Struct DocLenSlot 

pub struct DocLenSlot {
    pub fact: FactId,
    pub len: u16,
    pub distinct: u16,
    pub sig: u64,
}

Per-document record: [fact 4 | len u16 | distinct u16 | sig u64], Uniform arena.

Beyond the length BM25 scores with, the slot carries a summary of the document’s term set: how many distinct terms it has, and one bit per term hashed into a 64-bit word. That summary is what lets the write path bound the term-set overlap of two facts without re-reading and re-tokenizing their texts (see Memory::find_similar). It is written when the document is indexed, where the term set is already in hand, so it costs nothing to produce.

Fields

fact: FactId

The document (fact) id — the key.

len: u16

Token count of the document, saturated at u16::MAX.

distinct: u16

Number of distinct terms, saturated at u16::MAX. Zero means “unknown”: a document indexed before the signature existed, read through the legacy migration.

sig: u64

Union of sig_bit over the document’s distinct terms. A term absent from this word is definitely absent from the document; a term present may still be absent (bits collide). Zero alongside distinct == 0 means “unknown”.

Implementations

impl DocLenSlot

pub fn has_signature(&self) -> bool

Whether the term-set summary is present. A legacy document carries none, and callers must fall back to reading its text.

pub fn overlap_bound(&self, terms: &[u32]) -> usize

An upper bound on how many of terms this document also holds.

Exact in the direction that matters: a term whose bit is clear cannot be in the document, so the true intersection is never larger than the count returned here. Callers use it to rule overlap out.

Trait Implementations

impl Clone for DocLenSlot

fn clone(&self) -> DocLenSlot

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for DocLenSlot

impl Debug for DocLenSlot

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

Formats the value using the given formatter.

impl Eq for DocLenSlot

impl PartialEq for DocLenSlot

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

Equality operator ==.

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

Inequality operator !=.

impl Slot for DocLenSlot

const SIZE: usize = doclen_at::SIZE

Total size of the record in bytes.

const KEY_LEN: usize = doclen_at::KEY_LEN

Length of the key prefix; sorting and lookups compare only these leading bytes. Must not exceed SIZE.

fn write(&self, out: &mut [u8])

Serializes the record into out (out.len() == SIZE), filling every byte.

fn read(bytes: &[u8]) -> Self

Reconstructs the record from bytes (bytes.len() == SIZE).

impl StructuralPartialEq for DocLenSlot