columnwise

Columnwise refers to operations, storage, or organization that are oriented along the columns of a two-dimensional array or table. In data analysis and programming, columnwise processing means applying an operation independently to each column of a matrix or dataframe. This contrasts with rowwise processing, where the operation is applied across rows. Columnwise operations are common when features correspond to columns and observations to rows, such as computing per-column means, standard deviations, normalizing by column, or performing columnwise aggregations. Many languages and libraries expose this through functions that operate along columns, for example by specifying a columnwise margin or axis parameter, or by applying a function to each column of a data structure.

In computer memory, column-major order—where consecutive elements within a column are stored contiguously—affects how columnwise access patterns perform. Column-major storage is used by languages and environments such as Fortran, MATLAB, and R, whereas row-major order is typical for C and C++. The choice of storage order can influence cache locality and performance for columnwise versus rowwise algorithms.

The term is also used informally to describe columnwise formatting in tabular data or to describe column-oriented databases and columnar storage systems, which organize data by column to improve read efficiency for analytics workloads. Overall, columnwise describes a perspective or approach focused on columns rather than rows.