Abstract
We propose a query language L ARE for graphs whose edges are labelled by a finite alphabet and nodes store unbounded data values. L ARE is based on a variant of regular expressions with memory. Queries of L ARE can compare quantities of memorised graph nodes and their neighbourhoods. These features allow us to express a number of natural properties, while keeping the data complexity (with a query fixed) in non-deterministic logarithmic space. We establish an algorithm that works efficiently not only with L ARE, but also with a wider language defined using effective relational conditions, another formalism we propose.