Module sort 

Sorting Algorithms Kernels Module - Array Sorting and Ordering Operations

Sorting kernels for ordering operations across Arrow-compatible data types with null-aware semantics and optimised comparison strategies. Essential foundation for analytical operations requiring ordered data access and ranking computations.

Regular sorts here alter the actual data. The argsort variants return the indices.

Structs

ArgsortConfig

Configuration for argsort with algorithm selection

Enums

SortAlgorithm

Sort algorithm selection for argsort operations

Functions

argsort

Comparison-based argsort for any Ord type - returns indices that would sort the data

argsort_auto_i32

Argsort for i32 with automatic algorithm selection

argsort_auto_i64

Argsort for i64 with automatic algorithm selection

argsort_auto_u32

Argsort for u32 with automatic algorithm selection

argsort_auto_u64

Argsort for u64 with automatic algorithm selection

argsort_boolean

Argsort for boolean arrays - false sorts before true

argsort_categorical_custom

Argsort for CategoricalArray with custom category order

argsort_categorical_lexical

Argsort for CategoricalArray - lexical ordering by string values

argsort_float

Argsort for floats with proper NaN handling

argsort_radix_i32

LSD Radix argsort for i32 - uses sign-bit flipping for proper signed ordering

argsort_radix_i64

LSD Radix argsort for i64 - uses sign-bit flipping for proper signed ordering

argsort_radix_u32

LSD Radix argsort for u32 - O(n·k) where k=4 (bytes)

argsort_radix_u64

LSD Radix argsort for u64 - O(n·k) where k=8 (bytes)

argsort_str

Argsort for string slices - returns indices that would sort the data lexicographically

argsort_string_array

Argsort for StringArray (offset-based string storage)

sort_boolean_array

Sorts a BooleanArray in-place by value: all false first, then true. Nulls sort first if present.

sort_categorical_lexical

Sorts a CategoricalArray lexically by its unique values, returning new indices and mask. The category dictionary is preserved. Nulls sort first.

sort_float

Total ordering for f32/f64 as per IEEE 754

sort_int

Performs in-place unstable sorting of integer slices with optimal performance.

sort_slice_with_mask

Sorts array data and applies the same permutation to the null mask.

sort_str

Performs in-place unstable sorting of string slice references with lexicographic ordering.

sort_string_array

For StringArray as contiguous utf8 buffer plus offsets. Assumes offsets + values as in minarrow StringArray.

sorted_float

Returns a newly sorted Vec64, leaving the original slice untouched.

sorted_int

Creates a new sorted copy of integer data in a Vec64 container.

sorted_str

Creates a new sorted collection of owned strings from string references.

total_cmp_f

TLDR: Instead of comparing the raw bit patterns directly (which places all negative‐sign bit values after the positives), we transform each bit pattern into a “sort key”: => If the sign bit is 1 (negative or negative‐NaN), we invert all bits: !bits. => Otherwise (sign bit 0), we flip only the sign bit: bits ^ 0x80…0.