Load a real graph offline, or browse the frozen package catalogue. No network, Python, or external viewer is needed. The catalogue includes the curated comparison gallery and all six examples used by Zheng, Pawar and Goodman. Gallery membership and original-six membership are separate columns.
suitesparse.graph(name)suitesparse.graphs()
`suitesparse.graphs()` returns a data frame with `id`, `name`, `group`, `vertices`, `edges`, `conversion_type`, `components`, `original_vertices`, `retained_fraction`, `gallery`, and `original_six`.
`suitesparse.graph()` returns a list with `n` (including isolated vertices), `edges` (a two-column integer matrix of unique, lexicographically sorted, one-based undirected edges), `edge_weights` (unit traversal lengths), `vertex_data`, and `provenance`. Vertex data contain consecutive `vertex`, one-based `source_index`, `label`, `source_partition` (`row` or `column`), `original_vertex` (one-based index before component extraction), and one-based `component`, `vertex_id` (saved graph identity), and `source_vertex_id` (identity before extraction). In bipartite graphs, source indices refer separately to matrix rows and columns. Les Miserables retains character names.
Provenance includes source and download URLs, source matrix headers with original credits, CC BY 4.0 licensing, SHA-256 checksums, conversion recipe, original graph identity, and component retention information. The original six also retain their earlier metadata under `original_example`.
Full graph ID (for example `"HB/494_bus"`) or a unique short name (`"494_bus"`). Names are case-sensitive. Component-derived IDs include the `__LCC_` suffix and never resolve as the original full graph.
Square matrices use the union of off-diagonal numerical support with its transpose, after summing duplicates and removing numerical zeros. Rectangular matrices use distinct row and column vertices. Matrix signs, magnitudes and physical units are not traversal lengths: every edge has length one. Some full graphs are disconnected. Largest-component graphs are explicitly identified and retain their original vertex mapping.
The installed selection is fixed; changing viewer favorites does not change package data. See `extdata/suitesparse/selection-30_sept_2026.json` and `extdata/suitesparse/PROVENANCE.html` in the installed package.
Davis, T. A. and Hu, Y. (2011). The University of Florida Sparse Matrix Collection. tools:::Rd_expr_doi("10.1145/2049662.2049663").
Zheng, J. X., Pawar, S. and Goodman, D. F. M. (2018). Graph Drawing by Stochastic Gradient Descent. tools:::Rd_expr_doi("10.1109/TVCG.2018.2859997").
catalogue <- suitesparse.graphs()
head(catalogue)
g <- suitesparse.graph("494_bus")
identical(g, suitesparse.graph("HB/494_bus"))
c(vertices = g$n, edges = nrow(g$edges))
head(suitesparse.graph("lesmis")$vertex_data)
Run the code above in your browser using DataLab