Struct BatchWriter 

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

Batch writer for high-throughput operations

Supports two modes:

• Buffered mode (auto_commit=false): All ops buffered, single commit at execute()
• Streaming mode (auto_commit=true): Auto-commits every max_batch_size ops

Implementations

impl<'a> BatchWriter<'a>

pub fn new(conn: &'a ToonConnection) -> Self

Create new batch writer

pub fn max_batch_size(self, size: usize) -> Self

Set maximum batch size

In streaming mode (auto_commit=true), a commit is triggered when the batch reaches this size. Recommended values:

• 100-500: Low latency, more fsyncs
• 1000-5000: Balanced (default)
• 10000+: Maximum throughput, higher latency spikes

pub fn auto_commit(self, enabled: bool) -> Self

Enable auto-commit when batch is full

When enabled, the batch writer will automatically commit transactions when they reach max_batch_size operations. This bounds memory usage to O(max_batch_size) and provides predictable commit latency.

pub fn insert(self, table: &str, values: Vec<(&str, ToonValue)>) -> Self

Add insert operation

pub fn insert_slice( self, table: &str, row_id: u64, values: Vec<Option<ToonValue>>, ) -> Self

Add insert operation using slice-based values (zero-allocation path)

Values must be in schema column order. Use None for NULL values. This is the fastest insert path, matching benchmark performance.

Example

batch.insert_slice("users", 1, vec![
    Some(ToonValue::UInt(1)),
    Some(ToonValue::Text("Alice".into())),
    None, // NULL
])

pub fn update( self, table: &str, key_field: &str, key_value: ToonValue, updates: Vec<(&str, ToonValue)>, ) -> Self

Add update operation

pub fn delete(self, table: &str, key_field: &str, key_value: ToonValue) -> Self

Add delete operation

pub fn pending_count(&self) -> usize

Get number of pending operations (in current unflushed batch)

pub fn total_count(&self) -> usize

Get total operations processed so far (including flushed chunks)

pub fn execute(self) -> Result<BatchResult>

Execute all pending operations

In streaming mode, this commits any remaining operations in the final partial batch. Returns cumulative results from all chunks.