Data Warehouse Fixer: Fixing Inconsistencies in Data Warehouses

Abstract

Dimensions in Data Warehouses (DWs) are set of elements connected by a hierarchical relationship. Usually, dimensions are required to be strict and covering to support summarizations at different levels of granularity. A dimension is strict if all they rollup relations are functions, and is covering if every element in a category is connected with an element in its ancestor categories. We present the Data Warehouse Fixer (DWF), a system that restores consistency of dimensions that fail to satisfy their strictness constraints. The system checks consistency, computes minimal repairs for inconsistent dimensions by implementing Datalog programs with negation and weak constraints, and also fixes inconsistent dimensions.