Private
Public Access
Campaign: address the parent code_path_audit_20260607 Finding 1 (CRITICAL)
Metadata 4.01e22 effective codepaths via 3 SSDL techniques.
3 children, sequential, with budget gates:
1. metadata_nil_sentinel_20260624 (>= 10% drop): introduce
NIL_METADATA sentinel + migrate 6 nil-check functions.
2. metadata_generational_handle_20260624 (>= 20% drop,
BLOCKED_BY 1): wrap Metadata in (index, generation) handle;
collapse lifetime branches to 1 lookup + 1 cmp.
3. metadata_field_cache_20260624 (>= 30% drop, BLOCKED_BY 2):
MetadataFieldCache keyed by (handle.index, field_name);
123 string-keyed entry.get('key', default) sites become
cache lookups.
Each child has its own spec/plan/metadata/state. Budget gate
after each child: re-measure effective codepaths; if drop < threshold,
PAUSE the campaign and report to user.
End-of-campaign TRACK_COMPLETION captures the cumulative reduction
vs the 4.01e22 baseline. Deferred follow-up: apply the same
3 SSDL primitives to the 4 other dict[str, Any] aliases
(FileItem, CommsLogEntry, HistoryMessage, ToolDefinition, ToolCall).
16 files committed: 4 directories x 4 files each (spec, plan,
metadata, state).
6.5 KiB
6.5 KiB
Track Specification: metadata_generational_handle_20260624
Overview
Child 2 of the metadata_ssdl_defusing_20260624 campaign. Wraps Metadata in a (index, generation) handle resolved through a registry. Collapses lifetime branches to 1 lookup + 1 generation comparison. BLOCKED_BY child 1 (the nil sentinel is the fallback path on generation mismatch).
Current State Audit (master @ child-1-SHIPPED, after metadata_nil_sentinel_20260624)
NIL_METADATAsentinel exists (from child 1)- The 6 nil-check functions use sentinel-return
- The 3466 branch points in the parent audit include lifetime checks (e.g., "is this handle still valid?")
src/aggregate.pyandsrc/ai_client.pycontain lifetime checks; they should be replaced with handle lookup + generation comparison
Goals
| ID | Goal | Acceptance |
|---|---|---|
| G1 | MetadataHandle (or equivalent) introduced: a (index: int, generation: int) pair |
The handle type is exported; can be created and queried |
| G2 | MetadataHandleRegistry (or equivalent) introduced: stores index -> generation mapping |
The registry has O(1) lookup, bump-generation, and get-with-validation methods |
| G3 | Production Metadata is wrapped in a handle at the consumer entry points |
Consumers can do handle.registry_lookup() instead of if entry.lifetime != current_lifetime: ... |
| G4 | 1 behavioral test for the handle | tests/test_metadata_generational_handle.py exists; asserts lookup, generation mismatch returns NIL_METADATA, bump invalidates cached lookups |
| G5 | Budget gate met: effective-codepaths drop ≥ 20% vs post-child-1 measurement | Re-measurement shows the drop |
Non-Goals
- Touching the 4 other
dict[str, Any]aliases — out of scope (deferred to follow-up campaign) - Touching the list-typed aggregates — out of scope
- Replacing the 3 candidate placeholders — blocked on
any_type_componentization_20260621 - Adding a generational handle for the inner Metadata values within nested structures (the campaign handles the top-level Metadata; nested handles are out of scope)
- Cache invalidation (that's child 3; this child just provides the identity)
Functional Requirements
FR1: Handle + Registry types
In a sensible location (likely src/aggregate.py per AGENTS.md §File Naming Convention, OR a NEW module if the user explicitly approves):
@dataclass(frozen=True)
class MetadataHandle:
index: int
generation: int
class MetadataHandleRegistry:
def __init__(self) -> None:
self._generations: dict[int, int] = {}
def register(self, metadata: Metadata) -> MetadataHandle:
# returns a new handle with a fresh index + generation
def lookup(self, handle: MetadataHandle) -> Metadata | None:
# if handle.generation != self._generations[handle.index], return None
# otherwise return the stored Metadata (or sentinel for now)
def bump_generation(self, index: int) -> None:
# invalidate the cached entry for this index
(Exact API up to Tier 2; the contract is: handle + registry with O(1) lookup + generation-based invalidation.)
FR2: Migrate lifetime-branch code
For each site in src/ that does lifetime checks (e.g., "is this Metadata still the one I cached?"):
- Replace with
handle = registry.register(metadata)+value = registry.lookup(handle)+if value is None: use NIL_METADATA
FR3: Behavioral test
tests/test_metadata_generational_handle.py with at least 3 tests:
test_register_returns_handle: assertregistry.register(metadata)returns aMetadataHandlewith distinctindexandgenerationtest_lookup_returns_none_after_bump: assertregistry.lookup(handle)returns None afterregistry.bump_generation(handle.index)test_lookup_returns_none_for_unknown_index: assertregistry.lookup(MetadataHandle(index=999, generation=1))returns Nonetest_lookup_returns_value_for_valid_handle: assertregistry.lookup(handle)returns the registered Metadata
Non-Functional Requirements
- NFR1: 1-space indentation
- NFR2: CRLF line endings on Windows
- NFR3: No comments in source code
- NFR4: Per-task atomic commits with git notes
- NFR5: No new pip dependencies
- NFR6:
Result[T]returns for fallible registry methods (perconductor/code_styleguides/error_handling.md) - NFR7: No new
src/<thing>.pyfiles (per AGENTS.md) — unless the user explicitly approves; default is to putMetadataHandle+MetadataHandleRegistryinsrc/aggregate.pyor another existing module
Architecture Reference
src/aggregate.py(the parent module forMetadata)NIL_METADATA(from child 1) — the fallback returned bylookupon generation mismatchdocs/reports/code_path_audit/2026-06-22/AUDIT_REPORT.mdFinding 1 Fix 3 — the Generational Handle proposalconductor/code_styleguides/data_oriented_design.md— canonical DOD reference
Out of Scope
- The 4 other
dict[str, Any]aliases (deferred) - The 3 candidate placeholders (blocked)
- The cache (child 3)
- Runtime profiling (Track F from the previous menu; deferred)
Verification Criteria (Definition of Done)
| # | Criterion | Verification command |
|---|---|---|
| VC1 | MetadataHandle and MetadataHandleRegistry exist |
grep -rn "class MetadataHandle|class MetadataHandleRegistry" src/ |
| VC2 | Production code uses handle + registry at the entry points | grep -rn "registry.lookup|registry.register" src/ returns ≥ 1 hit |
| VC3 | Behavioral test exists and passes | uv run pytest tests/test_metadata_generational_handle.py -v |
| VC4 | Budget gate met | compute_effective_codepaths(Metadata_profile) returns number ≥ 20% smaller than post-child-1 measurement |
| VC5 | Full test suite remains green | uv run python scripts/run_tests_batched.py → 11/11 tiers PASS |
| VC6 | 4 audit gates remain clean | weak_types ≤ 112, type_registry in sync, main_thread_imports clean, no_models_config_io clean |
Risks
| # | Risk | Likelihood | Mitigation |
|---|---|---|---|
| R1 | The handle breaks code that expects raw Metadata |
medium | The handle is a wrapper; consumers can extract the raw value via .value or similar. Behavioral test verifies backwards-compat for the common cases. |
| R2 | The registry's lookup is not actually O(1) | low | The registry uses a dict[int, int] for index -> generation; lookup is O(1) by construction. |
| R3 | Budget gate fails (drop < 20%) | low | The 3466 branch points include lifetime checks; replacing with handle lookup should drop the count. If not, the SSDL math is wrong. |