... GROUP BY { {col1 col2 ...} | (column_syntax) } ... .
The addition GROUP BY combines groups of rows that have the same content in their specified columns col1 col2 ... in the resulting set into a single row.
The use of GROUP BY has the prerequisite that SELECT only individual columns, not all the columns, are specified using *. If GROUP BY is used, all columns that are specified directly after SELECT and not specified as an argument of an aggregate function must be listed there. Conversely, if GROUP BY is used, all the columns listed after SELECT that are not specified after GROUP BY must be specified as an argument of an aggregate function. The aggregate functions define how the content of these columns is determined in the combined row from the contents of all the rows of a group.
After GROUP BY, the same column identifiers must be specified as after SELECT. The specification can either be specified statically as a list col1 col2 ... or dynamically as a brackted data object column_syntax that - at execution of the statement - contains the syntax of the staticspecification or is set to initial value. For column_syntax, the same applies as for the dynamic column specification after SELECT.
If the content of column_syntax initial, either all the rows or no rows at all are grouped together. The columns after SELECT must then be listed either solely as arguments of aggregate functions or solely directly. If not, prior to Release 6.10 you could trigger a runtime error. As of Release 6.10, this would trigger an exception CX_SY_OPEN_SQL_DB that can be handled.
The columns listed after GROUP BY must not be of type STRING or RAWSTRING, and if GROUP BY is used,
pool or cluster tables cannot be accessed.
With the use of GROUP BY, the statement SELECT avoids the SAP buffering.