Struct InMemoryCoverFetch 

pub struct InMemoryCoverFetch<'a> { /* private fields */ }

§6E-C / Task #24.2 — in-memory CoverFetch adapter.

Wraps a pre-materialized (cover_bits, costs) pair and slices on-demand per segment. Suitable for orchestrator integrations (Task #24.3) where the encoder’s Pass-1 cover already lives in memory and the streaming-Viterbi memory savings come from the STC-internal O(√n) checkpoint + back-pointer working set, not the cover side.

Memory bound: O(n) cover (caller-provided) + O(K × w) per segment fetch + O(num_segments × num_states × 8 B) checkpoints inside stc_embed_streaming_segmented. For long-clip video stego where the encoder’s full per-GOP cover materialization is itself the OOM source, a follow-on per-GOP-replay adapter (left as v1.1+ work) replaces the cover materialization with repeated Pass 1 invocations bounded by the segment’s GOP range. That follow-on requires encoder restartability at arbitrary GOP boundaries — out of scope for v1.0 (mobile clips are short enough that in-memory cover fits).

Implementations

impl<'a> InMemoryCoverFetch<'a>

pub fn new( cover_bits: &'a [u8], costs: &'a [f32], m: usize, w: usize, segment_size_in_blocks: usize, ) -> Option<Self>

Construct a new in-memory cover fetcher. cover_bits.len() must equal costs.len() and must equal m * w. Returns None on length-mismatch.

Trait Implementations

impl<'a> CoverFetch for InMemoryCoverFetch<'a>

fn total_positions(&self) -> usize

Total cover position count n.

fn num_segments(&self) -> usize

Number of segments. Equals m.div_ceil(K) where K is the segment size.

fn segment_size_in_blocks(&self) -> usize

Segment size in message blocks (constant across segments except possibly the last). The last segment may have fewer blocks if m doesn’t divide evenly by K.

fn fetch_segment(&mut self, seg_idx: usize) -> (Vec<u8>, Vec<f32>)

Fetch one segment’s cover bits and costs.