Solution: to transform the data source by splitting into many rows and using 1 column.
Inside computeExpression computeTYPE1 nodes:
Inside Type_TYPE1 Computed Field:
this is text, which is the field name
Inside Value_TYPE1 Computed Field:
this is numeric, which is the field value
Do the same for computeExpression Type 2 and Type 3. Then, combine all the data using append node
Once we have all the values spread across rows, use sliceDataset transformation to drop the original Type 1, Type 2, Type 3 fields.