DatabaseSupported Program Execution

点击翻译按钮获取中文摘要

We continue our research into database-supported programming environments in which developers are no longer required to be bilingual (programming and database language) to construct efficient data-rich applications. Instead, developers continue to use the well-known constructs and idioms of the programming language to perform database operations. New compilation techniques identify program fragments that can be evaluated by an underlying database system (as opposed to the pro- gramming language runtime system), close to the data. In effect, the database system acts as a capable and scalable co-processor to the language runtime system.In this continuation of the project, we extend database-supported program execution to embrace idiomatic language constructs that are essential in expressing data-intensive algorithms: new generalizations of iterative constructs (generalized monad comprehensions), higher-order functions, or closures. Further, new embedding techniques for database operations aim to remove irritating seams between the host language and the database language.Whenever operations on data items are mutually independent, we now try harder to benefit. New compilation strategies emit vector instructions which can be exploited by modern database kernels, CPUs, and other platforms to perform data-parallel evaluation. At the same time, we modularize and re-organize compiler internals to incorporate new powerful optimizations and to exploit the project in more usage scenarios.