Fan out

Imagine a spreadsheet in which every formula mentions four other cells.  Something like =A1+B2+C3+D4, or =SUM(X5:Y6).

Then any cell that contains a formula has four precedents.  And each of those precedents likely contains a formula that has a further four precedents.  So the first cell has ...

•4 direct precedents

•16 precedents one level closer to the spreadsheet's inputs

•64 precedents two levels away

•256 precedents three levels away

...and so on.  Eventually, the fan-out will stop, because the chain of precedents reaches the spreadsheet's inputs (unless the spreadsheet is circular).  In real spreadsheets, this fan-out of precedents can be explosively large, and can be an obstacle to gaining any useful overview of how particular numbers are derived.

This fan out is evident in the physical layout of the example given of Discrepancy analysis. Taming it is a priority, as it undermines the usefulness of the discrepancy analysis output.  OAK seeks to do this by making available as options in the discrepancy analysis.

•a limit to the number of levels of precedence that the discrepancy analyzer examines

•pruning of expressions: the discrepancy analyzer confines its attention to the branches of conditional expressions that are actually active, given the current spreadsheet inputs, and does not provide any analysis of non-active branches; this is the same mechanism as is offered by the OAK4 | Formula | Prune command

•the ability to specify a limit beyond which large ranges will not be followed.

Fan-out is also a consideration in the reconstructions of calculations that OAK can prepare, and the Formula | Reconstruct command offers similar facilities with the same purpose in mind.