Struct DmlOptimizer 

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

DML cost-based optimizer

Provides optimization decisions for DML operations based on cost estimation.

Implementations

impl<'a> DmlOptimizer<'a>

pub fn new(db: &'a Database, table_name: &'a str) -> Self

Create a new DML optimizer for a table

Arguments

• db - Database reference for metadata lookups
• table_name - Name of the table being operated on

Returns

DmlOptimizer instance with cost estimation configured

pub fn get_stats_with_fallback(&self) -> TableStatistics

Get or compute table statistics with fallback for missing stats

When actual statistics are unavailable (new table, no ANALYZE run), this creates fallback statistics based on available metadata.

pub fn optimal_insert_batch_size(&self, total_rows: usize) -> usize

Determine optimal batch size for INSERT operations

Uses cost estimation to decide between small and large batch sizes. High-cost tables (many indexes) benefit from smaller batches to avoid index maintenance overhead accumulating.

Arguments

• total_rows - Total number of rows to insert

Returns

Recommended batch size

pub fn should_chunk_delete(&self, rows_to_delete: usize) -> bool

Determine if DELETE should be chunked to avoid long locks

Large deletes on high-cost tables can cause long pauses due to index maintenance. Chunking allows other operations to proceed between chunks.

Arguments

• rows_to_delete - Number of rows to be deleted

Returns

true if delete should be chunked, false for single operation

pub fn delete_chunk_size(&self) -> usize

Get recommended chunk size for chunked deletes

pub fn should_trigger_early_compaction(&self) -> bool

Check if early compaction should be triggered

Based on the current deleted ratio, determines if compaction should be triggered sooner than the default 50% threshold.

Returns

true if early compaction is recommended

pub fn compute_indexes_affected_ratio( &self, changed_columns: &HashSet<usize>, schema: &TableSchema, ) -> f64

Compute the ratio of indexes affected by an UPDATE operation

This is used for UPDATE cost estimation. If the update only touches non-indexed columns, the ratio is 0.0 and index maintenance is skipped.

Arguments

• changed_columns - Set of column indices being modified
• schema - Table schema for looking up index column info

Returns

Ratio of indexes affected (0.0 to 1.0)

pub fn estimate_update_cost( &self, row_count: usize, indexes_affected_ratio: f64, ) -> f64

Estimate the cost of an UPDATE operation

Arguments

• row_count - Number of rows to update
• indexes_affected_ratio - Ratio of indexes affected (from compute_indexes_affected_ratio)

Returns

Estimated cost in arbitrary units

pub fn index_info(&self) -> Option<&TableIndexInfo>

Get the current table index info (for external use)