Running SQL via SQLWorkbench versus via Tableau Prep

Question

I have developed some SQL that reads from a redshift table, does some manipulation (esp listagg some fields), and then writes to another redshift table.

When I run the SQL using SQLWorkbench it executes successfully. When I embed it in a Tableau Prep flow (as "Complex SQL") I get several of these errors: "System error: AqlProcessor evaluation failed: [Amazon][Support] (40550) Invalid character value for cast specification." Presumably these relate to my treatment of data types. What I don't is what is so difference in the environment that would cause different results like this? Is it because SQLWorkbench and Tableau Prep use different SQL interpreters? Or is my question too broad to even speculate without going through the actual code?

Answer 1

Best guess is that Tableau, which has knowledge of DDL, is add some CAST() operations to the SQL. SQLWorkbench is simpler and is pushing the SQL to Redshift as written. This is based on there being no explicit CASTs in your SQL but an error message that identifies a CAST().

Look at stl_querytext for these two queries and see if they are being given to Redshift differently by the two benches. I suspect this will give you some clues to go on.

If there are no differences in the SQL then the issue may be with user / connection differences and more info will likely be needed about the issue.