Deriving optimized integrity monitoring triggers from dynamic integrity constraints

M. Gertz, U.W. Lipeck
1996 Data & Knowledge Engineering  
Modern approaches to integrity monitoring in active databases suggest to generate triggers from constraints as part of database design and to utilize constraint simplification techniques for trigger optimization. Such proposals, however, have been restricted to static conditions only. In this paper, we show how to derive triggers from dynamic integrity constraints which describe properties of state sequences and which can be specified by formulas in temporal logic. Such constraints can
more » ... tly be transformed into transition graphs which describe such life cycles of database objects that are admissible with respect to the constraints: Nodes correspond to situations in life cycles, and edges give the (changing) conditions under which a change into another situation is allowed. If object situations are stored, integrity monitoring triggers can be generated from transition graphs for all situations and all critical database operations. Additionally, new simplification techniques can be developed by identifying characteristic preconditions in the graphs and by utilizing invariants. Maintenance of object situations can be supported by triggers as well. In this paper we want to study dynamic integrity constraints, i.e. constraints on state sequences instead of single states as in the static case. Based on our former work on monitoring schemes for such dynamic constraints [21, 18, 25 ] and on rules for transforming constraints into transaction specifications [18, 19] , we now want to explain how to use such constraint analysis techniques in the presence of ad-hoc transactions, i.e. arbitrary unforeseen sequences of insert-, update-, and delete-operations, provided that appropriate trigger mechanisms are available. By using temporal logic with operators like always, sometime, before, etc. rather general dynamic constraints can be expressed, starting with conditions on state transitions and also including long-term relations between database states. We have shown
doi:10.1016/s0169-023x(96)00010-9 fatcat:26lx7usxcrfphehih4ofzv3eqq