Struct Sq8Vector 

pub struct Sq8Vector {
    pub min: f32,
    pub max: f32,
    pub bytes: Vec<u8>,
}

SQ8-quantized vector: every dimension stored as one byte plus a shared (min, max) reconstruction frame.

Reconstruction: x_i ≈ min + (byte_i / 255) * (max - min). Quantization error bound (per element): (max - min) / 510 — half the quantization step (max - min) / 255 thanks to round-to-nearest.

Fields

min: f32

max: f32

bytes: Vec<u8>

Implementations

impl Sq8Vector

pub fn dim(&self) -> usize

Dimension of the original f32 vector (= byte count).

impl Sq8Vector

pub fn to_bytes(&self) -> Vec<u8>

Serialise to the standalone byte format. Always succeeds.

Panics if the dimension exceeds u32::MAX — but DataType::Vector already caps dim at u32 at the type level, so this is a no-op invariant on real inputs.

pub fn from_bytes(input: &[u8]) -> Result<Self, QuantizeError>

Parse the standalone byte format. Strict — body length must equal the declared dim exactly (no extra trailing bytes).

pub const fn encoded_size_for(dim: usize) -> usize

Bytes encoded by to_bytes for an instance of dim dimensions. Handy for buffer pre-sizing on the segment writer side.

Trait Implementations

impl Clone for Sq8Vector

fn clone(&self) -> Sq8Vector

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Sq8Vector

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

Formats the value using the given formatter.

impl PartialEq for Sq8Vector

fn eq(&self, other: &Sq8Vector) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Sq8Vector