The Dataiku Frontrunner Awards have launched to recognize your achievements! SUBMIT YOUR ENTRY

SQL like recipe creation should respect recipe initiation mode setting

In Project settings on the Engines & connections tab, there is a setting to indicate "How to create default code for new Hive, SQL, SparkSQL and Impala recipes". The setting name is "SQL-like recipes init mode".

However, at least when creating SQL recipes, this setting only partially used.

We certainly find it valuable to make our SQL datasets relocatable as described here:

Since we typically work with SQL Script recipes for greater control and input and output dataset references are fully resolved when inserted in the code (i.e., hover to right of input/output dataset and click "insert" on recipe window), we are having to type in project variables manually (i.e., ${projectKey}). It's certainly doable but seems completely unnecessary since the setting suggests it should be done for you. Moreover, including the project key variable is no longer the path of least resistance so use of these relocatable references is no longer consistent across our user base.

In fact, it was done for us through version 5. In version 6, this functionality apparently disappeared.

So for example if "SQL-like recipes init mode" is set to "Table reference with variables", when inserting a reference to a new SQL (or SQL like) recipe (by clicking the "insert" button to the right of the dataset name) the table reference should include the variable as the prefix (${projectKey}_TABLENAME) rather than the actual resolved project key prefix (MYPROJ_TABLENAME).