Data structures

Whereas control structures organize algorithms, data structures organize information. In particular, data structures specify types of data, and thus which operations can be performed on them, while eliminating the need for a programmer to keep track of memory addresses. Simple data structures include integers, real numbers, Booleans (true/false), and characters or character strings. Compound data structures are formed by combining one or more data types.

The most important compound data structures are the array, a homogeneous collection of data, and the record, a heterogeneous collection. An array may represent a vector of numbers, a list of strings, or a collection of vectors (an array of arrays, or mathematical matrix). A record might store employee information—name, title, and salary. An array of records, such as a table of employees, is a collection of elements, each of which is heterogeneous. Conversely, a record might contain a vector—i.e., an array.

Record components, or fields, are selected by name; for example, E.SALARY might represent the salary field of record E. An array element is selected by its position or index; A[10] is the element at position 10 in array A. A FOR loop (definite iteration) can thus run through an array with index limits (FIRST TO LAST in the following example) in order to sum its elements:

• FOR i ← FIRST TO LAST
• SUM ← SUM + A[i]

Arrays and records have fixed sizes. Structures that can grow are built with dynamic allocation, which provides new storage as required. These data structures have components, each containing data and references to further components (in machine terms, their addresses). Such self-referential structures have recursive definitions. A bintree (binary tree) for example, either is empty or contains a root component with data and left and right bintree “children.” Such bintrees implement tables of information efficiently. Subroutines to operate on them are naturally recursive; the following routine prints out all the elements of a bintree (each is the root of some subtree):

• PROCEDURE TRAVERSE(ROOT: BINTREE)
• IF NOT(EMPTY(ROOT))
• TRAVERSE(ROOT.LEFT)
• PRINT ROOT.DATA
• TRAVERSE(ROOT.RIGHT)
• ENDIF

Abstract data types (ADTs) are important for large-scale programming. They package data structures and operations on them, hiding internal details. For example, an ADT table provides insertion and lookup operations to users while keeping the underlying structure, whether an array, list, or binary tree, invisible. In object-oriented languages, classes are ADTs and objects are instances of them. The following object-oriented pseudocode example assumes that there is an ADT bintree and a “superclass” COMPARABLE, characterizing data for which there is a comparison operation (such as “<” for integers). It defines a new ADT, TABLE, that hides its data-representation and provides operations appropriate to tables. This class is polymorphic—defined in terms of an element-type parameter of the COMPARABLE class. Any instance of it must specify that type, here a class with employee data (the COMPARABLE declaration means that PERS_REC must provide a comparison operation to sort records). Implementation details are omitted.

• CLASS TABLE OF <COMPARABLE T>
• PRIVATE DATA: BINTREE OF <T>
• PUBLIC INSERT(ITEM: T)
• PUBLIC LOOKUP(ITEM: T) RETURNS BOOLEAN
• END

• CLASS PERS_REC: COMPARABLE
• PRIVATE NAME: STRING
• PRIVATE POSITION: {STAFF, SUPERVISOR, MANAGER}
• PRIVATE SALARY: REAL
• PUBLIC COMPARE (R: PERS_REC) RETURNS BOOLEAN
• END
• EMPLOYEES: TABLE <PERS_REC>

TABLE makes public only its own operations; thus, if it is modified to use an array or list rather than a bintree, programs that use it cannot detect the change. This information hiding is essential to managing complexity in large programs. It divides them into small parts, with “contracts” between the parts; here the TABLE class contracts to provide lookup and insertion operations, and its users contract to use only the operations so publicized.