Skip to content

pivoterpy.graph

The graph module defines the Graph type used by pivoter. All public constructors are also exposed on the package as pvt.from_edge_list, pvt.from_adj_matrix, and pvt.from_networkx.

Graph

Represents an undirected simple graph: internal vertex IDs are contiguous 0 .. n-1, with an optional map back to original labels.

Attributes

Name Type Description
edges list[tuple[int, int]] Undirected edges as (u, v) with u < v, internal IDs
n int Number of vertices
m int Number of edges
degrees list[int] Degree per internal vertex ID
nodes list[int] nodes[i] is the original label for internal vertex i
experimental — Reserved (None after construction); not used by current backends

Constructor (__init__)

Normally you should use a classmethod below. The direct constructor builds a graph from already-normalized data:

from pivoterpy.graph import Graph

edges = [(0, 1), (1, 2)]
G = Graph(edges=edges, n=3, nodes=[10, 20, 30])  # optional node id map
  • edges: internal (u, v) pairs, u < v, each edge once.
  • n: vertex count (must cover all indices appearing in edges).
  • nodes: length n; defaults to [0, 1, ..., n-1] if omitted.

from_edge_list

Graph.from_edge_list(array: list[tuple[int, int]]) -> Graph

Builds a graph from arbitrary integer vertex labels. Labels are compressed to 0 .. n-1, duplicates and self-loops are dropped, and nodes records the inverse map.

import pivoterpy as pvt

G = pvt.from_edge_list([(0, 1), (1, 2), (5, 0)])

Raises if no valid edges remain after filtering.

from_adj_matrix

Graph.from_adj_matrix(array: Sequence[Sequence[int | float | bool]]) -> Graph

Expects a square matrix. Only the strict upper triangle is read; an edge (i, j) is added when array[i][j] > 0 (so True counts as an edge). Diagonal and lower triangle are ignored.

import pivoterpy as pvt

matrix = [
    [0, 1, 0],
    [0, 0, 1],
    [0, 0, 0],
]
G = pvt.from_adj_matrix(matrix)

Vertex i in the matrix stays internal vertex i (no relabeling).

from_networkx

Graph.from_networkx(G: object) -> Graph

Imports G.edges() and uses G.number_of_nodes() as n. This matches graphs whose nodes are already integers 0 .. n-1 (for example nx.path_graph(5)).

If your graph uses arbitrary labels (strings, non-contiguous integers, etc.), prefer relabeling first or use from_edge_list(list(your_graph.edges())), which applies the same compression as the edge-list path.

import networkx as nx
import pivoterpy as pvt

Gnx = nx.complete_graph(4)
G = pvt.from_networkx(Gnx)

NetworkX is not a hard dependency of the core package; install it when you need this constructor.