Interoperability

hypermesh.interop bridges a HyperMesh hypergraph snapshot to the wider network-science ecosystem and to portable, library-agnostic exchange formats. This makes HyperMesh a first-class citizen alongside NetworkX, HyperNetX and XGI, and lets researchers move data in and out without lock-in.

pip install "hypermesh[interop]"   # numpy + scipy + networkx + hypernetx + xgi
pip install "hypermesh[pandas]"    # for to_pandas()

Every exporter takes either a HypergraphPy snapshot or an embedded Connection (with table=):

import hypermesh as hm

db = hm.connect("/var/lib/hypermesh/data")
hg = db.to_hypergraph("CoProximity")          # explicit snapshot

g = hm.interop.to_networkx(hg)                 # from a snapshot
g = hm.interop.to_networkx(db, table="CoProximity")  # straight from a Connection

Library bridges

Function	Returns	Notes
`to_networkx(hg, kind="bipartite")`	`networkx.Graph`	Lossless star expansion (node ↔ hyperedge incidence).
`to_networkx(hg, kind="clique")`	`networkx.Graph`	Weighted two-section (clique) expansion; lossy.
`to_hypernetx(hg)`	`hypernetx.Hypergraph`	Edge props: `weight`, `event_ts`, `size`.
`to_xgi(hg)`	`xgi.Hypergraph`	Integer edge ids, same edge attributes.
`to_pandas(hg, orient="edgelist")`	`pandas.DataFrame`	One row per incidence: `edge, node, weight, event_ts`.
`to_pandas(hg, orient="incidence")`	`pandas.DataFrame`	Dense `n_nodes × n_edges` 0/1 matrix.

g = hm.interop.to_networkx(hg)                 # bipartite by default
import networkx as nx
print(nx.density(g))

clq = hm.interop.to_networkx(hg, kind="clique")
H   = hm.interop.to_hypernetx(hg)
X   = hm.interop.to_xgi(hg)
df  = hm.interop.to_pandas(hg)                 # edge-list DataFrame

NetworkX node namespaces

In the bipartite projection the two node populations share one graph, so IDs are prefixed to avoid collisions:

member nodes → "n{node_id}" with bipartite=0, kind="node"
hyperedge nodes → "e{j}" with bipartite=1, kind="hyperedge" and the weight / event_ts / size attributes.

Exchange formats

HIF — Hypergraph Interchange Format

HIF is the portable JSON standard shared by HyperNetX, XGI and others. Round-trips through HyperMesh cleanly:

hm.interop.write_hif(hg, "graph.hif.json")        # export
hg2 = hm.interop.read_hif("graph.hif.json")        # import → HypergraphPy

doc  = hm.interop.to_hif(hg)                        # in-memory dict
hg3  = hm.interop.from_hif(doc)                     # rebuild from dict/JSON str

The document carries incidences, per-edge attrs (weight, event_ts, size), and a nodes list so isolated nodes survive the round-trip.

GraphML

hm.interop.to_graphml(hg, "graph.graphml")                 # bipartite
hm.interop.to_graphml(hg, "clique.graphml", kind="clique") # two-section

GraphML is written through the NetworkX projection, so it opens directly in Gephi, yEd, Cytoscape and similar tools.

Remote clients

Interop builds the hypergraph in-process and therefore needs an embedded Connection. From a remote Client, fetch rows and build the snapshot yourself:

from hypermesh.analytics import build_hypergraph

rows = list(client.execute("MATCH HYPEREDGE (he:CoProximity) RETURN *"))
hg   = build_hypergraph(rows)
g    = hm.interop.to_networkx(hg)