Reuse of software documents requires a repository that permits to find relevant documents easily. For later retrieval each document is indexed with keywords inependent of others. Keywords and documents together form a formal context which implies a concept lattice. The concept lattice is a canonical clustering of objects and their attributes derived from the indexing. It permits an efficient incremental retrieval with context sensitive feedback to the user. All operations are formally introduced together with examples. For an implementation of the proposed method a suitable representaion of the concept lattice is shown its typical complexity investigated by the means of some experiments. At least the proposed method is compared with related methods.