Struct Handle 

pub struct Handle { /* private fields */ }

Implementations

impl Handle

pub async fn open(location: &Url) -> Result<Self>

Open without storage options or explicit cache caps. Backend-aware defaults from [runtime] apply.

pub async fn open_with_options( location: &Url, storage_options: HashMap<String, String>, caps: RuntimeCaps, ) -> Result<Self>

Open with object-store options handed through to Lance verbatim, plus the resolved [runtime] cache caps. Object-store keys are the object_store crate’s standard config names; pond does not parse them. Opening datasets never performs index work; index lifecycle lives under Handle::optimize_table. parts.lance opens lazily on first use.

pub fn location(&self) -> &Url

pub fn storage_options(&self) -> &HashMap<String, String>

Read-only view of the storage_options the handle was opened with. pond status needs them to instantiate a raw object_store client that can LIST the remote bucket for sizing.

pub async fn row_counts(&self) -> Result<(usize, usize, usize)>

pub async fn optimize_table( &self, table: Table, intents: &[IndexIntent], progress: Option<&OptimizeProgressFn>, policy: &MaintenancePolicy, ) -> TableOptimizeOutcome

Run the table-local maintenance cycle for the supplied index intents. BTree is rebuilt from scratch to dodge Lance v7.0.0-beta.16’s flat BTree combine bug; Bitmap, FTS, and IVF_PQ fold via append.

spec.md#substrate 3.7 (lance-index-maintenance): indices and compaction commit independently and use independent retry budgets, so a hot writer that starves compaction (Rewrite) does not abort the index build (Update) the operator actually asked for.

pub async fn optimize_table_indices_only( &self, table: Table, intents: &[IndexIntent], progress: Option<&OptimizeProgressFn>, ) -> PhaseOutcome

Run only the indices phase for one table. Used by pond embed’s tail to fold newly written vectors into the indices without paying the compaction retry budget while embed itself may still be writing.

pub async fn rebuild_index( &self, table: Table, intent: &IndexIntent, ) -> Result<()>

pub async fn index_status( &self, table: Table, intents: &[IndexIntent], ) -> Result<Vec<IndexStatus>>

pub async fn scan(&self, table: Table, opts: ScanOpts<'_>) -> Result<Scanner>

Single read entry point: prefilter via predicate, optionally project, return the prepared Scanner (spec.md#lance-chokepoints-read).

pub async fn count_rows(&self, table: Table) -> Result<usize>

pub async fn initialized(&self) -> Result<bool>

Whether the store holds synced data yet. open eagerly creates empty sessions/messages datasets, but parts opens lazily on first write (see open_with_options), so its presence is the “has been synced” signal - letting read-only surfaces (pond status, pond storage) render an empty state instead of erroring on the first parts describe.

pub async fn table_sizes(&self) -> Result<TableSizes>

On-disk byte totals for the three datasets plus the data-dir remainder. Every byte is sized by listing through Lance’s object store (spec.md#lance-chokepoints-storage), identical for file:// and s3://.

Trait Implementations

impl Debug for Handle

fn fmt(&self, formatter: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.