Struct BinaryCodec 

pub struct BinaryCodec {
    pub dim: usize,
}

Binary quantization codec (sign-bit encoding, Hamming distance).

Stores only dim — no learned parameters. Suitable as a coarse pre-filter before exact reranking with a higher-fidelity codec.

Fields

dim: usize

Trait Implementations

impl VectorCodec for BinaryCodec

type Query = Vec<u8>

Pre-encoded query bits (ceil(dim/8) bytes).

fn encode(&self, v: &[f32]) -> Self::Quantized

Encode an FP32 vector into binary sign bits.

Panics

UnifiedQuantizedVector::new fails only on outlier-count/bitmask mismatches. With outlier_bitmask = 0 and an empty outliers slice this can never happen. The expect is therefore unreachable in practice.

fn prepare_query(&self, q: &[f32]) -> Self::Query

Pre-encode the query to binary sign bits for fast Hamming comparison.

fn adc_lut(&self, _q: &Self::Query) -> Option<AdcLut>

Binary codec has no ADC table — returns None.

fn fast_symmetric_distance( &self, q: &Self::Quantized, v: &Self::Quantized, ) -> f32

Symmetric Hamming distance between two binary-encoded vectors.

fn exact_asymmetric_distance(&self, q: &Self::Query, v: &Self::Quantized) -> f32

Asymmetric Hamming distance: pre-encoded query bits vs stored candidate.

type Quantized = BinaryQuantized

The packed quantized form. Must be convertible to a UnifiedQuantizedVector reference via AsRef.

fn train(&mut self, samples: &[&[f32]]) -> Result<(), CodecError>

Optional: fit the codec’s learned parameters on a set of training vectors.