The Lambert W function is a multivalued complex function, first named in the computer algebra system Maple. We present iterative schemes and strategies for the numerical evaluation of all branches of the scalar complex Lambert W function to hardware precision with high computational efficiency, and present a set of rules for the simplification of special symbolic arguments. We also extend the numerical and symbolic computations to the Lambert W function in C nxn , for n > 1. In order to achieve high precision and computational efficiency, we evaluate a series of high order and classical iterative methods and strategies for the evaluation of the scalar Lambert W function. We then construct optimal iterative schemes for the evaluation of the complex Lambert W function in the IEEE oating point model. The schemes consist of variations on Newton and Halley iterations together with initial estimates generated using a variety of series approximations. We also study several classes of exact simplifications for the Lambert W function for symbolic arguments and give rules for their application. Finally, we consider the solutions of the matrix equation S exp(S) = A, where S and A are n x n matrices. The solutions are expressed in terms of extensions of the scalar Lambert W function to C nxn . The solutions of the matrix equations consist not only of the matrix functions W(A); other solutions also exist. We focus first on solving the matrix equation in C 3x3 , and implement solutions in the floating-point case, and the symbolic case, using Maple.