optimize.sx adds three result-preserving passes: dce (keep outputs + ancestors,
preserve ids), cse (==build; structural sharing is free from content addressing),
and fuse (collapse 1-to-1 fusible unary chains into an artdag/pipeline node fed by
the chain head's input; leaves/fan-out/non-fusible ops never fuse). fusing-runner
replays pipeline stages, output-equivalent to the unfused dag. optimize 18/18,
total 87/87.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
execute.sx folds a plan, runs each node via an injected runner (perform in
prod, op-table in tests), and memoizes results in a lib/persist kv backend
keyed by content-id. Incremental recompute falls out of content addressing:
a leaf change reassigns ids across its dirty closure, so re-running hits the
unchanged nodes and recomputes only the closure (cold 5 -> rerun 0 -> change 3).
Cross-dag subgraph sharing verified. execute 15/15, total 69/69.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
plan.sx schedules a dag into Kahn-wave batches (parallel-safe), splits waves
wider than a cap into sub-batches, and plans incrementally over the dirty
closure only (out-of-set deps treated as satisfied cache hits). plan 18/18,
total 54/54.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
analyze.sx projects DAG edges to (edge in out) facts and runs recursive
reachable rules for deps-of/dependents-of/reachable-from/ancestors-of, plus
dirty-closure (dirty(Y):-edge(X,Y),dirty(X)) for incremental recompute. Keystone:
changing a mid node dirties only it + downstream. analyze 16/16, total 36/36.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
