onesparse

one_sparse : Sparse linear algebra and graph extension for PostgreSQL 18

Overview

Packages

Source

pig build pkg one_sparse;		# 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 one_sparse;		# install via package name, for the active PG version
pig install onesparse;		# install by extension name, for the current active PG version

pig install onesparse -v 18;   # install for PG 18

Create this extension with:

CREATE EXTENSION onesparse;

Usage

Sources: homepage, release v1.0.0, control file at v1.0.0, intro docs, matrix docs, vector docs, algorithm examples

OneSparse is a PostgreSQL extension that binds SuiteSparse:GraphBLAS into Postgres and exposes sparse linear algebra and graph algorithms as new types, functions, and operators. The docs treat matrix as the core type, with vector and scalar built on top of the same model. The v1.0.0 release exists, while the extension control file at that tag still declares SQL default_version = '0.1.0'.

Core Setup

CREATE EXTENSION onesparse;
SET search_path TO public,onesparse;

SELECT 'int32'::matrix;
SELECT 'int32'::vector;
SELECT 'int32:42'::scalar;

The docs site organizes the API around matrix, vector, and scalar, with interactive examples using casts and constructors.

Matrix and Vector

The matrix page shows common operations such as construction, print(), draw(), assignment, extraction, cast_to(), resize, and aggregation. The vector page documents the matching vector API including nvals(), size(), eadd(), emult(), reduce_scalar(), choose(), and apply().

SELECT print('int32(4:4)'::matrix);
SELECT draw('int32(4:4)[1:2:1 2:3:2 3:1:3]'::matrix);
SELECT eadd('int32[0:1 1:2 2:3]'::vector, 'int32[0:1 1:2 2:3]'::vector, 'plus_int32');
SELECT reduce_scalar('int32[0:1 1:2 2:3]'::vector, 'plus_monoid_int32');

Graph Algorithms

The examples page uses Matrix Market input and graph visualization with draw(...). The documented graph algorithms include:

• bfs(graph, 1) for level and parent BFS
• sssp(cast_to(graph, 'int32'), 1::bigint, 1) for single-source shortest path
• pagerank(graph) for ranking vertices by link structure
• triangle_centrality(graph) for triangle-based centrality
• betweenness(graph, ARRAY[...]) and square_clustering(graph) for additional graph analysis

Representative example from the docs:

SELECT draw(triu(graph), (SELECT level FROM bfs(graph, 1)), false, false, true, 0.5)
FROM karate;

The same guide shows loading a graph with mmread('/home/postgres/onesparse/demo/karate.mtx').