The paper is concerned with the construction of an adaptive beamformer, which uses the dominant subspace information of a sample covariance matrix. We assume that the sample covariance matrix and its eigenvalue decomposition (EVD) are to be updated as data arrive.  We present different structures for updating the EVD and introduce a robust method for adapting the rank of the dominant subspace of the EVD. We use the EVD of the sample covariance matrix to build a fast rank- and weight-adaptive beamformer. A useful component of the beamforming routine is a fast algorithm for updating the estimates of the direction of arrival of the dominant sources.