Examples

Maintain a live stream

Keep the matcher as the owner of graph state. Feed it updates in order and inspect the matching after each operation.

from axiom import Matcher

matcher = Matcher(n=8, mode="basic")
updates = [
    ("insert", 0, 1),
    ("insert", 1, 2),
    ("insert", 4, 5),
    ("delete", 0, 1),
]

for operation, u, v in updates:
    getattr(matcher, operation)(u, v)
    assert matcher.maximal()

print(matcher.matching())
print(matcher.partners())
print(matcher.stats())

Replay a deterministic trace

Generate a valid insert/delete sequence from a seeded random generator, save it with your experiment, and replay it against another matcher.

import random
from axiom import Matcher
from axiom.simulation import random_updates, replay

rng = random.Random(7)
trace = list(random_updates(n=50, steps=100, rng=rng))

matcher = Matcher(50, mode="multilevel")
replay(matcher, trace)
assert matcher.maximal()
print(matcher.size(), matcher.stats()["matching_size"])

Compare modes

compare runs the same seeded workload in independent worker processes. Compare throughput only under the same graph size, update count, and seed.

from axiom.parallel import compare

if __name__ == "__main__":
    results = compare(n=100, updates=1_000, seed=42)
    for mode, result in results.items():
        print(mode, round(result.updates_per_sec), result.is_maximal)

Run this from a Python file or module. The main guard is required on platforms that use the spawn multiprocessing start method.

Command line smoke test

$ axiom --n 20 --mode basic --updates 200 --seed 42
Completed 200 updates in 0.001s
Final edges: 12
Matching size: 8
Maximal: True

Integration checklist

• Validate update ordering at your input boundary.
• Persist the update trace if you need reproducibility.
• Assert maximality in staging and benchmark runs.
• Pin the Axiom version and record the selected mode.
• Do not share one matcher instance across concurrent writers.