Struct MaxSim 

pub struct MaxSim<'a> { /* private fields */ }

Computes per-query-vector maximum similarities to document vectors.

For each query vector qᵢ, finds the maximum similarity (minimum negated inner product) to any document vector:

scores[i] = minⱼ -IP(qᵢ, dⱼ)

Implements DistanceFnMut for various matrix types (e.g., MatRef<Standard<f32>>).

Usage

• Create with MaxSim::new, providing a mutable scores buffer.
• Call DistanceFnMut::evaluate with query and document matrices.
• Read results from the scores buffer.

Implementations

impl<'a> MaxSim<'a>

pub fn new(scores: &'a mut [f32]) -> Result<Self, MaxSimError>

Creates a new MaxSim with the provided scores buffer.

Errors

Returns an error if scores is empty.

pub fn size(&self) -> usize

Returns the number of score slots in the buffer.

pub fn get(&self, i: usize) -> Result<f32, MaxSimError>

Returns the score at index i.

pub fn set(&mut self, i: usize, x: f32) -> Result<(), MaxSimError>

Sets the score at index i.

pub fn scores_mut(&mut self) -> &mut [f32]

Returns a mutable reference to the internal buffer of scores.

This is useful for implementations external to crate as well as optimized implementations to access the buffer if needed.

Trait Implementations

impl<'a> Debug for MaxSim<'a>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<const NBITS: usize> DistanceFunctionMut<QueryMatRef<'_, MinMaxMeta<NBITS>>, MatRef<'_, MinMaxMeta<NBITS>>> for MaxSim<'_>

where Unsigned: Representation<NBITS>, InnerProduct: for<'x, 'y> PureDistanceFunction<BitSlice<'x, NBITS, Unsigned>, BitSlice<'y, NBITS, Unsigned>, MathematicalResult<u32>>,

fn evaluate( &mut self, query: QueryMatRef<'_, MinMaxMeta<NBITS>>, doc: MatRef<'_, MinMaxMeta<NBITS>>, )

Compute distances between the left and right arguments.

impl<T: Copy> DistanceFunctionMut<QueryMatRef<'_, Standard<T>>, MatRef<'_, Standard<T>>, Result<(), MaxSimError>> for MaxSim<'_>

where InnerProduct: for<'a, 'b> PureDistanceFunction<&'a [T], &'b [T], f32>,

fn evaluate( &mut self, query: QueryMatRef<'_, Standard<T>>, doc: MatRef<'_, Standard<T>>, ) -> Result<(), MaxSimError>

Compute distances between the left and right arguments.