Table of Contents
Fetching ...

Characterizing Data Dependencies Then and Now

Phokion G. Kolaitis, Andreas Pieris

TL;DR

The chapter surveys J. Makowsky's foundational contributions to data dependencies, first linking their logical structure to decidability issues and then providing finite axiomatizability characterizations. It shows that for full TGDs and EGDs, a finite axiomatization is equivalent to 1-criticality, domain independence, modularity, and closure under subdatabases and direct products, with an alternative characterization via closure under intersections. It then extends these ideas to arbitrary TGDs and EGDs through locality concepts, establishing a general theorem: a collection is finitely axiomatizable by $(n,m)$-TGDs and $n$-EGDs if and only if it is 1-critical, $(n,m)$-local, and closed under direct products. The discussion connects old results with modern frameworks (Robinson diagrams, McKinsey’s method) and highlights applications to data exchange, integration, and ontologies, demonstrating the enduring relevance of Makowsky’s structural approach to data dependencies.

Abstract

Data dependencies are integrity constraints that the data of interest must obey. During the 1980s, Janos Makowsky made a number of contributions to the study of data dependencies; in particular, he was the first researcher to characterize data dependencies in terms of their structural properties. The goal of this article is to first present an overview of Makowsky's work on characterizing certain classes of data dependencies and then discuss recent developments concerning characterizations of broader classes of data dependencies.

Characterizing Data Dependencies Then and Now

TL;DR

The chapter surveys J. Makowsky's foundational contributions to data dependencies, first linking their logical structure to decidability issues and then providing finite axiomatizability characterizations. It shows that for full TGDs and EGDs, a finite axiomatization is equivalent to 1-criticality, domain independence, modularity, and closure under subdatabases and direct products, with an alternative characterization via closure under intersections. It then extends these ideas to arbitrary TGDs and EGDs through locality concepts, establishing a general theorem: a collection is finitely axiomatizable by -TGDs and -EGDs if and only if it is 1-critical, -local, and closed under direct products. The discussion connects old results with modern frameworks (Robinson diagrams, McKinsey’s method) and highlights applications to data exchange, integration, and ontologies, demonstrating the enduring relevance of Makowsky’s structural approach to data dependencies.

Abstract

Data dependencies are integrity constraints that the data of interest must obey. During the 1980s, Janos Makowsky made a number of contributions to the study of data dependencies; in particular, he was the first researcher to characterize data dependencies in terms of their structural properties. The goal of this article is to first present an overview of Makowsky's work on characterizing certain classes of data dependencies and then discuss recent developments concerning characterizations of broader classes of data dependencies.
Paper Structure (10 sections, 16 theorems, 24 equations, 2 figures)

This paper contains 10 sections, 16 theorems, 24 equations, 2 figures.

Key Result

theorem 1

Let $C$ be a collection of databases. The following statements are equivalent:

Figures (2)

  • Figure 1: $C$ is $(n,m)$-locally embeddable in $D$.
  • Figure 2: The function $\lambda = \mu_L \circ g$ in the proof of Lemma \ref{['lem:locality']}.

Theorems & Definitions (26)

  • theorem 1: Theorem 5 in DBLP:journals/acta/MakowskyV86
  • theorem 2
  • proof
  • lemma 1
  • proof : Proof of Lemma \ref{['lem:aux-lemma-1-full-instersection']}
  • lemma 2
  • proof : Proof of Lemma \ref{['lem:aux-lemma-2-full-instersection']}
  • lemma 3
  • definition 1: Locality
  • lemma 4
  • ...and 16 more