ABAP CDS recursive association using hierarchy

Former Member · ‎01-13-2020

Introduction

In this article, I will show a practical example of self-association in CDS. You will learn how to define a simple hierarchy in your CDS and how to consume it in your ABAP code.

The problem

Recently I have faced a quite interesting problem. There is a concept of something like ‘frame contract’. It means that out of this document many others of the same type (or not necessairely) can be created. Let’s take a look at VBFA table (Sales Document Flow). I am interested only in subsequent Sales Orders, therefore I have filtered out only documents of category ‘C’:

What we can see is a ‘chain’ of documents, or even to be more specific – a tree. Let’s take a look at document 24. More friendly and eye-catching representation of the relationship between those documents would be:

Root parent of the above tree would be document 24 – our frame contract. Below it you will see it’s ‘children’ – subsequent documents. Now imagine that you would like to extract all documents related to your frame contract. Lot’s of twisted ABAP! Maybe there is an easier way of solving this riddle?

Gentle solutions

Yes! Fortunately, there is – we can avoid looping, reading, assigning and dumping.

CDS

We can define a CDS view which would have an association with itself. Let’s take a look:

define view Z_VBFA

  as select from vbfa

  association [0..1] to Z_VBFA as _Recursion on $projection.PrecedingDocument = _Recursion.SubsequentDocument

                                             and $projection.PrecedingItem = _Recursion.SubsequentItem

{

  key vbelv   as PrecedingDocument,

  key posnv   as PrecedingItem,

      vbeln   as SubsequentDocument,

      posnn   as SubsequentItem,

      vbtyp_n as SubsequentDocCategory,

      _Recursion

}

where

  vbtyp_n = 'C'

I have defined a view Z_VBFA based on VBFA and added an association to Z_VBFA. As you can see preceding documents will be treated as parents and subsequent ones as children. The cardinality of the above association is [0..1] as there cannot be any multiple parents – one contract can only have one preceding document, however it can have multiple children. In order to allow consumption of such CDS annotation needs to be added:

@Hierarchy.parentChild: [ { name: 'Recursion', recurseBy: '_Recursion' } ]

It defines parent-child hierarchy and points at our self-association. There is also a second way of defining a hierarchy:

@Hierarchy.parentChild: { name: 'Recursion',

                          recurse : { parent: ['PrecedingDocument','PrecedingItem'], child ['SubsequentDocument','SubsequentItem'] }

}

I believe that the first one is much nicer. Remember that in that approach you should not use association.

Hierarchy

You can also define a hierarchy object Z_VBFA_HIERARCHY which would look like this:

define hierarchy Z_VBFA_HIERARCHY 

  as parent child hierarchy (

    source Z_VBFA

    child to parent association _Recursion

    siblings order by PrecedingDocument ascending, PrecedingItem ascending

    orphans root

  )

{

    key PrecedingDocument,

    key PrecedingItem,

    SubsequentDocument,

    SubsequentItem,

    $node.parent_id as ParentNode,

    $node.node_id as ChildNode,

    $node.hierarchy_is_orphan as HierarchyIsOrphan,

    $node.hierarchy_level as HierarchyLevel,

    $node.hierarchy_rank as HierarchyRank,

    $node.hierarchy_parent_rank as HierarchyParentRank,

    $node.hierarchy_tree_size as HierarchyTreeSize  

}

Source – Z_VBFA – previously defined view with self-association

Child to parent association – points at recursion association in the source

Siblings order – order of sibling nodes

Orphans root – orphans will be treated as roots. In our case contract 24 would be such node.

Multiple parents – this syntax is not used here as multiple parents are not allowed in my case

I have also listed out all $node fields. These are hierarchy-specific fields describing a hierarchy. We will take a look at them later. After running the hierarchy you will see a standard table view:

Gentle consumption

Now let’s take advantage of the work we have done here and let’s consume our objects in ABAP. There are few useful HIERARCHY functions – you can read more about them here:

https://help.sap.com/viewer/4fe29514fd584807ac9f2a04f6754767/2.0.01/en-US/fc59f81a5c494f399cc2ff70b9...

We will Focus on three of them:

HIERARCHY – generates a hierarchy

HIERARCHY_DESCENDANTS – Returns all descendants of a set of start nodes in a hierarchy.

HIERARCHY_ANCESTORS – Returns all ancestors of a set of start nodes in a hierarchy.

Let’s implement it in ABAP.

HIERARCHY

A hierarchy from CDS view with defined hierarchy will be generated.

SELECT parent_id AS parent,

         node_id AS node,

         hierarchy_level AS level,

         hierarchy_rank AS rank,

         hierarchy_tree_size AS tree_size

    FROM HIERARCHY( SOURCE z_vbfa

                    CHILD TO PARENT ASSOCIATION _recursion

                    START WHERE precedingdocument = '0000000024' AND precedingitem = '000000'

                    SIBLINGS ORDER BY precedingdocument ASCENDING, precedingitem ASCENDING )

    INTO TABLE @DATA(lt_hierarchy).

Selected fields are hierarchy nodes that are always available and describe the structure. You need to select them explicitly and add an alias. Let’s take a look at the SOURCE:

CHILD TO PARENT ASSOCIATION – points at recursive association in the view

START WHERE – points at a starting node in the hierarchy

Result:

INDEX	PARENT	NODE	LEVEL	RANK	TREE_SIZE
1	10,0000000024;6,000000	10,0000000025;6,000000	1	1	11
2	10,0000000025;6,000000	10,0000000067;6,000000	2	2	1
3	10,0000000025;6,000000	10,0000000061;6,000000	2	3	1
4	10,0000000025;6,000000	10,0000000060;6,000000	2	4	1
5	10,0000000025;6,000000	10,0000000063;6,000000	2	5	1
6	10,0000000025;6,000000	10,0000000064;6,000000	2	6	1
7	10,0000000025;6,000000	10,0000000065;6,000000	2	7	1
8	10,0000000025;6,000000	10,0000000066;6,000000	2	8	1
9	10,0000000025;6,000000	10,0000000068;6,000000	2	9	3
10	10,0000000068;6,000000	10,0000000069;6,000000	3	10	1
11	10,0000000068;6,000000	10,0000000070;6,000000	3	11	1

We get all relationships between nodes, starting from contract 24. What’s great all children are listed out in the node column, so our goal has been achieved. Let’s dig deeper into it however and take a look at hierarchy specific fields which may help us with traversing a hierarchy. Below pictures explain two of them:

HIERARCHY_DESCENDANTS

In this example however we are going to consume our CDS hierarchy directly.

  SELECT *

    FROM HIERARCHY_DESCENDANTS( SOURCE z_vbfa_hierarchy

                                START WHERE precedingdocument = '0000000024' 

                                        AND precedingitem = '000000' )

    INTO TABLE @DATA(lt_descendants).

Result:

INDEX	PRECEDING DOCUMENT	PRECEDING ITEM	SUBSEQUENT DOCUMENT	SUBSEQUENT ITEM	PARENT NODE	CHILD NODE	HIERARCHY ISORPHAN	HIERARCHY LEVEL	HIERARCHY RANK	HIERARCHY PARENTRANK	HIERARCHY TREESIZE
1	0000000024	0	0000000025	0	10,0000000024;6,000000	10,0000000025;6,000000	1	1	2	0	11
2	0000000025	0	0000000067	0	10,0000000025;6,000000	10,0000000067;6,000000	1	2	3	2	1
3	0000000025	0	0000000061	0	10,0000000025;6,000000	10,0000000061;6,000000	1	2	4	2	1
4	0000000025	0	0000000060	0	10,0000000025;6,000000	10,0000000060;6,000000	1	2	5	2	1
5	0000000025	0	0000000063	0	10,0000000025;6,000000	10,0000000063;6,000000	1	2	6	2	1
6	0000000025	0	0000000064	0	10,0000000025;6,000000	10,0000000064;6,000000	1	2	7	2	1
7	0000000025	0	0000000065	0	10,0000000025;6,000000	10,0000000065;6,000000	1	2	8	2	1
8	0000000025	0	0000000066	0	10,0000000025;6,000000	10,0000000066;6,000000	1	2	9	2	1
9	0000000025	0	0000000068	0	10,0000000025;6,000000	10,0000000068;6,000000	1	2	10	2	3
10	0000000068	0	0000000069	0	10,0000000068;6,000000	10,0000000069;6,000000	1	3	11	10	1
11	0000000068	0	0000000070	0	10,0000000068;6,000000	10,0000000070;6,000000	1	3	12	10	1

I used * sign in select, therefore, all fields from defined earlier CDS hierarchy were selected. Notice that in this case Z_VBFA_HIERARCHY is consumed, not Z_VBFA.

HIERARCHY_ANCESTORS

This function will return the ancestors of a given node.

SELECT *

    FROM HIERARCHY_ANCESTORS( SOURCE z_vbfa_hierarchy

                              START WHERE precedingdocument = '0000000024' 

                                      AND precedingitem = '000000' )

    INTO TABLE @DATA(lt_ancestors).

Result:

INDEX	PRECEDING DOCUMENT	PRECEDING ITEM	SUBSEQUENT DOCUMENT	SUBSEQUENT ITEM	PARENTNODE	CHILDNODE	HIERARCHY ISORPHAN	HIERARCHY LEVEL	HIERARCHY RANK	HIERARCHY PARENTRANK	HIERARCHY TREESIZE
1	0000000024	0	0000000025	0	10,0000000024;6,000000	10,0000000025;6,000000	1	1	2	0	11

I have added orphans root annotation therefore one relationship in the hierarchy is displayed.

Alternative syntax

You can combine HIERARCHY and HIERARCHY_DESCENDANTS/HIERARCHY_ANCESTORS:

SELECT

    FROM HIERARCHY_DESCENDANTS( SOURCE HIERARCHY( SOURCE z_vbfa

                                                    CHILD TO PARENT ASSOCIATION _recursion

                                                    START WHERE precedingdocument = '0000000024'

                                                            AND precedingitem = '000000'

                                                    SIBLINGS ORDER BY precedingdocument ASCENDING, precedingitem ASCENDING )

    START WHERE precedingdocument = '0000000024' AND precedingitem = '000000' )

    FIELDS    node_id,

              parent_id,

              hierarchy_rank,

              hierarchy_level,

              hierarchy_tree_size,

              hierarchy_distance

    INTO TABLE @DATA(lt_descendants_2).

It will dynamically return hierarchy and apply HIERARCHY_DESCENDANTS on the result. Notice that we are not using Z_VBFA_HIERARCHY here but Z_VBFA. The result will be exactly the same as in HIERARCHY_DESCENDANTS.

Summary

I hope you can see the advantage of such solution. With few annotations and one select, we can extract the whole tree. With additional parameters like hierarchy_rank etc. it is incredibly easy to traverse through the whole structure. You can even select all other business data and have them all with one simple select. When possible avoid writing complex code!