Foundations of relational similarity-based query language RESIQL

We present a query language for similarity-based relational model of data with support of ordinal ranking and imperfect matches. The model utilized in this paper results from Codd's model of data by considering general residuated partially ordered scales of degrees of satisfaction in place of the two-element Boolean scale implicitly used in the Codd model. As a result, we look at the traditional relational model of data from the perspective of general scales allowing us to represent partial matches of relational queries. From the data representation point of view, the basic structures representing both the stored data and results of queries are so-called ranked data tables, generalizing the ordinary data tables (relations). Compared to other approaches dealing with ranking and imperfect matches, ours has several benefits: (i) it is based on solid logical foundations-we do not introduce an additional ranking module on top of the Codd model but have all the necessary concepts as an integral part of the model, (ii) the model is very general and several ranking procedures introduced in the past can be shown to be particular cases of our model, (iii) the model is not oriented just to query answering, it allows to handle similarity-based data dependencies, approximate constraints and other issues that cannot be properly formalized in the classic Codd model, (iv) despite of its generality, query execution in the model is tractable. In this paper, we present logical foundations of the query language and show how it implements the underlying similarity-based model of data.