[Feature Request]: Handling columns with different meta types than String in Row Normalizer

[vdwals]

What would you like to happen?

In the context of #2312 and #3633, an issue has been identified regarding the normalization of rows with different meta types. It's unclear if this affects only non-String meta types or any mix of types.

Currently, the value of any type is simply copied without casting or conversion into the result rows, but its field receives the meta type String. This causes some functionalities to fail:

• Sampling of rows after pipeline execution (#2312)
• Sorting rows, which failed in one instance when using a temporary file for caching.

Potential solutions include adapting the Row Normalizer Transformation or aligning the documentation. Proposed solutions are:

• Allowing users via the UI to specify format masks and other transformation-related details, enabling the Row Normalizer to transform non-String values.
• Restricting users from selecting non-String fields in the UI.
• Introducing a flag to cast/convert everything to String during the process.
• Adding a dropdown to select the output field type, with String as the default for backward compatibility.

Further discussion is encouraged to determine the optimal solution.

Issue Priority

Priority: 2

Issue Component

Component: Transforms

[vdwals]

This examples shows the error message appearing in the logs mentioned in #2312: feature-request-3636.zip

[vdwals]

.take-issue