Window: analytics functions

The “window” recipe allows you to perform analytics functions on any dataset in Dataiku Agent Management, whether it’s a SQL dataset or not. This is the equivalent of a SQL “over” statement. The recipe offers visual tools to setup the windows and aliases. The “window” recipe can have pre-filters and post-filters. The filters documentation is available here.

See also

For more information, see also the following articles in the Knowledge Base:

• Concept | Window recipe

• Tutorial | Window recipe

Engines

Depending on the input dataset types, Dataiku Agent Management will adjust the engine it uses to execute the recipe, and choose between Hive, Impala, SparkSQL, plain SQL, and internal Dataiku Agent Management. The available engines can be seen and selected by clicking on the cog below the “Run” button.

Note

The Dataiku Agent Management engine has different default window behavior than when using a SQL engine. When using the Dataiku Agent Management engine, the window will default to the whole frame if no window is specified. As a result, you can see different window behavior when switching between the Dataiku Agent Management engine and a SQL engine. In order to see the same result when using the Dataiku Agent Management Engine as you would with the SQL engine, you can enable the Window Frame option with both the “Limit preceding rows” and “Limit following rows” options unchecked.

Notes

• Since Dataiku DSS v4.1, null values are ordered in a specific way, take a look at Null values handling