This idea attempts to address the potential to easily misconfigure Omniscope to show the wrong data through unsuitable combinations of sum/mean/formula and aggregation/grouping.

These ideas would not affect pre-configured IOK files; they would only change default settings, and hide inactive unusual choices.

1. In the function picker, only show functions which have the same resulting data type (in the same "logical value domain") as the field they are applied to.
   
   Visible choices (depending on field data type):
   
   • Sum
   • Mean
   • Median
   • Mode
   • Minimum
   • Maximum
   • Singleton value
   • First value
   • Last value
   • (Formula result)
   
   
   Power-user hidden choices:
   
   • Record count
   • Empty count
   • Non-empty count
   • Unique value count
   • Unique values list
   • Range
   • Standard deviation
   
   
   You would still be able to get at the power-user hidden choices easily but would not see these by default.
   
   This is necessary for (3), to avoid aggregation/grouping from changing the data structure and preventing formulas from working. Formulas are defined to operate over underlying raw data, and if your aggregation changes a text field into a numeric through the use of "Empty count", the formula will most likely fail.
   
   This change would occur in grouping and aggregation, but perhaps also elsewhere for consistency and to remove choices which are confusing - e.g. a value labelled "Sales volume" but showing "Range" is incorrect.
   
   
2. In the Table view, we provide a "Formula result" option for each field alongside sum/mean/etc., just like with aggregation, instead of the "calculate formulas over grouped results" precedence setting.
   
   
3. In aggregation and grouping, we make "(Formula result)" the default for any formula field. This would include the Table view grouping, in-view aggregation, and the Aggregation operation in the Data menu and in DataManager.
   
   Note: While the bar view aggregates automatically, we couldn't allow choosing "Formula result", since Omniscope couldn't tell whether to sum/mean/whatever in any fields referenced by formulas. Instead, you would use the new "Formula" measure in 2.7, which allows you to write your own field-independent formula where you are required to use SUBSET_SUM/MEAN as appropriate.
   
   

This would mean that, if you have formulas configured in the underlying data, and you are rolling up a view to show the total sales per region, your formulas would re-execute by default.

However, due to the power and complexity of formulas, we would like to get feedback from our power users as to whether this would be a net-positive change. Would it reduce the incidence of incorrect configuration, for common models?