Groupby

Groupby, often written as group by in SQL and as groupby in programming libraries, is a data processing operation that partitions a dataset into subgroups based on one or more keys and then applies a calculation to each subgroup. The result is a table of summary values, with one row per group, illustrating metrics such as totals, means, or counts across defined categories.

In SQL, GROUP BY groups rows by the specified expressions. Each distinct value (or combination) of the grouping expressions defines a group, and aggregate functions such as SUM, AVG, COUNT, MIN, and MAX are computed for each group. The result typically includes only the grouped columns and the aggregates, unless a HAVING clause filters groups after aggregation.

In data analysis libraries, groupby constructs manage the grouping logic and support subsequent operations such as aggregation, transformation, or custom functions. Examples include Python’s pandas (df.groupby(keys).agg(...), df.groupby(keys).transform(...)), R’s dplyr (group_by(...) followed by summarise or mutate), and data.table (by=). Grouping can be performed on one or multiple keys and can operate on numeric, string, or date/time data.

Common aggregations include sum, mean, median, count, and standard deviation. Grouping is used for categorical, temporal, or derived keys, enabling cross-group comparisons and subsetting. Performance depends on data size, key cardinality, and memory, with implementations often using hash-based or sort-based strategies and, for very large datasets, streaming or chunked processing.

Related concepts include windowed operations that compute values within a group across a sequence, and the distinction between aggregations (reducing to one value per group) and transforms (producing a value per row within its group).