Specify index types-PolarDB(PolarDB)-阿里云帮助中心

GanosBase supports two index types for trajectory data. They differ in how they partition a trajectory into bounding boxes, which determines query filtering precision and storage cost.

Index types

Index type	How it works	Storage	Filtering efficiency
Single bounding box	Represents each trajectory as one bounding box	Lower	Lower
Multi-bounding box	Splits a trajectory into multiple segments, each as a separate bounding box	Higher	Higher — spatial queries are at least twice as fast as queries without bounding box indexing

Single bounding box

Available operator classes: trajgist_2d (two-dimensional), trajgist_3dt (three-dimensional and time).

Multi-bounding box

Operator class: multi.

Multi-bounding box indexing is not supported in some versions of GanosBase.

View the default index type

Run the following queries to check the current default index configuration:

-- List all trajectory-related operator classes and identify the default
SELECT * FROM pg_opclass WHERE opcname LIKE '%traj%' AND opcdefault = true;

-- Show the default split configuration (applies to trajgist_ops_multi)
SHOW ganos.trajectory.index_split_config;

Create and manage indexes

Create a two-dimensional and time index

CREATE INDEX idx_2dt ON trajectory_table USING GiST(traj trajgist_2dt);

Create a multi-bounding box index

Before creating a multi-bounding box index, configure how many bounding boxes represent each trajectory. The following example sets four bounding boxes:

SET ganos.trajectory.index_split_config = '{"cut_point.even_divide":4}';

Then create the index:

CREATE INDEX idx_2dt ON trajectory_table USING GiST(traj multi);

Drop an index

DROP INDEX idx_2dt;