Export .. to shared buffer ...

Question

I try to use this command but I don't understand something in the example of the Help. (F1 on shared buffer). (the bold)
TABLES INDX.
TYPES: BEGIN OF ITAB3_TYPE,
          CONT(4),
       END OF ITAB3_TYPE.
DATA: INDXKEY LIKE INDX-SRTFD VALUE 'KEYVALUE',
      F1(4), F2 TYPE P,
      ITAB3 TYPE STANDARD TABLE OF ITAB3_TYPE WITH
                 NON-UNIQUE DEFAULT KEY INITIAL SIZE 2,
      WA_INDX TYPE INDX.
Fill data fields before CLUSTR
before the actual export
INDX-AEDAT = SY-DATUM.
INDX-USERA = SY-UNAME. Export data.
EXPORT F1    FROM F1
       F2    FROM F2
       ITAB3 FROM ITAB3
       TO SHARED BUFFER INDX(ST) FROM WA_INDX ID INDXKEY.
Frédéric
(SAP 4.6C)

Answer 1

Because when everybody works on the same table, you cannot guarantee data consistency. Furthermore, the "free components" of INDX are just an example. You might want to define your own fields with your own specific information. Here comes the full description how such at table must be created. Note the second to the last entry. You are totally free to add components as you like and you are not bound to SAP's INDX-example!
<b>INDX-type structure:</b>
The first field must be a key field named MANDT of type CLNT for the client, if you want to store the data objects client-specifically. For a cross-client storage, this component does not apply.
The second field must be a key field named RELID of type CHAR and length 2. It stores the area ar specification.
The third field must be a key field of type CHAR named SRTFD with a maximum length of 55 characters. It stores the identifier specified in id.
The fourth field must be a key field named SRTF2 of type INT4. It contains the row numbers of a stored data cluster that can extend over several rows and is filled automatically by the system.
Then any number of components with freely selectable names and types may follow. They are provided with values by the specification of FROM wa. Addition TO wa of the IMPORT statement exports these fields again.
The last two components must be named CLUSTR and CLUSTD and be of types INT2 and LRAW of any length. In CLUSTR, the current length of field CLUSTD of each row is stored, while CLUSTD contains the actual data cluster.