Throughout our discussion of sorting, we have assumed that the elements to be sorted are simply integers. Frequently, we need to sort large structures by a certain key. For instance, we might have payroll records, with each record consisting of a name, address, phone number, financial information such as salary, and tax information. We might want to sort this information by one particular field, such as the name. For all of our algorithms, the fundamental operation is the swap, but here swapping two structures can be a very expensive operation, because the structures are potentially large. If this is the case, a practical solution is to have the input array contain pointers to the structures. We sort by comparing the keys the pointers point to, swapping pointers when necessary. This means that all the data movement is essentially the same as if we were sorting integers. This is known as indirect sorting; we can use this technique for most of the data structures we have described. This justifies our assumption that complex structures can be handled without tremendous loss efficiency.