Columnar Storage Is Normalization

7 hours ago

Columnar storage transforms row-oriented data into column-oriented arrays, where each column is stored separately.
Row-oriented storage stores all columns of a row together, making row insertion and retrieval efficient.
Columnar storage optimizes for analytical queries by allowing efficient access to specific columns without reading entire rows.
The process of columnarization can be viewed as an extreme form of database normalization, akin to splitting a table into separate tables with a primary key.
Reconstructing a row from columnar data is analogous to performing a join on the implied ordinal position (like a primary key).
This perspective unifies traditional query operations (like projections and joins) with data format manipulation, though it often remains an implementation detail.

Hasty Briefsbeta