Abstract: A method for database processing includes allocating primary, secondary and tertiary data structures for each column. Columnar content is written into the primary data structure. Updated values are written into the tertiary data structure. Storage keys indicate correspondence to the primary data structure. Query responses involve first reading updated values from the tertiary data structure, sorting by storage keys, storing into the secondary data structure, and clearing the tertiary data structure. Then the primary and secondary data structures are consulted to respond to the query, reading the value for each row from the secondary data structure where present, otherwise from the primary data structure. A commit phase produces a new version of the primary data structure by merging the secondary data structure and an old version of the primary data structure, consulting them in a similar manner.