AQL: an Alignment Based Language for Querying String Databases

AQL stands for Alignment Query Language, and it is an extension of SQL. AQL allows for declarative querying of databases containing strings of characters as entries. For instance, in applications such as genomic databases, the DNA-sequences are strings, and they need to be queried based on their structure. The LIKE-operation in SQL does not go very far. Existing systems for string databases have a very ad hoc nature, usually giving the user a set of predefined simple query forms. On the other hand, there has recently been considerable theoretical advancement in the field of string databases. Several languages have been proposed, and their expressive power is becoming illuminated. However, there has been no practical demonstration of the feasibility of these languages. In this paper we describe an implementation of Alignment Calculus, a powerful string database language based on a modal logic extension of relational calculus. The implementation yields AQL. In AQL the user can write declarative string predicates in the WHERE-clause of a query. The predicates can be of an arbitrary arity, expressing for instance a unary property of a string being a palindrome, or expressing a binary property of one string being the reversal on another, and so on. The string predicates are translated into a novel form of a high level finite state automaton, and they are evaluated using an inference engine for chasing these automata. The AQL system has been implemented as an extension of a relational database management system. Our experiments show that the inference engine has the efficiency of an equivalent query computed in compiled Prolog. This is efficiency achieved already without using optimisation, something which the string queries due to their declarative nature are amenable to.