Struct BulkWriter 

pub struct BulkWriter { /* private fields */ }

Bulk writer for high-throughput data ingestion.

Buffers vertices and edges, deferring index updates until commit. Supports constraint validation, automatic checkpointing when buffer limits are exceeded, and proper rollback via LanceDB version tracking.

Use abort() to discard uncommitted changes and roll back storage to its pre-bulk-load state.

Implementations

impl BulkWriter

pub fn stats(&self) -> &BulkStats

Returns a snapshot of the current bulk load statistics. Updated after each batch flush.

pub fn touched_labels(&self) -> Vec<String>

Returns the set of vertex labels that have been written to.

pub fn touched_edge_types(&self) -> Vec<String>

Returns the set of edge types that have been written to.

pub async fn insert_vertices( &mut self, label: &str, vertices: impl IntoArrow, ) -> Result<Vec<Vid>>

Insert vertices in bulk.

The vertices are buffered until batch_size is reached, then written to storage. When constraint validation is enabled, constraints are checked before each flush. When the buffer size exceeds max_buffer_size_bytes, a checkpoint is triggered.

Errors

Returns an error if:

• The label is not found in the schema
• Constraint validation fails (when enabled)
• Storage write fails

pub async fn insert_edges( &mut self, edge_type: &str, edges: Vec<EdgeData>, ) -> Result<Vec<Eid>>

Insert edges in bulk.

Edges are buffered until batch_size is reached, then written to storage. When the buffer size exceeds max_buffer_size_bytes, a checkpoint is triggered. Indexes are NOT updated during these writes.

Errors

Returns an error if the edge type is not found in the schema or if storage write fails.

pub async fn commit(self) -> Result<BulkStats>

Commit all pending data and rebuild indexes.

Flushes remaining buffered data, rebuilds deferred indexes, and updates the snapshot manifest.

Errors

Returns an error if flushing, index rebuilding, or snapshot update fails.

pub async fn abort(self) -> Result<()>

Abort bulk loading and roll back all changes.

Rolls back LanceDB tables to their pre-bulk-load versions using LanceDB’s version API. Tables created during the bulk load are dropped. Buffered data that hasn’t been flushed is discarded.

Errors

Returns an error if rollback fails. The error message includes details about which tables failed to roll back.