v2.0.9 - bugfixes

What's Changed

• Fix biological gene-disease layer assignments by @SkBlaz in #1512
• Paper feedback by @popchanovska in #1464
• Add GC-0002 grant to paper acknowledgements by @bkolosk1 in #1516
• Fix biological layer density edge counts by @SkBlaz in #1515
• Fix NetworkX export in quick reference by @SkBlaz in #1513
• Fix layer-specific biological network metrics by @SkBlaz in #1514
• Remove obsolete CLI coverage plan by @SkBlaz with @Copilot in #1518
• Fix temporal conversion of multilayer replica edges by @SkBlaz in #1519
• Fix UTC Z timestamp parsing on older Python versions by @SkBlaz in #1520
• No code changes: Lean failure not reproduced by @SkBlaz with @Copilot in #1522
• Prevent infinite temporal iteration from floating-point rounding by @SkBlaz in #1523
• Return empty results for non-positive SELECT limits by @SkBlaz in #1524
• Reject incomplete WHERE clauses in legacy SELECT queries by @SkBlaz in #1527
• Reuse layer iterators across temporal windows by @SkBlaz in #1528
• Honor configured timestamp fields in temporal views by @SkBlaz in #1529
• Filter parallel edges independently in temporal views by @SkBlaz in #1530
• Preserve zero run seeds in AutoCommunity benchmark UQ by @SkBlaz in #1535
• Preserve zero seeds in community uncertainty queries by @SkBlaz in #1534
• Retain converged scores in streaming PageRank by @SkBlaz in #1532
• Reject explicit zero steps in windowed DSL queries by @SkBlaz in #1531
• Count parallel edges in streaming PageRank contributions by @SkBlaz in #1533
• Add deterministic, semantics-preserving DSL optimizer rewrites by @SkBlaz with @Copilot in #1526
• Preserve union metrics for false-valued node identifiers by @SkBlaz in #1543
• Preserve four-field edge identities in DataFrame exports by @SkBlaz in #1541
• Preserve positional metrics in result symmetric difference by @SkBlaz in #1542
• Redistribute dangling-node mass in streaming PageRank by @SkBlaz in #1537
• Retain single-timestamp networks in inferred temporal windows by @SkBlaz in #1539
• Keep unsupported AutoCommunity wins mode from aborting benchmarks by @SkBlaz in #1536
• Align attribute lists when subtracting query results by @SkBlaz in #1540
• Count undirected self-loops once in streaming PageRank transitions by @SkBlaz in #1538
• Sample directed edges in Erdos-Renyi null models by @SkBlaz in #1545
• Scope cached method results to receiver and function identity by @SkBlaz in #1544
• Preserve directed multigraphs in NetworkX edge exports by @SkBlaz in #1546
• Preserve scalar node and layer identities through Arrow imports by @SkBlaz in #1548
• Prevent numeric timestamp strings from being parsed as calendar dates by @SkBlaz in #1549
• Fix empty and multigraph network exports and preserve edge data by @SkBlaz in #1547
• Restore benchmark metric registry after property tests by @SkBlaz in #1550
• Return no classification labels when top-k is zero by @SkBlaz in #1553
• Allow external Node2Vec wrapper use without optional gensim by @SkBlaz in #1551
• Respect undirected edge identity in layer Jaccard similarity by @SkBlaz in #1552
• Keep dependency tests importable on supported older Python by @SkBlaz in #1556
• Disambiguate colliding replica export identities by @SkBlaz in #1557
• Preserve scalar identifiers in Arrow and Parquet round trips by @SkBlaz in #1558
• Separate approximation sampling from metric uncertainty budgets by @SkBlaz in #1563
• Retain execution settings in canonical query provenance by @SkBlaz in #1560
• Read approximation and embedding specs from AST JSON by @SkBlaz in #1561
• Apply configured parameters during approximation dispatch by @SkBlaz in #1564
• Preserve weights in multilayer edge-list exports by @SkBlaz in #1565
• Preserve graph data when parsing gpickle layer labels by @SkBlaz in #1555
• Retain embeddings in scoped canonicalization by @SkBlaz in #1568
• Isolate canonical ASTs from mutable input settings by @SkBlaz in #1569
• Preserve selected column order in canonical query replay by @SkBlaz in #1567
• Respect parallel-edge weights in approximate betweenness by @SkBlaz in #1573
• Retain result stages when reconstructing query builders by @SkBlaz in #1562
• Preserve typed targets in AST JSON round trips by @SkBlaz in #1559
• Accept mixed node identities in weighted betweenness sampling by @SkBlaz in #1572
• Reload multilayer rows written by the edge-list exporter by @SkBlaz in #1570
• Correct scaling of sampled betweenness and standard errors by @SkBlaz in #1571
• Compute robustness paths over reachable directed pairs by @SkBlaz in #1554
• Correct self-distance bias and disconnected closeness scaling by @SkBlaz in #1577
• Make seeded closeness reproducible across Python hash seeds by @SkBlaz in #1575
• Respect directed arcs and reciprocal capacities in maximum flow by @SkBlaz in #1579
• Use standard normalization for undirected streaming degree centrality by @SkBlaz in #1580
• Preserve component coverage and budget in closeness sampling by @SkBlaz in #1576
• Attach computed metrics to selected multigraph edge data by @SkBlaz in #1585
• Honor explicit direction conversion for gpickle multigraphs by @SkBlaz in #1583
• Retain configured directedness when loading networks by @SkBlaz in #1582
• Preserve isolated replicas and graph data in GML loading by @SkBlaz in #1581
• Implement the documented GraphML network loader by @SkBlaz in #1588
• Sum parallel edge capacities in maximum flow by @SkBlaz in #1584
• Make seeded layer shuffles stable across Python hash seeds by @SkBlaz in #1578
• Preserve selected multigraph edge keys and source attributes by @SkBlaz in #1586
• Honor requested direction when loading NetworkX graphs by @SkBlaz in #1587
• Preserve weighted adjacency values in MAT graph parsing by @SkBlaz in #1589
• Match inward directed semantics in approximate closeness by @SkBlaz in #1574

New Contributors

• @bkolosk1 made their first contribution in #1516

Full Changelog: v2.0.8...v2.0.9