Struct KvSchema 

pub struct KvSchema { /* private fields */ }

Implementations

impl KvSchema

pub fn new(client: StoreClient) -> Self

pub fn table( self, name: impl Into<String>, columns: Vec<TableColumnConfig>, primary_key_columns: Vec<String>, index_specs: Vec<IndexSpec>, ) -> Result<Self, String>

pub fn orders_table( self, table_name: impl Into<String>, index_specs: Vec<IndexSpec>, ) -> Result<Self, String>

pub fn table_versioned( self, name: impl Into<String>, columns: Vec<TableColumnConfig>, entity_column: impl Into<String>, version_column: impl Into<String>, index_specs: Vec<IndexSpec>, ) -> Result<Self, String>

Create a table with a versioned composite primary key.

The entity column and version column (UInt64) together form the composite primary key. The entity can be any supported primary-key type, including variable-length logical keys encoded through the crate’s ordered variable-length Utf8 mapping.

Versions sort numerically via big-endian encoding, so a reverse range scan from (entity, V) downward with LIMIT 1 yields the latest version <= V. See examples/versioned_kv.rs for the basic query pattern plus an immutable-friendly companion watermark table pattern for out-of-order batch uploads.

pub fn table_count(&self) -> usize

pub fn register_all(self, ctx: &SessionContext) -> DataFusionResult<()>

pub fn batch_writer(&self) -> BatchWriter

pub async fn backfill_added_indexes( &self, table_name: &str, previous_index_specs: &[IndexSpec], ) -> DataFusionResult<IndexBackfillReport>

Backfill secondary index entries after adding new index specs.

previous_index_specs must represent the index list used when existing rows were written. The current schema’s index list must be an append-only extension of that list (same order/layout for existing indexes, with new indexes only added at the tail).

Operational ordering requirement: start writing new rows with the new index specs before backfilling historical rows, or rows written during the backfill window may be missing from the new index.

pub async fn backfill_added_indexes_with_options( &self, table_name: &str, previous_index_specs: &[IndexSpec], options: IndexBackfillOptions, ) -> DataFusionResult<IndexBackfillReport>

Backfill secondary index entries after adding new index specs, with configurable row page size for the full-scan read.

pub async fn backfill_added_indexes_with_options_and_progress( &self, table_name: &str, previous_index_specs: &[IndexSpec], options: IndexBackfillOptions, progress_tx: Option<&UnboundedSender<IndexBackfillEvent>>, ) -> DataFusionResult<IndexBackfillReport>

Backfill secondary index entries after adding new index specs, with configurable row page size for the full-scan read and an optional progress event channel.

Progress events are emitted only after buffered ingest writes for the reported cursor are flushed, so Progress.next_cursor can be persisted and used to resume later.