There is also the possibility to handle complex numbers and vectors/matrices with computable real entries.
The concept on which the representation is based is accuracy. A real/complex number or a vector/matrix is represented by an object that maintains: a rational approximation and its accuracy; a function which, given an accuracy, returns an approximation of that accuracy, and/or a function that gives a more accurate approximation.
When implementing the arithmetic, a thorough analysis of error propagation must be done: the accuracy of an approximation of the result depends on the accuracies of approximates of the operands; this dependency has to be determined.
In the polynomial arithmetic, root isolation and extraction of a root as an erna object are currently implemented.
In linear algebra, the solution of linear systems with singular or nonsingular matrix is implemented.
The package is written in Maple.