In smart building systems, the automatic control of devices relies on matching the sensed environment information to customized rules. With the development of wireless sensor and actuator networks (WSANs), low-cost and self-organized wireless sensors and actuators can enhance smart building systems, but produce abundant sensing data. Therefore, a rule engine with ability of efficient rule matching is the foundation of WSANs based smart building systems. However, traditional rule engines mainly focus on the complex processing mechanism and omit the amount of sensing data, which are not suitable for large scale WSANs based smart building systems. To address these issues, we build an efficient rule engine. Specifically, we design an atomic event extraction module for extracting atomic event from data messages, and then build a $\beta$-network to acquire the atomic conditions for parsing the atomic trigger events. Taking the atomic trigger events as the key set of MPHF, we construct the minimal perfect hash table which can filter the majority of the unused atomic event with O (1) time overhead. Moreover, a rule engine adaption scheme is proposed to minimize the rule matching overhead. We implement the proposed rule engine in a practical smart building system. The experimental results show that the rule engine can perform efficiently and flexibly with high data throughput and large rule set.