This APAR is done in v97 only as a temporary internal workaround
for a specific customer. This APAR is not a problem and can be
ignored by any users.
As per IEEE 754 standard for floating point datatypes, floating
point numbers are a binary approximation of an actual number.
As such, when these datatypes are stored, there is potential for
a loss of accuracy. For some numbers, in particular
(but not limited to) those numbers that have a large amount of
scale digits, this gives a number that may lose the
precision and accuracy that is needed in some applications.
In DB2, the floating point datatypes are DOUBLE, FLOAT, and REAL
(see documentation about DB2 datatypes). If having precision and
accuracy is required for the application, then it is not
recommended to use floating point types, and the DECIMAL or
DECFLOAT datatype should be used instead.
This APAR provides an alternate approach from float to decimal
for some numbers that are suffering from the
binary approximation accuracy concerns.
Since this is a behaviour change in how floating point numbers
are treated by DB2, this change is not provided for general use.
If this APAR is required, then please contact DB2 support team,
who can provide guidance on how to enable this alternate mode
for converting floating point types to DECIMAL. |
