Recover scheduler throughput lost to the snapshot-staleness check

[wbarnha]

Description

Stacked on #751 — based on claude/faust-cython-rewrites-vwltqs, not master, because it optimises code that only exists there. Retarget to master once #751 lands.

Fixing the record-loss bug reported in review of #751 cost most of the scheduler's speedup. The length-only shortcut it replaced was unsound — it could not tell "unchanged" from "swapped" — but the identity check that replaced it walks the mapping with .items(), allocating a view, an iterator and a tuple per entry. A single-partition topic reaches a pass boundary on every record, so that walk lands squarely on the hot path, and the accelerator fell from 5.7–7.7x over pure Python to 2.2–2.8x.

Two changes get it back without weakening the guarantee.

TopicBuffer._buffers becomes a plain dict

It was an OrderedDict, carrying this comment:

# note: this is a regular dict, but ordered on Python 3.6
# we use this alias to signify it must be ordered.

Dicts have been insertion-ordered by language guarantee since 3.7, and Faust requires 3.10, so the subclass bought nothing. It also cost something: OrderedDict is not an exact dict, so it cannot be walked with PyDict_Next.

Both _snapshot_is_current() methods use PyDict_Next

When the mapping is an exact dict, the check walks it with no allocation at all, comparing borrowed references by identity. Anything that is not an exact dict still goes through .items(), so a custom mapping handed to records_iterator keeps working.

Benchmarks

ns per record, CPython 3.11:

topology	python	cython	before (this PR)	after
1t × 1p × 500	525.8	90.2	2.23x	5.83x
1t × 8p × 200	422.5	78.2	—	5.40x
4t × 8p × 100	336.6	64.3	2.63x	5.23x
8t × 16p × 50	324.8	69.3	2.77x	4.68x

Back to roughly where the unsound version was, with the correctness guarantee kept.

Correctness

Unchanged from #751 and re-verified rather than assumed:

• Both mid-iteration replacement scenarios (a TopicBuffer swapped in under an existing topic name, and an iterator swapped in under an existing TP) still match the pure-Python iterator exactly.
• The randomised differential test is at 4000/4000.
• Full suite passes both ways: 2314 with the extension, 2308 with NO_CYTHON=1.

New tests cover the two things this PR makes load-bearing:

• a non-dict mapping still iterating correctly, exercising the .items() fallback;
• _buffers actually being an exact dict, since PyDict_Next silently would not apply otherwise and the speedup would quietly disappear.

Note on the OrderedDict change

This is a small behavioural surface worth a reviewer's eye: anything relying on _buffers being an OrderedDict specifically — move_to_end, popitem(last=False), equality being order-sensitive — would be affected. Nothing in the tree does; _buffers is only assigned, popped and iterated. Flagging it because it is the one change here that is not purely internal to the .pyx.

🤖 Generated with Claude Code

https://claude.ai/code/session_019oX4oGCQGgvPJHabjwfaBk

---

Generated by Claude Code