ABAP Keyword Documentation → ABAP - Reference → Processing Internal Data → Internal Tables → Processing Statements for Internal Tables → LOOP AT itab → LOOP AT itab - GROUP BY
LOOP AT itab - group_key
Other versions:
7.31 | 7.40 | 7.54
Syntax
... key | ( key1 = dobj1 key2 = dobj2 ...
[gs = GROUP SIZE] [gi = GROUP INDEX] ) ...
Extras
1. ... gs = GROUP SIZE
2. ... gi = GROUP INDEX
Effect
Group key expression for constructing the group key in the grouping of internal tables. The group key is the result of the group key expression, whose data type is defined as follows:
-
If
key
is specified, the group key is a single data object with any data type. -
If ( key1 = dobj1 key2 = dobj2 ...
) is specified, the group key is a structure with the components
key1
,key2
, ... The components can have any names and any data types. The structure type is defined here as a bound data type of the structure.
key
or key1
, key2
, ... are
general expression
positions in which the current row can be used in accordance with the output behavior defined in
result
. The data types of
key
or key1
, key2
, ... must be complete and statically identifiable.
For each row of the internal table itab
read in the grouping phase, the value
of the group key is calculated from the data objects, expressions, or calls specified here using the group key expression.
If a group key binding is defined in the output behavior
group_result
of the group loop, the data object or field symbol specified here can be used to access the value of the group key of every group in the group loop.
Notes
-
If the grouping is to be useful, the value of the group key calculated for every read row of
itab
must have a suitable relationship to the row in question. The following are some special cases:
- If the group key is a value that depends on the row, there is only one group to which all read rows belong.
- If the group key is different for every read row, there are the same number of groups as rows and each group contains only one row.
- The ability to directly construct any type of structure makes it easy to create multi-component group keys without needing to declare a matching structure first or to concatenate the key components in, for example, a string.
-
The syntax shown here applies both to groupings with
LOOP AT itab ... GROUP BY and to groupings with
FOR GROUPS ... OF
.
Addition 1
... gs = GROUP SIZE
Addition 2
... gi = GROUP INDEX
Effect
Declares additional components for a structured group key. These components are not part of the group key and are used as a repository of group-specific information instead:
-
The component
gs
is filled with the number of group members for each group. -
The component
gi
is filled with a group index for each group. The group index of the group key constructed first is 1 and is raised by 1 for each new group key.
Before the additional components can be declared, a group key binding must be defined in the output
behavior group_result
.
The additional components can be accessed in the group loop using the data object or field symbol specified here.
The components gs
and gi
can be given any name.
The language elements GROUP SIZE
or GROUP INDEX
on the right side give the components their special semantics.
Notes
-
If the groups are not sorted using
ASCENDING
orDESCENDING
, the groups in the group loop are processed in the order of the group index. If the groups are sorted in different ways, the original position is noted in the group index. -
The additional components can still be evaluated even if the addition
WITHOUT MEMBERS
is used and the group members cannot be accessed. -
Implicitly, no further values are created that match the results of
aggregate expressions
when
GROUP BY
is used in the Open SQL statementSELECT
, except in the componentgs
with the right sideGROUP SIZE
. Aggregations like this are easy to program (see the example), particularly when applying constructor operators such asREDUCE
to the group members. The Open SQL additionHAVING
for restricting the loops of groups is a similar case. Here, for example, the filter operatorFILTER
can be used.
Examples
See the examples for grouping internal tables.