User-defined operators

In recent years, complex data mining and machine learning algorithms have become more common in analytics. Several specialized systems exist to evaluate these on ever-growing sets, which are built efficiently execute different types of analytics queries. However, using various comes at a price. Moving out traditional database is often slow as it requires exporting importing data, typically performed the relatively inefficient CSV format. Additionally, usually offer strong ACID guarantees, lost when adding new, external systems. This disadvantage can be detrimental consistency results. Most scientists still prefer not use classical for The main reason why RDBMS used that SQL difficult work with due its declarative set-oriented nature, easily extensible. We present User-Defined Operators (UDOs) concept include custom into modern query engines. Users write idiomatic code programming language their choice, then directly integrated existing show our implementation compete tools engines while retaining all beneficial properties system.