onesparse

Overview

PG18 only; upstream release v1.0.0 ships extension SQL version 0.1.0

Version

Build

You can build the RPM / DEB packages for one_sparse using pig build:

pig build pkg one_sparse         # build RPM / DEB packages

Install

You can install one_sparse directly. First, make sure the PGDG and PIGSTY repositories are added and enabled:

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

Install the extension using pig or apt/yum/dnf:

• Install
• pig
• dnf
• apt

pig install one_sparse;          # Install for current active PG version

pig ext install -y one_sparse -v 18  # PG 18

dnf install -y onesparse_18       # PG 18

apt install -y postgresql-18-onesparse   # PG 18

Create Extension:

CREATE EXTENSION onesparse;

Usage

Sources: homepage, intro docs, Matrix, Vector, and Algorithms.

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.

Core Setup

CREATE EXTENSION onesparse;
SET search_path TO public,onesparse;

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

The intro docs note that OneSparse keeps its API in the onesparse schema, and the matrix/vector pages show the same search_path pattern for interactive use.

Matrix and Vector

The matrix page shows common operations such as constructing, printing, drawing, resizing, casting, and aggregating matrices. The vector page shows the matching vector API, including nvals, size, set_element, get_element, 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 getting-started docs use graph examples built from Matrix Market files and random graphs. They highlight these algorithms:

• 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 also shows graph loading with mmread(...) and graph visualization with draw(...).

Scope

The documentation set is broad. This stub captures the core interface and the main examples that repeat across the intro, matrix, vector, and algorithms pages, without reproducing the full GraphBLAS catalog.