[Paper Review] View Synthesis from Schema Mappings

Diego Calvanese, Giuseppe De Giacomo+2 more|arXiv (Cornell University)|Mar 5, 2010

Advanced Database Systems and Queries35 references3 citations

TL;DR

This paper introduces view synthesis from schema mappings, a novel problem in data integration and warehousing that automatically generates target schema views from source-to-target schema mappings. It proposes techniques for relational and semistructured data using conjunctive and regular path queries, establishing complexity upper bounds: ExpTime for 2RPQs and 2ExpSpace for CRPQs.

ABSTRACT

In data management, and in particular in data integration, data exchange, query optimization, and data privacy, the notion of view plays a central role. In several contexts, such as data integration, data mashups, and data warehousing, the need arises of designing views starting from a set of known correspondences between queries over different schemas. In this paper we deal with the issue of automating such a design process. We call this novel problem "view synthesis from schema mappings": given a set of schema mappings, each relating a query over a source schema to a query over a target schema, automatically synthesize for each source a view over the target schema in such a way that for each mapping, the query over the source is a rewriting of the query over the target wrt the synthesized views. We study view synthesis from schema mappings both in the relational setting, where queries and views are (unions of) conjunctive queries, and in the semistructured data setting, where queries and views are (two-way) regular path queries, as well as unions of conjunctions thereof. We provide techniques and complexity upper bounds for each of these cases.

Motivation & Objective

• Address the challenge of automatically generating target schema views from source-to-target schema mappings in data integration and warehousing.
• Formalize view synthesis as a query rewriting problem where source queries must be rewritable over synthesized views.
• Provide decidability and complexity analysis for view existence under different query languages: conjunctive queries, 2RPQs, CRPQs, and UC2RPQs.
• Extend the framework to semistructured data models using regular path queries and their extensions.
• Explore the theoretical foundations for view synthesis with sound and exact rewriting semantics.

Proposed method

• Model view synthesis as a query rewriting problem: for each source query, synthesize a target view such that the source query is logically equivalent to a rewriting over the view.
• Use automata-theoretic techniques to reduce query containment problems to language containment over word automata (NWA and 2NWA).
• Apply tree automata and congruence class techniques to manage view closure and minimize view complexity.
• Reduce containment of complex queries (e.g., CRPQs) to containment of regular languages via word automata constructions.
• Leverage reductions from query containment to automata language inclusion to enable decidability and complexity analysis.
• Construct view-defining automata from source and target mappings and verify inclusion constraints to ensure correctness of synthesized views.

Experimental results

Research questions

• RQ1 Can we automatically synthesize target schema views from a set of source-to-target schema mappings such that source queries are rewritable over the views?
• RQ2 What are the computational complexity bounds for the view existence problem under different query languages (e.g., 2RPQs, CRPQs)?
• RQ3 How does the complexity of view synthesis vary between sound rewriting (certain answers) and exact rewriting (logical equivalence) semantics?
• RQ4 Can automata-theoretic methods be effectively used to reduce query containment and enable view synthesis in semistructured data models?
• RQ5 What are the theoretical limits of view synthesis in terms of decidability and complexity for richer query languages like CRPQs and UC2RPQs?

Key findings

• The view-existence problem is decidable for both relational and semistructured data models, with complexity upper bounds depending on the query language.
• For 2RPQs, the view-existence problem is in ExpTime and the exact view-existence problem is in ExpSpace.
• For CRPQs, the view-existence problem is in 2ExpTime and the exact view-existence problem is in 2ExpSpace.
• For UC2RPQs, the complexity bounds match those of CRPQs due to the use of 2RPQ views and automata-based containment checks.
• The use of word automata (NWA and 2NWA) enables reduction of complex query containment problems to language inclusion, forming the core of the synthesis technique.
• The congruence-class technique allows for view minimization and closure under equivalence, improving efficiency and correctness of synthesized views.