pub fn harmonic_centrality_source_ordered(
graph: &Graph,
sources: &[&str],
) -> HarmonicCentralityResultExpand description
Compute harmonic centrality while replaying sources in an explicit order.
NetworkX accumulates each target’s score in its outer sources set order:
one forward BFS per source, then centrality[target] += 1 / distance.
Floating-point addition is not associative, so the level-grouped kernels
below can differ by one ULP even though they discover identical distances.
Python bindings use this entry point with the exact CPython set iteration
order to preserve observable NetworkX values.