target ).
  WRITE: / lr_target->get_state( ).

[SuhaSaha]

In the method violate_basic_oo_principles you are changing the attribute MV_STATE of the instance LR_TARGET which is of type lcl_oo_violation_demo. I don't see how this violates the OO principle.

I can do the same in Java too.

public class oo_violation {
	private String state;
	
	public String getState() {
		return this.state;
	}
	public oo_violation() {
		this.state = "I am initial";
	}
	public void changeState( oo_violation anotherInstance ) {
		anotherInstance.state = "I have been changed";
	}
}

class Demo {
	public static void main(String args[]) {
		oo_violation source = new oo_violation();
		System.out.println("Source State: " + source.getState());
		
		oo_violation target = new oo_violation();
		source.changeState(target); // Change the state of the Target Instance
		System.out.println("Target State: " + target.getState());
		
	}
}

[hubert_heitzer]

You claimed, that it violates basic OO principles, not me!

"
A private attribute of an instance of class X can be changed by every other Instance of class X.

It is not true, unless:

• the attribute is static, or
• the instances of the class X are the same, i.e., class X is a singleton!

Otherwise it would violate basic OO principles.
"

Attribute mv_state is neither static nor are lr_source and lr_target referencing the same instance.

Btw.: I am pretty sure of C++ not tolerating this (at least the GNU compiler)

[SuhaSaha]

A private attribute of an instance of class X can be changed by every other Instance of class X.

Ahh, now i see what you meant by the statement. Sorry for the misunderstanding.

CLASS lcl_x DEFINITION CREATE PUBLIC.
  PUBLIC SECTION.
    DATA state TYPE string READ-ONLY.
    METHODS:
      constructor,
      main.
  PROTECTED SECTION.
  PRIVATE SECTION.
    DATA: output TYPE REF TO if_demo_output.
    METHODS: call_badi.
ENDCLASS.

CLASS lcl_x IMPLEMENTATION.
  METHOD constructor.
    me->output  = cl_demo_output=>new( ).
    me->state   = |Initial State|.
    me->output->write( |State of the current: "{ me->state }"|  ).
  ENDMETHOD.
  METHOD call_badi.
    DATA(xb)  = NEW lcl_x( ).
    xb->state = |Changed in the BAdI|.
    me->output->write( |State of the new instance: "{ xb->state }"|  ).
    
    ASSERT me->state <> xb->state. " <<< This is what i meant by the violation >>> 
  ENDMETHOD.
  METHOD main.
    me->call_badi( ).
    me->output->write( |Current State: "{ me->state }"|  ).
    me->output->display( ).
  ENDMETHOD.
ENDCLASS.
START-OF-SELECTION.
  NEW lcl_x( )->main( ).

When i said,

It is not true, unless:

• the attribute is static, or
• the instances of the class X are the same, i.e., class X is a singleton!

This is what i meant -

ASSERT me->state <> xb->state.

I hope this clarifies the misunderstanding.

Anyway i found another interesting thread -> https://archive.sap.com/discussions/thread/3482419

[former_member186338]

"this is a complicated business logic which is already live for a period of time. I cannot explain this on this forum" - top secret! 🙂

I have written write back badi many times for different systems and never have the issues like you have.

But I never set ls_wb_param-update_audit = abap_false. and disable work status check in my badi - to my mind it's a bad practice in general.

All data written to the cube (by input form, script logic, badi etc) has to be reflected in audit log. Same for work status lock.

[matt]

This would be easier to answer if you posted the actual code. Use the "code" button in the editor.

Run in debugger and look at the instance numbers of Xa and Xb. Are they the same?