Wildcards in GROUP BY clause in OpenSQL? - abap

Question

I have a select similar to the one below:
SELECT DISTINCT
SCARR~CARRID,
SCARR~CARRNAME,
MIN( SPFLI~DISTANCE ) AS MIN_DISTANCE
FROM SCARR JOIN SPFLI ON SPFLI~CARRid = SCARR~CARRid
GROUP BY
SCARR~CARRID,
SCARR~CARRNAME
INTO TABLE #DATA(result).
In the real case, these are other tables and I have many more fields in both SELECT and GROUP BY.
Can I simplify the GROUP BY to not have to write again all the fields that are in the SELECT clause from the table SCARR?
I know other options are to use MIN for all the fields of table SCARR except its ID, or not GROUP BY and instead remove duplicates after the select, but I was trying to do something like GROUP BY scarr~*

Answer 1

No, the OpenSQL syntax doesn't support this.

Answer 2

Here is one possible workaround using dynamic approach:
DATA: lcl_struc TYPE REF TO cl_abap_structdescr,
lt_grouped TYPE TABLE OF sflight.
lcl_struc ?= cl_abap_typedescr=>describe_by_name( 'SFLIGHT' ).
DATA(group_by) = REDUCE string( INIT line TYPE char1024 FOR <field> IN lcl_struc->get_components( ) NEXT line = COND #( WHEN line <> space THEN line && `, ` && `SFLIGHT~` && <field>-name ELSE line && `SFLIGHT~` && <field>-name ) ).
SELECT (group_by)
FROM scarr LEFT OUTER JOIN spfli
ON spfli~carrid = scarr~carrid
LEFT OUTER JOIN sflight
ON sflight~carrid = spfli~carrid
AND sflight~connid = spfli~connid
LEFT OUTER JOIN sairport
ON sairport~id = spfli~airpfrom
WHERE scarr~carrid = 'AA'
GROUP BY (group_by)
INTO CORRESPONDING FIELDS OF TABLE #lt_grouped.
Pay attention to several limitations here:
In the above sample I group only by SFLIGHT fields, because when specifying wildcarded fields together with single (SELECT spfli~*, sflight~carrid) the inline result table will have not single fields but sub-structures, so it's more sophisticated for processing or requires explicit declaration. Keep this in mind if you need group by joined tables.
As it stated in help chapter provided by Florian, dynamic group clause requires all-or-nothing:
The columns after SELECT must then be specified either solely as arguments of aggregate functions or only directly. If not, this would raise a handleable exception CX_SY_OPEN_SQL_DB. Invalid syntax raises a handleable exception from the class CX_SY_DYNAMIC_OSQL_ERROR.
So it may be useless in your particular case if you need one-field aggregation, but just in case I put it here.

Answer 3

No is not possible in plain OpenSQL.
You can create a new CDS view for aggregate SPFLI, the view return for each CARRID the minimum distance:
define view Z_CDS_SPFLI_MIN as select from SPFLI
{
carrid,
min( distance ) as min_distance
}
group by carrid
And then you can modify your query like this, without use group-by:
select SCARR~*, Z_CDS_SPFLI_MIN~min_distance
from SCARR
inner join Z_CDS_SPFLI_MIN
on SCARR~carrid = Z_CDS_SPFLI_MIN~carrid.