dense_rank_int

simd_kernels::kernels::window

Function dense_rank_int 

Source 
pub fn dense_rank_int<T: Ord + Copy>(
    window: IntegerAVT<'_, T>,
) -> Result<IntegerArray<i32>, KernelError>
Expand description

Computes SQL DENSE_RANK() ranking for integer data with consecutive rank assignment.

Assigns consecutive rank values where tied values receive identical ranks, implementing SQL DENSE_RANK() semantics. Unlike standard ranking, eliminates gaps in rank sequence when ties occur, providing compact rank numbering for analytical applications.

§Parameters

  • window - Integer array view containing values for dense ranking

§Returns

Returns Result<IntegerArray<i32>, KernelError> containing:

  • Success: Dense rank values with no gaps in sequence
  • Error: KernelError if capacity validation fails
  • Zero values for null elements
  • Null mask indicating positions with valid ranks

§Dense Ranking Semantics

  • DENSE_RANK() behaviour: Tied values receive same rank, next rank is consecutive
  • No rank gaps: Eliminates gaps that occur in standard RANK() function
  • Unique value counting: Essentially counts distinct values in sorted order
  • Ascending order: Smaller integer values receive lower (better) ranks

§Comparison with RANK()

  • RANK(): [1, 2, 2, 4] for values [10, 20, 20, 30]
  • DENSE_RANK(): [1, 2, 2, 3] for values [10, 20, 20, 30]

§Use Cases

  • Analytical queries: SQL DENSE_RANK() window function implementation
  • Categorical ranking: Creating compact categorical orderings
  • Percentile calculation: Foundation for percentile computations without gaps
  • Data binning: Assigning data points to consecutive bins based on value

§Examples

ⓘ
use minarrow::IntegerArray;
use simd_kernels::kernels::window::dense_rank_int;

let arr = IntegerArray::<i32>::from_slice(&[10, 30, 20, 30]);
let result = dense_rank_int((&arr, 0, arr.len())).unwrap();
// Output: [1, 3, 2, 3] - dense ranking with tied values