Struct SnapshotDecoder 

pub struct SnapshotDecoder(/* private fields */);

Decodes a snapshot without retaining its rows or write history. The caller stages records and commits only after finish succeeds. Its oracle must describe the causally closed applied writes recorded by this snapshot. Populate that oracle from section 1. Loss records add no oracle metadata; the consumer validates active losses against merge rows. Only the header, counts and previous record identities are retained between frames. The database consumer checks cross-section completeness, SQL schema rules and app-row visibility; merge’s removal computation checks which removal rules actually hold.

Implementations

impl SnapshotDecoder

pub fn history_frame( &mut self, bytes: &[u8], ) -> Result<Option<SnapshotRecord>, Error>

Extract original slots and claim settlement evidence before the store’s historical signing keys are known. Other sections are counted and skipped; this is not row validation. A loader must subsequently decode the entire authenticated snapshot with frame before installing its data.

pub fn start(frame: &[u8], prefix: &SnapshotObjectPrefix) -> Result<Self, Error>

Read the initial header frame using coverage from the authenticated sealed prefix.

pub fn header(&self) -> &SnapshotHeader

Identity, schema and coverage of this snapshot.

pub fn frame<O: WriteSlots>( &mut self, bytes: &[u8], oracle: &O, ) -> Result<Option<SnapshotRecord>, O::Error>

where O::Error: From<Error>,

Decode a record or end marker. Merge checks row invariants using the supplied oracle. Every error leaves counts and order unchanged.

pub fn finish(&self) -> Result<(), Error>

Refuse truncation, including EOF between records without an end marker.