qbiocode.apps.quvine.data.prepare module

Summary

Classes:

PrepareGraphConfig

PrepareGraphConfig(subsample_nodes: 'bool' = False, max_nodes: 'int' = 3000, radius: 'int' = 2, sparsify_edges: 'bool' = True, retain_ratio: 'float' = 0.7, max_degree: 'int' = 40, scoring: "Literal['common_neighbors', 'prefer_low_degree']" = 'common_neighbors', verbose: 'bool' = False)

Functions:

keep_largest_connected_component	Return a connected component while preserving required nodes when possible.
prepare_graph	Step-by-step graph preparation:

Reference

keep_largest_connected_component(G, *, required_nodes=None)

Return a connected component while preserving required nodes when possible.

If required_nodes are supplied and they all lie in the same connected component, that component is preferred over the global largest component. Otherwise, fall back to the largest connected component.

Return type:

Graph

class PrepareGraphConfig(subsample_nodes=False, max_nodes=3000, radius=2, sparsify_edges=True, retain_ratio=0.7, max_degree=40, scoring='common_neighbors', verbose=False)

Bases: object

subsample_nodes: bool = False

max_nodes: int = 3000

radius: int = 2

sparsify_edges: bool = True

retain_ratio: float = 0.7

max_degree: int = 40

scoring: Literal['common_neighbors', 'prefer_low_degree'] = 'common_neighbors'

verbose: bool = False

prepare_graph(cfg, graph, seed, *, seeds=None, targets=None)

Step-by-step graph preparation:

1. Materialize a clean undirected simple graph (no views).

2. If cfg.subsample_nodes: subsample nodes while protecting seeds+targets,

   expand neighborhood (radius hops), fill remaining budget.

3. If cfg.sparsify_edges: degree-capped, biology-aware edge sparsification

   (triangle support or low-degree preference).

4. Return a clean nx.Graph.

Designed to avoid the KeyError / _succ / neighbor traversal issues you saw: - no subgraph views - no nx.is_connected / connected_components - no G.neighbors() calls - only adjacency dict access (G.adj[u])

Return type:

Graph