Add __slots__ to Node and Path

[invisig0th]

Summary

Add __slots__ to synapse.lib.node.Node and synapse.lib.node.Path to remove the per-instance __dict__ and reduce memory for these high-cardinality runtime objects (a Node/Path pair is created for every node touched during Storm execution).

Details

• Node.__slots__ includes:
  • __weakref__ -- Node instances are held in the snap livenodes WeakValueDictionary (synapse/lib/snap.py), so they must remain weak-referenceable.
  • isrunt -- set externally on the instance in synapse/lib/snap.py.
• Path.__slots__ includes display, which is set externally in synapse/lib/storm.py.
• Rewrote the coherency test in synapse/tests/test_lib_snap.py: it previously set a dynamic node._test attribute (which __slots__ forbids) to prove a re-fetched node was a different object. It now holds a weakref.ref to the original node and asserts it is collected after GC -- a more direct check that the node was released from the coherency cache.
• Added a feat changelog entry.

Backward compatibility

__slots__ prevents setting arbitrary attributes on Node/Path instances. All in-tree assignments are accounted for; out-of-tree code that monkey-patches attributes onto these instances would need updating (noted in the changelog).

Testing

python -m pytest synapse/tests/test_lib_node.py synapse/tests/test_lib_snap.py    # 29 passed
python -m pytest synapse/tests/test_lib_storm.py synapse/tests/test_lib_ast.py     # 120 passed

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.76%. Comparing base (186ba33) to head (afed0f3).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4966      +/-   ##
==========================================
- Coverage   97.81%   97.76%   -0.05%     
==========================================
  Files         299      299              
  Lines       63433    63435       +2     
==========================================
- Hits        62047    62019      -28     
- Misses       1386     1416      +30     

Flag	Coverage Δ
linux	97.70% <100.00%> (-0.05%)	⬇️
linux_replay	93.51% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[invisig0th]

Performance: max Storm node throughput (before vs after __slots__)

Method. In-memory test Cortex (getTestCore), 100k test:int nodes pre-added, then timing test:int | spin — this forces a Node + Path per node through the full Storm pipeline and then discards them, i.e. the max-throughput path. 2 warmup + 6 measured trials; best = min wall time. Only synapse/lib/node.py was toggled between master (no slots) and this branch (slots). Python 3.14, single host.

Throughput

	best	mean
before (no slots)	34,843 nodes/s	34,591 nodes/s
after (slots)	34,885 nodes/s	34,480 nodes/s

No measurable throughput change. This path is dominated by LMDB reads, msgpack decode and async scheduling, not per-instance __dict__ allocation, so removing the dict doesn't move the pipeline rate.

Where the win actually is: memory

Average bytes per live Node, sys.getsizeof(obj) + sys.getsizeof(obj.__dict__), over 20k lifted nodes:

	per Node
before (no slots)	200 B
after (slots)	136 B

-64 B/node (-32%). That delta is exactly the removed __dict__. At scale (large lifts / many simultaneously-live nodes) this is the real benefit; throughput is unchanged.

Caveat: in-memory test cortex on a single host under Python 3.14 (CI pins 3.11); absolute nodes/s will differ on real deployments, but the before/after comparison is apples-to-apples since only node.py changed.