The optimisation was mistakenly assuming data always had the un-parsed structure, when in fact serialize and test operate on the parsed data. This change also introduces a new optimisation of excluding never and excluding parsed values that do not have a serialize method
The optimisation was mistakenly assuming data always had the un-parsed structure, when in fact serialize and test operate on the parsed data. This change also introduces a new optimisation of excluding never and excluding parsed values that do not have a serialize method