Schedule - PostgreSQL Development Conference 2024

Multi-Entry Generalized Search Trees

Date: 2024-05-29
Time: 13:30–14:20
Room: Sauder(2270)
Level: Intermediate

Many complex data types cannot be efficiently indexed using the existing PostgreSQL indices, due to the fact that indexing frameworks represent a tuple as a single index entry. The only exception, GIN, is restricted to indexing 1-dimensional data.

In this talk, we present MGiST and MSP-GiST, multi-entry counterparts of the well-known GiST and SP-GiST frameworks. Through the use of an additional external method, each tuple can be split into multiple index entries before indexing. This allows the index to hold a more accurate representation of the indexed tuples. Examples of data types that can make use of this are: multi-ranges, geometry collections (multi-points, multi-polygons), line strings, time-series, spatio-temporal trajectories and more.

MGiST and MSP-GiST are implemented as PostgreSQL extensions and we detail their use for indexing trajectories in MobilityDB.


Maxime Schoemans