Class Insert
HNSW constructs a multi-layer graph, where each layer is a subset of the one below it. The top layers serve as fast entry points to navigate the graph, while the bottom layer contains all the data points. This structure allows for logarithmic-time complexity for search operations, making it suitable for large-scale, high-dimensional datasets.
The entry point for any interactions with the HNSW data structure is implemented in HNSW. Do not instantiate
this class directly.
-
Constructor Summary
Constructors -
Method Summary
Modifier and TypeMethodDescriptionGet the configuration of this hnsw.Get the executor used by this hnsw.Get the on-read listener.Get the on-write listener.Gets the subspace associated with this object.insert(Transaction transaction, Tuple newPrimaryKey, RealVector newVector) Inserts a new vector with its associated primary key into the HNSW graph.
-
Constructor Details
-
Insert
This constructor initializes a new insert operations object with the necessary components for storage, execution, configuration, and event handling.- Parameters:
locator- theLocatorwhere the graph data is stored, which config to use, which executor to use, etc.
-
-
Method Details
-
getLocator
-
getSubspace
Gets the subspace associated with this object.- Returns:
- the non-null subspace
-
getExecutor
Get the executor used by this hnsw.- Returns:
- executor used when running asynchronous tasks
-
getConfig
Get the configuration of this hnsw.- Returns:
- hnsw configuration
-
getOnWriteListener
Get the on-write listener.- Returns:
- the on-write listener
-
getOnReadListener
Get the on-read listener.- Returns:
- the on-read listener
-
insert
@Nonnull public CompletableFuture<Void> insert(@Nonnull Transaction transaction, @Nonnull Tuple newPrimaryKey, @Nonnull RealVector newVector) Inserts a new vector with its associated primary key into the HNSW graph.The method first determines a layer for the new node, called the
top layer. It then traverses the graph from the entry point downwards, greedily searching for the nearest neighbors to thenewVectorat each layer. This search identifies the optimal connection points for the new node.Once the nearest neighbors are found, the new node is linked into the graph structure at all layers up to its
top layer. Special handling is included for inserting the first-ever node into the graph or when a new node's layer is higher than any existing node, which updates the graph's entry point. All operations are performed asynchronously.- Parameters:
transaction- theTransactioncontext for all database operationsnewPrimaryKey- the uniqueTupleprimary key for the new node being insertednewVector- theRealVectordata to be inserted into the graph- Returns:
- a
CompletableFuturethat completes when the insertion operation is finished
-