Evaluating View Materialization Strategies for Complex Hierarchical Objects *

In many design applications, it is common practice to store complex hierarchical objects in a compact folded form to save storage space and to reduce processing costs for accessing the objects. In these folded representations, complex objects are built up from identical and otherwise indistinguishable design objects. However, it is often necessary, especially during the refinement of data, to distinguish between these identical folded objects by personalizing a subset of them. The established practice is to explicitly unfold the hierarchical objects and thus create space in which to store distinct personalization data for each object occurrence. However, this explicit unfolding is costly and time consuming, resulting in a potentially much larger structure, and substantially increasing the costs of querying and updating the design. Therefore, we propose an unfold view operator and provide the basis for updating of customized values for each hierarchical sub-object through the unfolded view. We propose alternative strategies for the maintenance of personalization values, representing various portions of the view materialization spectrum. We present a performance evaluation comparing these strategies as well as the traditional explicit unfolding approach. Our evaluation indicates the trade-offs in terms of storage and query costs and compares the costs to do implicit unfolding through a view rather than explicit unfolding of complex hierarchical objects.