Default rules for schema detection from JSON
By inspecting the beginning of JSON document, schema detection automatically determines lookup expressions for
repeating objects, locating row objects relative to the root,
column values, locating columns relative to row objects.
This process begins by finding potential columns. A potential column is a simple-valued (i.e., non-array, non-object) member of some parent object.
Their parent objects are candidates for the repeating object. For the first column candidate that occurs at least twice with the same locator, its parent object will be chosen as the repeating object. If there is none that occurs twice, the parent with the highest number of columns is chosen.
The column locator expressions are made up of the expressions that address members of (and relative to) the repeating object.
