Aggregator Totals Mismatch

[rameshrr3]

Hi

Im pretty sure that this has been discussed before, but my searches arent showing any threads.

Im trying to aggregate a flat file column that has 8000+ records. This column is defined as FLOAT. (Business expects data with variable number of digits after decimal point).
The source data is a delimited ASCII file.
The data sample that I have could well be handled by DECIMAL(15,2), but due to the said constraint , i need to define it as FLOAT data. None of the input records has data that contains more than 2 digits after the decimal point.(When seen using an OS Editor or Data browser)

However if I sum up this column using the aggregator stage, the output sum contains 4 digits after the decimal point. ( Again i define the 'sum' column as FLOAT). Now the user group comes back to me and says precision is important, and those 2 extra digits in the sum are a cause for concern. Is there anything that i could do to avoid/bypass this problem?

[ray.wurlod]

Search the web for information about why computers can not store FLOAT accurately, and the IEEE standards for representing a FLOAT. When a FLOAT is converted for display, there may be an arbitrary number of digits following the decimal point. It is not a DataStage issue; it's because you have chosen to use FLOAT as a data type. FLOAT can not be stored precisely.

[rafik2k]

so you requirement is to get only 2 digits after decimal point.

Either you can use round off function to get nearest values after decimal point
or
write a small basic rountine and call it in tranformer stage.
You need to manipulate column value using basic function.
Use string or field funtion to get exact 2 digits after decimal point.

Concatenate before decimal part and after decimal part.
then convert resultant string into float again.

[chulett]

Now the user group comes back to me and says precision is important, and those 2 extra digits in the sum are a cause for concern.

So, precision is important and yet they want a less precise number?

As noted, round or truncate the result, whichever is appropriate for your user group.

[ray.wurlod]

Use the Fix() function to specify the largest number of decimal places, then add 0 to that result to remove trailing non-significant zeroes.