Glossary

This glossary includes terms and definitions from (1) The American National Dictionary for Information Processing Systems, copyright 1982 by the Computer and Business Equipment Manufacturers Association (CBEMA). Copies may be purchased from the American National Standards Institute, 1430 Broadway, New York, New York 10018. Definitions are identified by the symbol (A) after the definition. (2) The Information Technology Vocabulary, developed by Subcomittee 1, Joint Technical Committee 1, of the International Organization for Standardization and the International Electrotechnical Committee (ISO/IEC JTC2/SC1). Definitions of published segments of the vocabularies are identified by the symbol (I) after definition; definitions from draft international standards, draft proposals, and working papers in development by the ISO/IEC JTC2/SC1 vocabulary subcommittee are identified by symbol (T) after definition, to indicate that final agreement has not yet been reached among participating members.

activity value: The value of a basic or nonbasic variable. The activity of nonbasic variables is either the upper or lower bound or a fixed value if the variable has been set to that value.
active node: In mixed-integer program processing, a node that has been created but has not yet been evaluated. The current node is also active.
adjacent arc: An incoming or outgoing arc at a node.
adjacency matrix of A: A matrix in which the adjacency of rows is indicated (A A^T). A nonzero entry a_ij in the adjacency matrix indicates there is at least one column with a nonzero entry in both rows i and j.
adjustable cell: A spreadsheet cell containing a numeric value that can be changed to satisfy constraints and yield an optimal value of the objective function.
advanced basis: Any basis other than a basis of all logical variables. An advanced basis might be obtained using EKKCRSH or EKKBASI.
APAR: Authorized Program Analysis Report. A report of a problem caused by a suspected defect in a current unaltered release of a program.
application: For the Optimization Library, your entire mathematical programming problem and the work area useed in which to store and solve it.
arc: A column of the LP matrix for a network programming problem. There is a +1 in the row that corresponds to the tail of the arc and a -1 in the row that corresponds to the head.
arc flow: See flow.
argument: A parameter passed between a calling program and a SUBROUTINE subprogram, a FUNCTION subprogram, or a statement function.
array: An ordered set of data items identified by a single name.
artificial variable: Logical variables associated with equality rows.
barrier parameter: In the formulation of the interior-point primal barrier method, a sequence of subproblems of the following form are solved: 'minimize' %% F(bix ; µ) % = % <bic sup T> bix % - % µ sum from j=1 to n of ln x sub j subject to: <biA bix % = % bib> labove <mu % gt % 0> The scalar µ is known as the barrier parameter.
basic variables: The m logical or structural variables, in an m-row linear programming model whose values can be computed as a solution of a system of equations, where the matrix is the basic matrix, and the right-hand side is computed by fixing the remaining variables at one of their bounds.
basic feasible solution (BFS): A feasible solution with m basic variables and n - m nonbasic variables.
basis: The set of names or indices that identifies the basic variables. The optimal base of a given model is often kept to be restored later as an advanced starting solution of a modified model.
BFS: Basic feasible solution.
block: A small component matrix. A large matrix may be created by first creating one or more blocks, then adding them together to get the entire matrix.
block angular form: Also known as a Dantzig-Wolfe problem, a problem with a collection of disjoint blocks in the constraint matrix and a small set of coupling constraints. The coupling constraints may have nonzero coefficients in columns corresponding to two or more of the blocks.
bound: A limit on a particular variable (column) or on a row of the constraint matrix. The value of the variable or row must stay between the upper and lower bounds.
branch-and-bound method: The mathematical algorithm used in the Optimization Library to solve mixed-integer programming problems. It uses a technique of successive tightening of integer variables to integer values starting from the continuous optimal solution.
chaining: For 0-1 integer variables only, if one variable is forced to a bound, a number of variables may be forced into one bound or the other. These variables are said to be chained.
character type: The data type for representing strings of alphanumeric characters; in storage, one byte is used for each character.
Cholesky factorization: Also known as Cholesky decomposition, the unique representation of positive definite matrix A as the product A=L L^T, where L is a lower triangular matrix. Also an algorithm for computing the elements of the matrix L from the elements of matrix A.
clique: A set of 0-1 integer variables, such that setting any member in this set to one bound will fix all other members, while fixing it to the other bound may not fix any other members.
coefficient: The factor that multiplies a variable in a row.
coefficient matrix: The m-row by n-column matrix formed by all the coefficients that link the rows to the variables of a linear programming model.
coefficient strengthening: Techniques for modifying coefficients of the model in such a way that a constraint is still valid when a 0-1 integer variable is 0 or 1, but is otherwise a stronger constraint.
column: Used in linear programming terminology as a synonym for variable.
common block: A storage area that may be referred to by a FORTRAN calling program and one or more FORTRAN subprograms.
complementarity violation: The product of a primal variable and its corresponding dual variable, such as a primal variable and its reduced cost.
configuration file: In the AIX operating system, a file that specifies the characteristics of a system or subsystems.
constraint cell: A cell in a spreadsheet containing a problem constraint. Constraint cells in a Lotus 1-2-3 spreadsheet should be expressed as logical formulas. For example, the formula 2*(A1-B1)<=5 can be used to define a constraint in a Lotus 1-2-3 spreadsheet.
constraints: The limiting rows and bounds on variables of a model. The model constraints are the relationships expressed in the rows and in the range between variables and bound vectors.
continuous model: A linear programming model with no integer variables.
continuous solution: The solution to the linear part of a mixed-integer programming problem (ignoring the fact that some variables must be integers).
continuous variable: A variable that is not constrained to be an integer.
control variable: A variable that affects all subroutine calls within a program, such as how frequently log information will be recorded.
convex: A closed set in which a line segment between any two points in the set is also in the set. For example, the set of feasible solutions to a linear programming problem is convex.
convex function: A function f from a nonempty convex set S subset <doubleR sup N> to <doubleR sup 1> is said to be convex if for every bix, biy memberof biS and for each real scalar lambda, 0 lt lambda lt 1, f (lambda bix + (1 - lambda) biy) le lambda f(bix) + (1 - lambda) f (biy).
convexity constraint: In spreadsheet format, an equality constraint with coefficients for the problem variables of 0 or 1 and a RHS of 1.
costs: The objective function coefficients, which are called costs because they often represent the dollar cost of a commodity. For example, in a feed blending problem in which we want to minimize costs, the objective function is the sum of the costs of each type of feed multiplied by the amount of each type of feed used.
coupling constraints: The constraints of a Dantzig-Wolfe problem that have nonzero coefficients in columns belonging to two or more of the subblocks.
crash: To find a starting basis.
current basis: The basis at any iteration of the simplex method, or after a basis is read in, or after crash processing.
current directory: For the AIX and SunOS operating systems, the directory that is the starting point for relative path names, which can be displayed with the pwd command.
Dantzig-Wolfe problem: Also known as a block angular form, a problem with a collection of disjoint blocks in the constraint matrix and a small set of coupling constraints. The coupling constraints may have nonzero elements in columns corresponding to two or more of the blocks.
data type: The structural characteristics, features, and properties of data that may be directly specified by a programming language; for example, integers, real numbers in FORTRAN; arrays in APL2.
demand node: A node with a negative upper bound on its exogenous flow. Contrast with supply node.
Devex pricing: In the simplex method, a way of selecting entering columns based on scaling the reduced costs with a standard frame of reference.
direct access storage: Storage in which the access time is in effect independent of the location of the data.
direction of unboundedness: In an unbounded problem in which the objective function can be driven infinitely downward (or upward) in a minimization (or maximization) problem without violating any constraints, the direction of unboundedness gives the direction the objective function can move without bounds.
directory: For the AIX operating system, a type of file containing the names and controlling information for other files or other directories. Directories in the AIX operating system are arranged in a tree-like structure.
doubleton row: A row containing two nonzero coefficients, while the rest of the row is made up of zeros.
doubleword: Eight bytes of storage. In the Optimization Library, "doublewords" is usually used to mean the memory required to store double precision real variables.
dspace: The user-supplied storage area where Optimization Library modules do their processing, addressable in doublewords.
dual problem: See primal problem.
dual algorithm: An algorithm for which all intermediate solutions are feasible for the dual problem. The dual algorithm in EKKSSLV and the primal-dual algorithms in EKKBSLV do not maintain dual feasibility of the intermediate solutions, but dual feasibility is gradually achieved.
efficient frontier: For portfolio analysis problems, the values of the solutions to a parametric family of quadratic programming problems for a range of values of the parameter.
entering arc: The nonbasic arc that has been selected to be added to the basis. Contrast with leaving arc.
environment variable: In the AIX and SunOS operating systems, a name for a data item that is used by a process. AIX and SunOS environment variables can correspond to VS FORTRAN unit numbers.
equality row: A constraint row of the form <biA sub <i &star.> bix> = <b sub i>
exogenous flow: The right-hand side element of a network constraint if the constraint is an equality. If the constraint is not an equality, the exogenous flow is the sum of the flows on the outgoing arcs minus the sum of the flows on the incoming arcs.
factorization: A representation of a matrix as two or more matrices so that when they are multiplied together in order, the original matrix results.
feasible region: The region consisting of all feasible solutions.
feasible solution: A solution where the activities of logical and structural variables are within their bounds.
fixed variables: Those variables that may take only a specified value. Each variable to be fixed will have its value specified as an entry in the bound vector.
flag: In the AIX and SunOS operating systems, a modifier, appearing on a command line, that defines the action of the command. Flags in the AIX and SunOS operating systems are almost always preceded by a dash.
flow: The value of a variable in a network problem.
free variables: Those variables that may take on any value between plus and minus infinity.
full path name: In the AIX and SunOS operating systems, the name of any directory or file expressed as a string of directories and files beginning with the root directory (/).
fullword: Four bytes of storage. In the Optimization Library, "fullwords" is usually used to mean the memory required to store single precision usually used for INTEGER*4 variables.
generalized upper-bound rows: Sets of rows with no columns in common and all coefficients 1.0. The rows may be L-type or G-type rows.
GUB: See generalized upper-bound rows.
ill-conditioned matrix: A matrix, where small changes to a vector x result in large changes to Ax.
incidence matrix: See node-arc incidence matrix.
incoming arc: The incoming arcs for node i are columns that have a -1 in row i in the coefficient matrix for a network programming problem. Contrast with outgoing arc.
independent rows and columns: A set of rows or columns in which no linear combination of any of them exists that is equal to another member of the set.
index: A position within an array.
indicator records: MPS format records containing a single word that specifies the type of data that follows.
infeasibility: A variable (column) or row that is outside its bounds. Information provided about infeasibilities includes both the number of infeasibilities (such as the number of rows and columns that are out of bounds) and their sum (such as the sum of all the amounts by which each row or column violates its bounds).
infeasible region: The complement of the feasible region.
infeasible solution: A solution where not all the activities of logical and structural variables are within their constraints.
input redirection: In the AIX and SunOS operating systems, the specification of an input source other than the standard one. The symbol < is used on the command line to indicate input redirection.
integer solution: A set of activities for the variables that satisfies the model constraints and in which the integer variables have integer values.
integer variables: Variables that must take only integer values (..., -2, -1, 0, 1, 2, ...) between their bounds.
interior-point method: A method of solving linear programming problems by stepping through the interior of the feasible region.
input cost: The cost of an activity input to the problem.
JCL: Job control language. (A)
job control language (JCL): A problem-oriented language designed to express statements in a job that are used to identify the job or describe its requirements to the operating system. (A)
leaving arc: The basic arc that has been selected to be removed from the basis. Contrast with entering arc.
linear programming (LP): A technique for finding the best of all possible solutions of a system of linear equalities and inequalities. The criterion for the best solution is the maximum or minimum value of a given linear function of bounded variables, called the objective function.
link: In the AIX operating system, a connection between the information pertaining to a file and one or more names associated with it.
logarithmic barrier method: A method of solving linear programming problems by reformulating the problem as: 'minimize' %% sum from j=1 to n of lbrace <c sub j x sub j - µ ln(x sub j - l sub j) - µ ln(u sub j - x sub j)> rbrace subject to: biA bix = bib
logical variable: An additional variable introduced for a constraint row. Logical variables are slack variables for inequality rows and artificial variables for equality rows.
lower bound: See bound.
LP: Linear programming.
LP model: A model where all constraints and the objective function are linear.
main program: In FORTRAN, a program unit, required for execution, that can call other program units but cannot be called by them.
mask: To use a pattern of characters to control the retention or elimination of portions of another pattern of characters. (I) (A)
mathematical programming (MP): A generic term covering the optimization algorithms for linear programming, mixed-integer programming, and nonlinear programming.
matrix: A rectangular array of elements, in rows and columns, that can be manipulated based on matrix algebra rules. (I) (A)
matrix block: See block.
maxint: The maximum integer allowed on your platform.
maxreal: The maximum real (doubleword) value allowed on your platform.
MIP: Mixed-integer programming.
MIP model: A model with linear constraints and both continuous and integer variables.
mixed-integer programming (MIP): The linear programming technique for models in which certain variables may take on only integer values.
model: (1) A specific mathematical programming problem to be solved. An application may contain one or more models. (2) A representation of a real-world system as a set of variables and constraining relationships.
MP: Mathematical programming.
MPS: Mathematical Programming System.
MPSX/370: Mathematical Programming System Extended/370.
MPS file: A sequential file containing model data in MPS format.
MPS format: An alphanumeric data format, widely adopted in mathematical programming to enter data into mathematical programming software.
mspace: The user-supplied storage area where Optimization Library modules do their processing, addressable in fullwords.
network programming problems: A linear programming problem whose matrix is a node-arc incidence matrix.
network simplex method: A specialization of the simplex method for network programming problems.
neutral row: A row that is not constrained (there is no active right-hand side and no range). Objective functions are examples of neutral rows.
node: A row of the LP matrix for a network programming problem.
node-arc incidence matrix: The constraint matrix of a (pure) network programming problem. Its rows correspond to nodes, and its columns correspond to arcs.
nonbasic variables: The (n-m) variables, in an m-row n-column linear programming model, that are not in the basis. The activity value of a nonbasic variable must be either its lower bound (zero if no explicit lower bound) or its upper bound.
null space: In matrix A, the set of vectors x, such that Ax=0.
objective cell: A cell in a spreadsheet that contains an objective function. The objective cell in a Lotus 1-2-3 spreadsheet should be a numeric formula, such as (plus C1 minus 2) '*' (C2 minus G5). For this function, any or all of the cells C1, C2, or G5 can be adjustable cells.
objective function: A neutral row, specified as the optimization target to be maximized or minimized.
objective row: In a simplex tableau, the row corresponding to the objective function.
optimal solution: The feasible solution to the set of constraints that gives the best possible value for the optimization target, that is, the maximum or minimum value for the objective function.
optimal integer solution: A solution that has all integer variables at integer values and gives a value to the objective function such that no other integer solution is better.
outgoing arc: The outgoing arcs for node i are the set of columns that have a +1 in the coefficient matrix for a network programming problem. Contrast with incoming arc.
output redirection: In the AIX and SunOS operating systems, the specification of an output destination other than the standard one. The symbol > is used on the command line to indicate output redirection.
parametrics: The practice of modifying a mathematical programming model by simultaneously adding incremental changes to one or more of the costs or bounds.
path: Given a tree containing two nodes, the set of arcs and nodes that connect them.
path name: In the AIX and SunOS operating systems, the sequential list of directory names that identifies the location of a particular directory or the sequential list of directory names and file name that identifies the location of a particular file in the file hierarchy. Each file has a full path name, beginning with the root directory (/) and ending with the file's name. The symbol / is used to separate directory and file names.
perturbation: A method of reducing degeneracy by adding or subtracting small tolerances to costs, matrix elements, right-hand side elements, or bounds.
phase 1: A part of the simplex or interior-point primal algorithm concerned with making the solution feasible.
phase 2: A part of the simplex or interior-point primal algorithm in which a feasible solution is improved toward optimality.
platform: A particular type and level of hardware and operating system environment.
positive definite matrix: Matrix biA in which bix sup T biA bix is positive for all nonzero vectors bix.
positive semidefinite matrix: Matrix biA in which bix sup T biA bix is nonnegative for all vectors bix.
predictor-corrector method: An efficient variant of the primal-dual interior point barrier method.
primal simplex algorithm: An algorithm for which all intermediate solutions are feasible for the primal problem. Simplex-based primal algorithms select a variable with a corresponding dual infeasibility and remove the infeasibility by changing the value of that variable and others. Interior-point methods generate a sequence of interior points by establishing a search direction at each point and selecting the next point to lie somewhere in this direction. The primal algorithms in EKKSSLV, EKKNSLV, EKKQSLV, and the primal-dual methods in EKKBSLV do not maintain primal feasibility of the intermediate solutions, but primal feasibility is gradually achieved.
primal problem: The original problem formulation for a linear or quadratic programming problem. An LP problem may be either a minimization or maximization problem. Convex quadratic programming problems with linear constraints must be minimization problems and must have a positive semidefinite quadratic matrix. Standard formulations for these problems are:
principal minor: Any square submatrix of a square matrix M that includes the upper left-hand element of M.
print unit: A unit used for printing output from a program. See unit.
probing: A logical technique for determining integer feasibility by setting 0-1 variables to 0 or 1 and observing all the consequences. This may show that setting the variable one way is feasible, enabling the variable to be fixed the other way.
problem variables: The unknown quantities of a mathematical programming model.
projection: In the interior-point barrier method for linear programming, finding the vector p in the direction of steepest descent, which lies in the null space of matrix A.
pseudocost: Usually applying to 0-1 variables, the estimated change in the objective function resulting from forcing an integer variable to its upper or lower limit (hence, up- and down-pseudocosts). Pseudocosts are used to decide which nodes and branches to pursue first.
QP: Quadratic programming.
quadratic: Pertaining to a problem that typically has linear constraints, but the objective function is of the form: bic sup T bix + 1 over 2 <bix sup T> biQ bix where biQ is a positive semidefinite matrix.
quadratic matrix: In the quadratic programming problem: <'minimize' %% f(bix) = <bic sup T> bix + 1 over 2 <bix sup T> biQ bix> subject to: <biA bix % = % bib> labove <bix % ge % bi0> The positive semidefinite matrix biQ, which has dimension <n times n>, is referred to as the quadratic matrix.
quadratic parametric family: A family of quadratic programming problems with objective function of the form: bic sup T bix + lambda bix sup T biQ bix or bic sup T bix + lambda bid sup T bix + <1 over 2> bix sup T biQ bix where biQ is a nonnegative semidefinite matrix and &lambda. is a nonnegative scalar. As &lambda. varies, a family of problems is described.
random pricing: In the simplex method, a way of selecting an entering column based upon the previous iteration's reduced costs. A random subset of the columns is selected, and a potential entering column is selected. This is generally more efficient early in the algorithms, since selecting the entering variable uses a lot of computer time.
range vector: A vector of elements comprising additional constraints on the rows of a model. Each range element consists of the distance between the existing right-hand side and the desired additional constraints.
rank: The rank of a matrix A is equal to the maximum number of linearly independent columns of A. It is also equal to the maximum number of independent rows of A.
rank deficient: The m % x % n matrix A is rank deficient if the rank of A is less than the minimum of m and n.
reduced cost: For a variable, the variation of the objective function if the variable is relaxed by one unit (made higher if the variable is at its upper bound, or made lower if the variable is at its lower bound), provided such relaxation is physically or economically possible. The reduced cost of a nonbasic variable during the optimization process indicates the value per unit of variable activity when increasing or decreasing the activity of the variable. A variable that is at one of its bounds in a linear programming solution can have a nonzero reduced cost.
reference row entry: A value used in determining the optimum branch to take during branch-and-bound processing in mixed-integer programming. An ordered set of these values is used with special ordered sets to compute relative values and make the branch determination.
relative path name: In the AIX and SunOS operating systems, the name of a directory or file expressed as a sequence of directories followed by a file name, beginning from the current directory.
rewind: To position a sequentially accessed file at the beginning of the first record of the file.
RHS: Right-hand side.
right-hand side vector: A vector of elements comprising the constant terms of the equalities and inequalities of the model.
row: In linear programming, the set of coefficients that expresses a linear relationship between variables. A row is either a constraint or a neutral row.
row activity: The current value of a constraint row.
scalar: (1) A quantity characterized by a single value. (I) (A). (2) Contrast with vector.
scalar processing: Execution of machine instructions that operate on scalars, where a scalar is a single data item. Contrast with vector processing.
scale: To improve the numerical and algorithmic stability of a problem by making the matrix coefficients differ in magnitude by at most 1.
shell: In the AIX and SunOS operating systems, a program that accepts and interprets commands for the operating system. Commonly used shells are the Bourne Shell, the C Shell, and the Korn Shell.
simplex method: A method of solving linear programming problems by going from one basic feasible solution of the feasible region to another until an optimal solution is reached.
sink node: A demand node. Contrast with source node.
slack arc: A slack variable for a network problem.
slack variables: The variables implied by inequality constraints whose activities form the difference between the row activities and the RHS value for the row.
SOS: Special ordered set.
SOS type 1: An ordered set of variables, where at most one variable may take on a nonzero value.
SOS type 2: An ordered set of variables, where at most two variables may take on nonzero values, and if two variables are nonzero, they must be adjacent in the set.
SOS type 3: A set of 0-1 variables only one of which may be selected to have the value 1, the other variables in the set having the value 0.
SOS row: The row that links a set of SOS variables.
source node: A supply node. Contrast with sink node.
solution cell: See adjustable cell.
sparse matrix: A matrix composed mostly of zeros.
staircase problem: A problem with identifiable blocks in the constraint matrix that are almost disjoint. Blocks do not share rows, and each block may share columns only with its two neighboring blocks.
standard input: In the AIX and SunOS operating systems, the primary source of data going into a command. Standard input (stdin) comes from the keyboard unless input redirection or piping is used, in which case standard input can be from a file or the output from another command.
standard output: In the AIX and SunOS operating systems, the primary destination of data coming from a command. Standard output (stdout) goes to the display unless output redirection or piping is used, in which case standard output can be to a file or another command.
stanza: In the AIX and SunOS operating systems, a group of lines in a file that together have a common function. Stanzas are usually separated by blank lines, and each stanza has a name.
starting basis: A basis used at the beginning of the simplex method. A common starting basis has all slack variables as basic, with others nonbasic.
steepest descent method: A method of solving optimization problems by selecting a direction within the feasible region to give the most rapid improvement in the objective function. An iteration involves a step in this direction, followed by a recalculation of the steepest descent direction.
structural variable: A variable explicitly defined in the model which represent modelled decisions and possible actions.
subscript: (1) A symbol associated with the name of a set to identify a particular subset or element. (I) (A) (2) A subscript expression or set of subscript expressions, enclosed in parentheses and used with an array name to identify a particular array element.
subscript expression: An integer expression in a subscript whose value and position in the subscript determine the index number for the corresponding dimension in the referenced array.
supernode: In the branch-and-bound method, a node for which all the preprocessing techniques used in EKKMPRE, such as probing and coefficient reduction, are used. Several variables may be fixed inside a supernode. When exiting EKKMPRE, one or more new nodes will be formed.
supply node: A node with a positive lower bound on its exogenous flow. Contrast with demand node.
tableau: Used in most linear programming textbooks for the simplex method, a table used to represent the state of the constraints and objective function at each iteration.
tolerances: Specific values that are used to handle problems of numerical accuracy, which arise in calculations on computers with finite real number representations.
traceback: A debugging aid that shows the subroutines called in reverse order. This gives a trail from the subroutine back to the main program.
transshipment network: See network programming problem.
transportation network: A network for which all nodes can be partitioned into two disjoint sets so that every node in the first set has no incoming arcs, and each node in the second set has no outgoing arcs.
tuning: The adjustment of standard parameters and tolerances to permit faster solution of a problem.
type declaration: The explicit specification of the type of a constant, variable, array, or function by use of an explicit type specification statement.
unbounded: Pertaining to a feasible optimization problem that has no finite optimal solution.
unit: In FORTRAN, a file or device. A unit may be a disk, a file, a printer, or a terminal. Before you can write data to or read data from a file, the file must be associated with--that is, connected to--a unit.
upper bound: See bound.
user exit subroutines: An option provided by an IBM software product that may be requested by Calls to specified subroutines, the user exit subroutines, that are built into Optimization Library modules to provide user contorllable interrupts to module processing. The sample user exit subroutines that are distributed with the library (in all but one case) simply pass contnrol directly back to the calling program. A user may replace these stubs with new subroutines to extensively customize execution of library modules.
variables: The unknown quantities of a mathematical programming model.
vector: A one-dimensional ordered collection of scalars. Contrast with scalar.
Vector Facility: The hardware feature of IBM 3090 and ES/9000 processors that provides the capability to perform mathematical operations on vectors.
vector processing: Execution of machine instructions that operate on vectors, where a vector is an ordered set of scalars. Contrast with scalar processing.
wildcard: A character used in pattern matching that represents the occurrence of zero or more characters.
work area: A storage area provided by the application program for the use of a subroutine. Throughout this book, the work area is referred to as dspace.
working directory: See current directory.
0-1 variables: Those integer variables that may only take the values zero or one.

[ Top of Page | Previous Page | Next Page | Table of Contents ]