Struct BatchAccumulator 

struct BatchAccumulator {
    batches: Vec<RecordBatch>,
    rows: usize,
    bytes: usize,
}

Coalesces consecutive ready record batches into one outgoing group.

The budgets are soft limits (flush thresholds) rather than strict admission caps for under-budget batches: push appends a batch before reporting whether to flush, so a group may exceed a budget by up to one batch. A batch that is itself at or over a budget is never appended: it is forwarded as a singleton, flushing the current group first when it was encountered inside a group.

Fields

batches: Vec<RecordBatch>

rows: usize

bytes: usize

Implementations

impl BatchAccumulator

const MAX_ROWS: usize = 4096

const MAX_BYTES: usize

const MAX_BATCHES: usize = 16

fn new() -> Self

fn reaches_budget(rows: usize, bytes: usize) -> bool

Whether a batch already reaches a budget on its own, in which case the stream forwards it as a singleton instead of coalescing it. This applies both to a batch that starts a new group and to a batch encountered inside a group (the accumulated group is flushed first).

fn is_empty(&self) -> bool

fn push(&mut self, batch: RecordBatch) -> bool

Appends batch and reports whether the group should be flushed now.

fn drain(&mut self) -> Vec<RecordBatch>

Removes and returns the accumulated batches in arrival order.

Merging is left to the caller, which owns the stream schema.