Struct SparseBlock 

pub struct SparseBlock {
    pub header: BlockHeader,
    pub doc_ids_data: Vec<u8>,
    pub ordinals_data: Vec<u8>,
    pub weights_data: Vec<u8>,
}

Fields

header: BlockHeader

doc_ids_data: Vec<u8>

ordinals_data: Vec<u8>

weights_data: Vec<u8>

Implementations

impl SparseBlock

pub fn from_postings( postings: &[(DocId, u16, f32)], weight_quant: WeightQuantization, ) -> Result<Self>

pub fn decode_doc_ids(&self) -> Vec<DocId>

pub fn decode_doc_ids_into(&self, out: &mut Vec<DocId>)

Decode doc IDs into an existing Vec (avoids allocation on reuse).

pub fn decode_ordinals(&self) -> Vec<u16>

pub fn decode_ordinals_into(&self, out: &mut Vec<u16>)

Decode ordinals into an existing Vec (avoids allocation on reuse).

pub fn decode_weights(&self) -> Vec<f32>

pub fn decode_weights_into(&self, out: &mut Vec<f32>)

Decode weights into an existing Vec (avoids allocation on reuse).

pub fn write<W: Write>(&self, w: &mut W) -> Result<()>

pub fn read<R: Read>(r: &mut R) -> Result<Self>

pub fn from_owned_bytes(data: OwnedBytes) -> Result<Self>

Zero-copy constructor from OwnedBytes (mmap-backed).

Parses the block header and sub-block length prefix, then slices the OwnedBytes into doc_ids/ordinals/weights without any heap allocation. The returned Vecs share the underlying mmap Arc via .to_vec() on the small sub-slices (typically < 1KB each).

pub fn with_doc_offset(&self, doc_offset: u32) -> Self

Create a copy of this block with first_doc_id adjusted by offset.

This is used during merge to remap doc_ids from different segments. Only the first_doc_id needs adjustment - deltas within the block remain unchanged since they’re relative to the previous doc.

Trait Implementations

impl Clone for SparseBlock

fn clone(&self) -> SparseBlock

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseBlock

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

Formats the value using the given formatter.