lance_index::scalar::btree

Trait BTreeSubIndex

Source 
pub trait BTreeSubIndex:
    Debug
    + Send
    + Sync
    + DeepSizeOf {
    // Required methods
    fn train<'life0, 'async_trait>(
        &'life0 self,
        batch: RecordBatch,
    ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
    fn load_subindex<'life0, 'async_trait>(
        &'life0 self,
        serialized: RecordBatch,
    ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ScalarIndex>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
    fn retrieve_data<'life0, 'async_trait>(
        &'life0 self,
        serialized: RecordBatch,
    ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
    fn schema(&self) -> &Arc<Schema>;
    fn remap_subindex<'life0, 'life1, 'async_trait>(
        &'life0 self,
        serialized: RecordBatch,
        mapping: &'life1 HashMap<u64, Option<u64>>,
    ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait;
}
Expand description

A trait that must be implemented by anything that wishes to act as a btree subindex

Required Methods§

Source

fn train<'life0, 'async_trait>( &'life0 self, batch: RecordBatch, ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Trains the subindex on a single batch of data and serializes it to Arrow

Source

fn load_subindex<'life0, 'async_trait>( &'life0 self, serialized: RecordBatch, ) -> Pin<Box<dyn Future<Output = Result<Arc<dyn ScalarIndex>>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Deserialize a subindex from Arrow

Source

fn retrieve_data<'life0, 'async_trait>( &'life0 self, serialized: RecordBatch, ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Retrieve the data used to originally train this page

In order to perform an update we need to merge the old data in with the new data which means we need to access the new data. Right now this is convenient for flat indices but we may need to take a different approach if we ever decide to use a sub-index other than flat

Source

fn schema(&self) -> &Arc<Schema>

The schema of the subindex when serialized to Arrow

Source

fn remap_subindex<'life0, 'life1, 'async_trait>( &'life0 self, serialized: RecordBatch, mapping: &'life1 HashMap<u64, Option<u64>>, ) -> Pin<Box<dyn Future<Output = Result<RecordBatch>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Given a serialized page, deserialize it, remap the row ids, and re-serialize it

Implementors§