The FIVM repository contains the implementation of Factorized Incremental View Maintenance (F-IVM) based on the SIGMOD 2018 paper: Incremental View Maintenance with Triple Lock Factorization Benefits.
F-IVM leverages higher-order IVM to reduce the maintenance of the input view to the maintenance of a tree of increasingly simpler views. F-IVM takes a SQL query and a variable order (tree decomposition) for that query as input and produces C++ code optimized for processing batch or single-tuple updates to input relations.
The frontend of F-IVM produces high-level update trigger specifications for the input query. The backend of F-IVM relies on DBToaster , which is extended to support custom ring types, to transform these trigger specifications into native code.
FIVM 1.0 is now released to the public.