Code-to-data: stop dragging the database into ABAP

For twenty years classic ABAP did the opposite of what made sense: it pulled raw data up to the application server and processed it there, row by row, inside a LOOP. With slow disks and dumb databases that had some logic. With HANA, it doesn’t. The code-to-data paradigm flips the rule: push the computation to where the data lives, not the other way round.

Data-to-code versus code-to-data

• Data-to-code (the old pattern): SELECT *, everything into an internal table, then ABAP sums, filters and groups in memory. The database only reads; the heavy lifting happens on the AS.
• Code-to-data (the HANA pattern): aggregation, filtering and joins are expressed in SQL and run by the database. Only the reduced result reaches the AS.

The difference isn’t cosmetic. Moving a million rows across the database interface to discard 99% in an IF is exactly the waste HANA was designed to eliminate.

flowchart TB
 subgraph O["Data-to-code (old)"]
 A1[("HANA tables")] -->|"1,000,000 rows"| B1["SELECT *"]
 B1 --> C1["LOOP + IF + sum on the AS"]
 C1 --> D1["Result"]
 end
 subgraph N["Code-to-data (HANA)"]
 A2[("HANA tables")] -->|"aggregate and filter in the database"| B2["SUM, WHERE, JOIN"]
 B2 -->|"250 rows"| C2["Result on the AS"]
 end

The antipattern you’ll find in every legacy system

" NO: fetch everything and sum in memory
DATA lv_total TYPE p DECIMALS 2.
SELECT * FROM vbap INTO TABLE @DATA(lt_items).
LOOP AT lt_items INTO DATA(ls_item).
  IF ls_item-matnr = lv_matnr.
    lv_total = lv_total + ls_item-netwr.
  ENDIF.
ENDLOOP.

That code reads the whole item table, copies it entirely to the AS and throws almost all of it away. The code-to-data version lets HANA do the work:

" YES: the database aggregates and returns only the total
SELECT SUM( netwr ) AS total
  FROM vbap
  WHERE matnr = @lv_matnr
  INTO @DATA(lv_total).

One row travels the wire instead of a million. And HANA, columnar and in-memory, sums that column at a speed no ABAP LOOP will ever match.

It’s not just SUM

Code-to-data covers much more than aggregation. It’s pushing CASE expressions, SQL functions (CONCAT, SUBSTRING, ROUND), currency conversion with CURRENCY_CONVERSION, common table expressions with WITH. Everything you used to do in ABAP with loops and function calls now has an equivalent that runs inside HANA.

Mental rule: if you’re about to write a LOOP to compute something the database already knows how to compute, you’ve pulled the data too far.

Next post we go deeper: what code pushdown actually is, how to tell whether your SELECT exploits it, and where the optimizer stops helping you.