Qlik Replicate SQL view replication: float data column inconsistency

[Michael_Litz]

Data inconsistencies can be observed when replicating an SQL view from SQL to Hadoop. The column in the underlying table is of the FLOAT data type.

Example:

The source has 5 decimals, while the target has more than 10.

SQL Source value of the float datatype = 11.14008 T
Hadoop Target value of the float datatype = 11.1400800000000001

 

Environment

• Qlik Replicate 2023.11 and later releases

Resolution

Qlik Replicate does not add precision explicitly. The behavior is caused by the nature of the FLOAT datatype; no configuration steps or workarounds exist to change the datatype's behavior.

Cause

FLOAT is, by definition, not a precise data type. 

For additional reading on FLOAT, see, for example:

• Why Floating-Point Numbers May Lose Precision (learn.microsoft.com)
• Precision and accuracy in floating-point calculations 

Internal Investigation ID(s)

RECOB-8880