LoginSign Up
← Back to all sparks
L

L1centrality

INFRA · APIS
Velocity0.0

Graph/Network Analysis Based on L1 Centrality

github.com/seungwoo-stat/l1centrality ↗

A graph-centrality package that spent 2026 making its existing measures usable at scale, then went quiet.

graph-analysiscentralityr-packagevisualizationparallel-computation
◆Current state
L1centrality implements L1 centrality and prestige for graphs, including group, local, and neighbourhood variants plus MDS-based visualization. The measure set has been stable since 0.3.0; the work since has gone into interfaces around it — S3 classes with print and summary methods, plot methods for every result class, and in 0.5.0 both multi-group evaluation and multicore computation for the local variant. The two releases since have been a warning-message pass and a typo pass.
◆Where it's heading
The package has moved from defining measures to operationalizing them. 0.5.0 was the inflection: parallel local computation and list-valued group input both target users running these measures over many vertex sets or large graphs rather than illustrating them on one. The same release renamed weight_transform and eta to edge_weight_transform and vertex_weight, and added an explicit message when a distance matrix is received — the signature of a maintainer fielding the same misuse repeatedly.
◆Prediction
The last two releases carry no functional change, so the near-term path is maintenance rather than new measures; a 0.6.0 would most likely extend parallelism beyond L1centLOC to the other computationally heavy variants.

◆Recent moves

  1. 1mo ago

    Typo fixes only

    A documentation-only release with no functional change, the second consecutive maintenance tag after 0.5.0's feature work.

    View source ↗
  2. 3mo ago

    Warning message wording updated

    Reworded warnings and nothing else. It continues the pattern from 0.5.0's distance-matrix message: the maintainer is tightening what the package tells users when input is ambiguous.

    View source ↗
  3. 3mo ago

    Multi-group prominence and multicore local centrality

    The most substantive release in the window and the one that shifts the package toward scale. L1centGROUP() now accepts a list of vertex sets and computes prominence for each, and L1centLOC() gains multicore execution. The accompanying renames from weight_transform to edge_weight_transform and eta to vertex_weight trade a small break in clarity's favour.

    View source ↗
  4. 9mo ago

    Plot methods for every result class, plus edge-weight transforms

    Plot methods arrive across the result classes, with Lorenz curves, scatter plots, and a directed local-median graph, and igraph vertex colors and labels carried through automatically. Together with the optional edge-weight transform, this is the release that made results inspectable without hand-built plotting code.

    View source ↗
  5. 1y ago

    Handles unnamed vertices; quantile type pinned

    Two small robustness fixes: printing no longer breaks on graphs without vertex names, and the quantile type used internally is pinned rather than left to the default.

    View source ↗
  6. 1y ago

    S3 classes for all results, plus a Gini coefficient

    Every computation function starts returning a classed object with print and summary methods, which is the groundwork the plot methods in 0.4.0 build directly on. Gini() and the heterogeneity documentation extend the package past centrality itself into how unevenly prominence is distributed.

    View source ↗