The timestamps of now-relative bitemporal databases are modeled as growing, shrinking, or rectangular regions. The shape of these regions makes it a challenge to design bitemporal operators that a) are consistent with the pointbased interpretation of a temporal database, b) preserve the identity of the argument timestamps, c) ensure locality, and d) perform efficiently. We identify the bitemporal split operator as the basic primitive to implement a wide range of advanced now-relative bitemporal operations. The bitemporal split operator splits each tuple of a bitemporal argument relation, such that equality and standard nontemporal algorithms can be used to implement the bitemporal counterparts with the aforementioned properties. Both a native database algorithm and an SQL implementation are provided. Our performance results show that the bitemporal split operator outperforms related approaches by orders of magnitude and scales well.
Mikkel Agesen, Michael H. Böhlen, Lasse Pouls