What is a GiST index used for? (Geometric data, range types, full-text.)

GiST — Generalized Search Tree — is a balanced-tree framework for data that has no single linear order: geometry, range types, and nearest-neighbor search. B-tree assumes "this key is less than that key." GiST drops that assumption: each internal node stores a predicate (a bounding box, a range, a region) that's true for everything below it, so a search prunes whole subtrees by asking "could a match be in here?"

What it's for

Geometric / spatial data. "Which shapes overlap this box?" is meaningless to a B-tree but natural to GiST's bounding-box pruning. This is the foundation PostGIS builds spatial indexes on.

CREATE INDEX ON shapes USING gist (geom);
SELECT * FROM shapes WHERE geom && box '((0,0),(10,10))';  -- overlaps

Range types — and exclusion constraints. Indexing tstzrange, int4range, etc. for overlap (&&) and containment. The standout use is preventing overlapping bookings:

CREATE TABLE bookings (
  room_id int,
  during  tstzrange,
  EXCLUDE USING gist (room_id WITH =, during WITH &&)
);
-- two reservations for the same room with overlapping times are rejected

Nearest-neighbor (KNN). GiST can answer ordered "closest to" queries directly via the distance operator <->, returning rows in distance order without a sort.

SELECT * FROM places ORDER BY location <-> point '(40.7,-74.0)' LIMIT 5;

Full-text search, as the alternative to GIN.

GiST vs GIN for full-text

Both index tsvector, and the trade-off is a classic interview follow-up:

GIN  → faster searches, slower to build/update, larger   → static/read-heavy text
GiST → faster updates, smaller, but lossy (rechecks rows) → churning/smaller corpora

GIN is the usual default for full-text because reads dominate; GiST shines when the indexed text changes constantly or when you also need KNN/geometry that GIN simply can't do.