We've made some changes to how we parse the SQL for object-centric transformation scripts when you publish them. We'll validate more things when you publish, instead of when you run the transformations, which makes the errors easier to identify and fix or mitigate. These changes apply to both your custom transformation scripts and our supplied transformation scripts for Celonis catalog processes. A phased rollout of the changes to teams starts from now.

As a result of these changes, when you publish your object-centric data model, you might see new validation errors that you weren't getting before. Here's what you might see, and how to fix or handle it:

• We'll now always add parentheses to expressions when we output the transformations. This might mean that an expression now fails to evaluate, or evaluates to a different result.

  • To fix this error, in your custom scripts, follow best practice and include parentheses in SQL expressions where the order can be ambiguous.

• We'll now validate that each column data type supplied from your source system matches, or can be assigned to, the required data type for the attribute in our underlying database. The data types we use are Boolean, long, float, timestamp and string. You might see a new issue if a column in your source system data has a data type that our Celonis catalog transformations don't expect.

  • To fix these errors, use the suggestions in Troubleshooting data extraction and pre-processing to account for the unexpected column data types.

  • To handle these errors, activate the Skip missing data option for the data connection, as described in Skipping missing data for objects and events. We'll cast the data types to match the expected data types. Note that it's best to fix the errors as this option might introduce other unexpected issues. It's possible that the transformation might fail to run even with the Skip missing data option enabled if a data type can't be cast to the one required. If that's the case, you'll need to fix the data type.