Python Analysis transform - Value was either too large or too small for UInt64

#1

I am running into an issue when using the Python Analysis transform, whenever the source data contains negative numbers I get the error “Value was either too large or too small for UInt64”, dispite the data type in the input not being a UInt.

I have set up a minimal cube to demonstrate the issue:
image

The manual query output is shown as an Int32, and contains negative numbers. If I preview the data of this transform I see the numbers all fine:
image

The python analysis transform contains a single statement that returns the input:
image

If I try to preview the data for this transform I get the error indicating it is trying to convert the input to a UInt64:

image
image.png755×190 16.6 KB

Is there something I am not configuring correctly, or is this an issue with the way the transform works? (The server is running Python 3.6.6)

#2

This issue (#80408) is fixed and under testing for a future revision. What is the severity of this issue for you?

#3

Hi Ilya,

Thats great, currently I am working around around the issue when I know it will occur by adding a calculated value transform to offset the negative values to be positive before they get to the python analysis, but obviously this is not ideal as I have to do this in many places and makes it more difficult for others to understand.

Do you know which release this is planned for?

#4

Another temporary solution would be to convert your values to a string before the python analysis. In the analysis, you could then go through that list, convert each value to float, and do further mathematical operations on those values.
// Declaring two empty lists for our values
floats = []
mathTest = []
// $value$ is a list of strings inputted into the analysis
for val in $value$:
// Convert each string to float type and append to list ‘floats’
floats.append(float(val))
// Test to see that math operations will work on the new list
for i in floats:
mathTest.append(i-1)
return mathTest

#5

Hi Andy,

The fix for this issue is now available in our minor release (Dundas BI v6.0.2).

The release notes are published:
https://www.dundas.com/Support/learning/documentation/release-notes/dundas-bi-release-notes?v=6.0.2

Thank you for your patience