Indexes and path constraints in semistructured data

In this paper, we study semistructured data and indexes preserving inclusion constraints. A semistructured datum is modelled by multi-rooted edge-labeled directed graphs. We consider regular path queries and inclusion constraints other these data. These constraints are binary relations over regular path expressions q and r, and are interpreted on a datum as "for this datum, the answer to query q is included in the answer to query r". We study how to represent inclusion constraints that are common to several data. Our work is based on two existing indexes: dataguide and 1-index. Given a set of data S, we extract from the dataguide of S a finite set C(S) of (finite) inclusion constraints such that an inclusion constraint is satisfied by a datum of S if and only if it is implied by C(S). We use 1-index which are covering indexes preserving inclusion constraints. Experiments compare the different ways of using the 1-index to index a set of data.