Struct CowRewriteBuilder 

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

Builder for orchestrating copy-on-write data file rewrites.

Implementations

impl<'a> CowRewriteBuilder<'a>

pub fn new(table: &'a Table) -> Self

Creates a copy-on-write rewrite builder for table.

pub fn with_predicate(self, predicate: Predicate) -> Self

Sets the row predicate used to plan candidate files.

pub fn with_snapshot_id(self, snapshot_id: i64) -> Self

Sets the snapshot id used to plan candidate files.

pub fn with_batch_size(self, batch_size: usize) -> Self

Sets the Arrow reader batch size.

pub fn with_case_sensitive(self, case_sensitive: bool) -> Self

Sets the case sensitivity used to bind the planning predicate.

pub fn with_rewriter(self, rewriter: Arc<dyn CowBatchRewriter>) -> Self

Sets the record batch rewriter.

pub async fn rewrite(self) -> Result<CowRewriteResult>

Plans, reads, rewrites, and writes replacement data files.

Memory usage

Batches read before a file’s first changed batch are buffered so that no replacement file is emitted for a source file that turns out unchanged. A file that never changes — or whose first change sits at its very end — therefore buffers its entire decoded contents in memory. Files are processed sequentially, so peak usage is one decoded source file at a time, but that can still be several GB for a compaction-sized file. A size-capped fallback that starts writing the replacement once the buffer crosses a threshold is left for follow-up work.