Behavior of Reverse Association (Meshes)

[SuhaSaha]

Hello SDNers,

This week i have been exploring the feature "Meshes" and have found a behaviour which i cannot understand.

Below is the sample code i'm using (ABAP 740 SP12):

REPORT ydemo_mesh_inverse_assoc.
CLASS demo DEFINITION.
  PUBLIC SECTION.
    CLASS-METHODS:
      main,
      class_constructor.
  PRIVATE SECTION.
    TYPES:
      BEGIN OF line1,
        col1 TYPE i,
      END OF line1,
      t_itab1 TYPE SORTED TABLE OF line1
                   WITH UNIQUE KEY col1,
      BEGIN OF line2,
        col1 TYPE i,
        col2 TYPE i,
      END OF line2,
      t_itab2 TYPE SORTED TABLE OF line2
                   WITH UNIQUE KEY col1 col2,
      BEGIN OF line3,
        col1 TYPE i,
        col2 TYPE i,
        col3 TYPE i,
      END OF line3,
      t_itab3 TYPE SORTED TABLE OF line3
                   WITH UNIQUE KEY col1 col2 col3,
      BEGIN OF MESH t_mesh,
        node1 TYPE t_itab1
             ASSOCIATION to_node2 TO node2 ON col1 = col1,
        node2 TYPE t_itab2
             ASSOCIATION to_node3 TO node3 ON col1 = col1
                                          AND col2 = col2,
        node3 TYPE t_itab3,
      END OF MESH t_mesh.
    CLASS-DATA
      mesh TYPE t_mesh.
ENDCLASS.
CLASS demo IMPLEMENTATION.
  METHOD main.
    DATA(idx) = 1.
    cl_demo_input=>request( CHANGING field = idx ).
    DATA(out) = cl_demo_output=>new(
      )->begin_section( 'node1'
      )->write( mesh-node1
      )->next_section( 'node2'
      )->write( mesh-node2
      )->next_section( 'node3'
      )->write( mesh-node3 ).
    IF line_exists( mesh-node3[ col1 = idx ] ).
      out->next_section( 'node3\^to_node3~node2'
        )->write( VALUE t_itab2(
           FOR <node2> IN
             mesh-node3\^to_node3~node2[ mesh-node3[ col1 = idx ] ]
             ( <node2> ) ) ).
      out->next_section( 'node3\^to_node3~node2\^to_node2~node1'
        )->write( VALUE t_itab1(
           FOR <node1> IN
             mesh-node3\^to_node3~node2[ mesh-node3[ col1 = idx ]
                                       ]\^to_node2~node1[ ]
             ( <node1> ) ) ).
    ELSE.
      out->write( `Enter a valid index for node3 ...` ).
    ENDIF.
    out->display( ).
  ENDMETHOD.
  METHOD class_constructor.
    mesh-node1 = VALUE t_itab1(
      ( col1 = 1 ) ( col1 = 2 ) ( col1 = 3 ) ).
    mesh-node2 = VALUE t_itab2(
      col1 = 1  ( col2 = 11 ) ( col2 = 12 )
      col1 = 2  ( col2 = 21 ) ( col2 = 22 )
      col1 = 3  ( col2 = 31 ) ( col2 = 32 ) ).
    mesh-node3 = VALUE t_itab3(
        col1 = 1  col2 = 11 ( col3 = 111 ) ( col3 = 112 )
        col1 = 1  col2 = 12 ( col3 = 121 ) ( col3 = 122 )
        col1 = 2  col2 = 21 ( col3 = 211 ) ( col3 = 212 )
        col1 = 2  col2 = 22 ( col3 = 221 ) ( col3 = 222 )
        col1 = 3  col2 = 31 ( col3 = 311 ) ( col3 = 312 )
        col1 = 3  col2 = 32 ( col3 = 321 ) ( col3 = 322 ) ).
  ENDMETHOD.
ENDCLASS.
START-OF-SELECTION.
  demo=>main( ).

My question is specifically for the reverse association -

FOR <node2> IN
mesh-node3\^to_node3~node2[ mesh-node3[ col1 = idx ] ]

Why does the path result return only 1 row of node2?

I referred to the ABAP Documentation and found this -

If the last path node contains multiple rows that fit the description, the first row found is read.

Refer - ABAP Keyword Documentation. But i am not sure, if this is the correct explanation of this behaviour.

BR,

Suhas

[Sandra_Rossi]

Hello Suhas,

I think this is because of mesh-node3[ col1 = idx ] which by definition ("A mesh path expression is a special type of table expression and can be used in the same way as a table expression"; so I think it's really like a table expression) returns only one line. If idx is 1, then it will match one of the four lines with col1 = 1 (depends on binary search SAP internal algorithm I guess). Consequently, it takes only one line from node2.

Regards,

Sandra

[SuhaSaha]

Hi Sandra,

Nice to see you back

"A mesh path expression is a special type of table expression and can be used in the same way as a table expression"

Refer: ABAP Keyword Documentation.

This statement explains the behaviour. I wrote a forward association to test this out and it is aligns with what you have mentioned.

out->next_section( 'Forward assoc: node2\to_node3'
)->write( VALUE t_itab3(
   FOR <node3> IN
  mesh-node2\to_node3[ mesh-node2[ col1 = idx ] ]
  ( <node3> ) ) ).

mesh-node2[ col1 = idx ] returns a row of node2, which is then used to analyse the follow-on node3.

Thanks for clarifying.

BR,

Suhas