Phase 6 of the decoupling refactor (behavior-preserving move):
- topology_export.py: body_geometry / surface_geometry / topology_records
(~290 lines) moved verbatim out of build123d_adapter.py as module-level
functions; includes the circle_center_mm/radius_mm concentric-circle
evidence added by the lk_dev integration
- build123d_adapter.py: the three methods are now one-line forwards, so
kernel construction and topology evidence export live in separate files
and can evolve independently
No protocol change (GeometryAdapter untouched). Verified by:
- golden comparison: full-corpus analyze (2881 docs) + 12 complete
rebuilds -- zero field differences vs f79c287
- full test suite: 435 tests, failure set identical to baseline
Phase 8: README now maps every module of the split package, documents
the executor registration + schema-contract workflow for new atomic
operations, and restates the schema maintenance rule for executors/.
Phase 7 of the decoupling refactor (behavior-preserving move):
- legacy/llm_compiler.py, legacy/llm_engine.py: frozen build_pack path
(unused by run_cdsl_only), moved with git mv for history
- legacy/exact_rebuild.py: _run_parameterized / _run_exact /
_apply_geometric_compensations moved out of rebuild.py, including
the part-specific compensation table
- rebuild.py keeps run_rebuild / run_cdsl_only / compare_with_gold and
imports the moved functions
- llm_compiler.py / llm_engine.py become compatibility shims
Frozen zone: new capability belongs in executors/ + schema contracts.
The project-specific compensation no longer sits in the main pipeline.
Phase 5 of the decoupling refactor (behavior-preserving move):
- translator/ir.py: SolidWorks plugin JSON to backend-IR conversion (70 syms)
- translator/codegen.py: backend IR to build123d source generation (64 syms)
- translator/runtime_lib.py: frozen generated-script runtime library,
spliced into generate_build123d_code as *RUNTIME_LIB_LINES
- translator/common.py: helpers shared by both sides
- translator/__init__.py: full historical symbol surface re-exported
Generated-code equivalence verified byte-for-byte against the pre-split
output for a representative IR sample; py_compile clean.
Phase 4 of the decoupling refactor (behavior-preserving):
- profile_schema.json: every operation contract now carries
runtime_capability {body_mutating, requires_active_body,
replayable, requires_selector, open_profile_ok} (single source of truth)
- operation_contracts.py: validates and forwards the flags
- capabilities.py: the five data-classification frozensets are now
derived from the schema at import time; dispatch-logic sets
(_HOLE_ATOMICS, _PATTERN_ATOMICS, extent constants) stay in code
- test_profile_schema.py: completeness + structural invariants test
Equivalence proven by flag counts (27/13/25/6/2) matching the previous
hand-written sets and by the unchanged test-baseline failure set.
Phase 3 of the decoupling refactor (behavior-preserving):
- registry.py: atomic_executor decorator, ALL_ATOMIC_IDS with
fail-fast registration validation, execute_node dispatcher
- executors/: one module per family (extrude, revolve, surfaces,
loft_sweep, bodies, context, primitives, parametric, holes,
dressup, patterns) + shared helpers in executors/common
- executors/__init__: explicit aggregation + completeness check
(registry must cover every declared atomic id at import time)
- runtime.py: slimmed to entry points (analyze_cdsl/rebuild_cdsl)
plus full historical re-exports incl. test-referenced privates
Adding an atomic operation now touches only one executor module and
its schema contract; the shared registry never changes. Verified
against baseline: zero new failures.
likang
ganjihong