nx-ladybug

LadybugDB storage provider for nx-arrow columnar graphs. nx-arrow owns the Arrow-backed graph and the arrow NetworkX backend; this package only adds persistence into LadybugDB. Build and analyze graphs with nx-arrow everywhere — the only nxladybug import is the persist step.

Install

pip install nx-arrow nx-ladybug

For development:

pip install -e ".[test]"

Use

import networkx as nx
from nxladybug import LadybugStorageProvider, persist

g = nx.DiGraph([(0, 1), (1, 2)], backend="arrow")

nx.out_degree_centrality(g, backend="arrow")
list(nx.topological_sort(g, backend="arrow"))

Persist

persist() bulk-loads per-type frames through a storage provider — pick LadybugDB or Parquet:

from nxarrow.parquet import ParquetStorageProvider
from nxladybug import LadybugStorageProvider, persist

# LadybugDB: one unified node table (NxNode) + one relationship table
# (NxEdge) + a meta table (NxMeta). Each frame is registered as an Arrow
# staging table and bulk-copied natively — COPY <table> FROM (MATCH ...) —
# so endpoints resolve through the node primary key inside the copy.
provider = LadybugStorageProvider("graph.lbdb")  # or ":memory:"
persist(g, provider, graph_attrs=g.graph)
restored = provider.load_graph()  # full round-trip

# ... or Parquet: one file per type plus a manifest.json
parquet = ParquetStorageProvider("graph_store")
persist(g, parquet, graph_attrs=g.graph)
restored = parquet.load_graph()

The Ladybug provider drops and recreates its tables on every persist, so re-persisting a smaller graph leaves no stale rows. :memory: databases live as long as the provider; file paths reopen with LadybugStorageProvider.load(path).

Custom stores implement store_nodes(frames) / store_edges(frames) and register with nxarrow.register_storage_provider.

Develop

pytest