ABAP Keyword Documentation → ABAP Programming Guidelines → Robust ABAP → Assignments, Calculations, and Other Types of Data Access
Rounding Errors
Other versions: 7.31 | 7.40 | 7.54
Background
In the case of assignments between floating point numbers (types f,
decfloat16, decfloat34) and fixed point numbers
(types i and p), rounding errors usually occur
that produce an incorrect value. However, values of type p (and also
decfloat16,decfloat34) assigned to type f are not always displayed accurately.
Rule
Avoid unnecessary rounding errors
Avoid unnecessary or frequent conversions between floating point numbers and fixed numbers as this can cause rounding errors.
Details
In a program, the value of a number should always be stored for as long as possible in a data object with the numerical data type of the highest required level of accuracy. This applies especially to saving intermediate results of calculations.
The data input or output requirements (for example, formatting on the screen or in a spool list) cannot
change the way numbers are saved internally. If a number is to be formatted using a specific number
of decimal places, the actual value should not be converted to the corresponding packed number. Instead,
a suitable format should be configured in a character-like field using the relevant constructs. These are character string templates and previously the statement WRITE TO.
Bad example
The following source code shows a calculation where results can be assigned to a numerical field intended for output. The result 55.55 is rounded to 56.00.
percentage TYPE decfloat34,
value TYPE decfloat34.
percentage = '55.55'.
value = '100.0'.
output = percentage / 100.
output = value * output.
Good example
The following source code corrects the above example by separating the code into data objects intended for calculations and a character-like data object for the formatted output.
percentage TYPE decfloat34,
value TYPE decfloat34.
DATA output TYPE c LENGTH 40.
percentage = '55.55'.
value = '100.0'.
result = percentage / 100.
result = value * result.
WRITE result TO output DECIMALS 2 EXPONENT 0.