istore

istore : Integer key/value types with arithmetic, aggregation, and GIN indexing

Overview

Packages

Source

pig build pkg istore;		# build rpm/deb

Install

Make sure PGDG and PIGSTY repo available:

pig repo add pgsql -u   # add both repo and update cache

Install this extension with pig:

pig install istore;		# install via package name, for the active PG version

pig install istore -v 18;   # install for PG 18
pig install istore -v 17;   # install for PG 17
pig install istore -v 16;   # install for PG 16
pig install istore -v 15;   # install for PG 15
pig install istore -v 14;   # install for PG 14

Create this extension with:

CREATE EXTENSION istore;

Usage

Sources:

• Official README
• Official extension SQL
• Official extension control file

istore provides compact integer-key maps for PostgreSQL. The istore type stores integer values, while bigistore stores bigint values; both are useful for sparse counters, distributions, and analytical aggregates keyed by integer identifiers.

Core Workflow

Text input resembles hstore, and array constructors can aggregate repeated keys:

CREATE EXTENSION istore;

SELECT istore(ARRAY[1, 2, 1, 3, 2, 2]);
SELECT '1=>4,2=>5'::istore -> 1;
SELECT '1=>4,2=>5'::istore + '1=>4,3=>6'::istore;

CREATE INDEX metrics_keys_idx
  ON metrics USING gin (values istore_key_ops);

Key lookup uses ->; ?, ?&, and ?| test key presence. || concatenates stores, while arithmetic operators combine matching keys. The istore_key_ops GIN operator class accelerates key-existence predicates.

Important Objects and Caveats

Constructors and transforms include istore(...), bigistore(...), akeys, avals, each, slice, compact, fill_gaps, accumulate, clamp, and delete helpers. sum_up summarizes values, and aggregate SUM returns bigistore so that totals have a wider value type. Per-key minimum and maximum aggregates are also provided.

The cast from istore to bigistore is implicit; narrowing assignments can overflow. Arithmetic and division still require application-level checks for overflow and zero divisors. Large values can be TOASTed, and GIN indexes add write and maintenance cost, so benchmark the actual distribution and update workload instead of treating upstream examples as performance guarantees.