accordsync-core

accordsync-core

Accord's merge core for Python: hybrid logical clocks, operations and the four merge strategies.

Pure Python, no I/O, no dependencies. Most programs use the client accordsync or the server accordsync-server, which depend on it. Python 3.11+.

pip install accordsync-core

set_() has a trailing underscore so it does not shadow Python's set.

import time

from accordsync_core import LocalWriter, conflict, counter, define_schema

schema = define_schema({"dossier": {"visits": counter(), "status": conflict()}})


def now() -> int:
    return time.time_ns() // 1_000_000


a = LocalWriter(schema, "phone-a", now)
b = LocalWriter(schema, "phone-b", now)

# Each device writes offline...
from_a = [a.inc("dossier:1", "visits", 1), a.assign("dossier:1", "status", "approved")]
from_b = [b.inc("dossier:1", "visits", 2), b.assign("dossier:1", "status", "rejected")]

# ...then they exchange ops, in any order, as often as the network repeats them.
for op in from_b + from_b:
    a.receive(op)
for op in reversed(from_a):
    b.receive(op)

assert a.replica.snapshot() == b.replica.snapshot()
print(a.replica.read("dossier:1"))
# {'visits': 3, 'status': {'conflicted': [{'value': 'approved', 'opId': 'phone-a:2'},
#                                         {'value': 'rejected', 'opId': 'phone-b:2'}]}}

It merges byte for byte like @accordsync/core: the test suite runs the shared golden vectors in every delivery order, and random scenarios generated by the TypeScript core must produce identical snapshots here.

Docs: accord.benhattab.pro/docs/schema · Source: crossben/accordsync-python · Licence: Apache-2.0