Phase 1: Strategy-Differentiating Revocation Artifacts and Figure Pipeline #81

Merged

erikinkinen merged 7 commits from 1-strategy-differentiating-revocation-artifacts-and-figure-pipeline into main

2026-03-06 13:49:09 +01:00

erikinkinen commented

2026-03-06 13:46:48 +01:00

Owner

Closes #79

Summary

Implements Issue #79: Strategy-Differentiating Revocation Artifacts and Figure Pipeline.

This PR shifts Phase 1 analysis from revoke-only/size-only comparisons to strategy trade-offs by:

adding strategy-separating workloads to CLI + sweep,
emitting per-run metrics_v2 artifacts in sweep outputs,
upgrading the figure pipeline to consume metrics_v2 + event logs,
producing deterministic trade-off plots.

What Changed

New workload support wired end-to-end

Added/connected:
- deep_tree_mid_subtree_revoke
- root_revoke_with_large_unrelated_background
- shared_descendant_revoke
aes simulate now validates flags/minima/forbidden-flag combos for all new workloads.
Sweep config parser accepts strict JSON schemas for the new workload shapes.

Sweep artifact model extended

Per-run outputs now include:
- event_log.jsonl
- metrics_v1.json
- metrics_v2.json
runs.csv now includes metrics_v2_path.

Figure pipeline migrated to strategy-differentiating metrics

Plotting now consumes:
- dataset-level runs.csv / metrics_long.csv contract,
- per-run metrics_v2.json and event_log.jsonl referenced by runs.csv.
Added deterministic plots (PNG + SVG):
- immediate_vs_total_cost_by_strategy
- completeness_vs_total_cost_by_workload
- residual_risk_vs_revoke_latency
- metadata_overhead_vs_completeness
- post_revoke_hot_path_cost
- depth_sensitivity_completeness
- depth_sensitivity_total_cost
- fanout_sensitivity_completeness
- fanout_sensitivity_total_cost

Canonical figure sweep configs updated

Cost/completeness canonical configs now include strategy-separating workloads for stronger differentiation in plots.

Documentation and tests updated

Updated docs/phase1.md for:
- new workload schemas and minima,
- sweep artifact updates (metrics_v2),
- new figure pipeline inputs/outputs.
Updated/expanded tests:
- CLI simulate/sweep coverage for new workloads and strict validation,
- sweep CSV expectations (metrics_v2_path),
- figure contract + determinism tests and smoke test output list.

Validation

Executed and passing:

aes_cli_simulate_tests
aes_cli_sweep_tests
aes_metrics_runner_tests
aes_revocation_outcome_metrics_tests
aes_phase1_figures_tests
aes_phase1_figures_smoke_test
plus regression slice:
- aes_event_log_reader_tests
- aes_event_log_replay_tests
- aes_revocation_strategy_tests
- aes_strategy_equivalence_tests
- aes_invalid_event_determinism_tests

Notes

This branch history was rewritten to align commit subjects with Issue #79.

Closes #79 ## Summary Implements **Issue #79: Strategy-Differentiating Revocation Artifacts and Figure Pipeline**. This PR shifts Phase 1 analysis from revoke-only/size-only comparisons to strategy trade-offs by: - adding strategy-separating workloads to CLI + sweep, - emitting per-run `metrics_v2` artifacts in sweep outputs, - upgrading the figure pipeline to consume `metrics_v2` + event logs, - producing deterministic trade-off plots. ## What Changed 1. New workload support wired end-to-end - Added/connected: - `deep_tree_mid_subtree_revoke` - `root_revoke_with_large_unrelated_background` - `shared_descendant_revoke` - `aes simulate` now validates flags/minima/forbidden-flag combos for all new workloads. - Sweep config parser accepts strict JSON schemas for the new workload shapes. 2. Sweep artifact model extended - Per-run outputs now include: - `event_log.jsonl` - `metrics_v1.json` - `metrics_v2.json` - `runs.csv` now includes `metrics_v2_path`. 3. Figure pipeline migrated to strategy-differentiating metrics - Plotting now consumes: - dataset-level `runs.csv` / `metrics_long.csv` contract, - per-run `metrics_v2.json` and `event_log.jsonl` referenced by `runs.csv`. - Added deterministic plots (PNG + SVG): - `immediate_vs_total_cost_by_strategy` - `completeness_vs_total_cost_by_workload` - `residual_risk_vs_revoke_latency` - `metadata_overhead_vs_completeness` - `post_revoke_hot_path_cost` - `depth_sensitivity_completeness` - `depth_sensitivity_total_cost` - `fanout_sensitivity_completeness` - `fanout_sensitivity_total_cost` 4. Canonical figure sweep configs updated - Cost/completeness canonical configs now include strategy-separating workloads for stronger differentiation in plots. 5. Documentation and tests updated - Updated `docs/phase1.md` for: - new workload schemas and minima, - sweep artifact updates (`metrics_v2`), - new figure pipeline inputs/outputs. - Updated/expanded tests: - CLI simulate/sweep coverage for new workloads and strict validation, - sweep CSV expectations (`metrics_v2_path`), - figure contract + determinism tests and smoke test output list. ## Validation Executed and passing: - `aes_cli_simulate_tests` - `aes_cli_sweep_tests` - `aes_metrics_runner_tests` - `aes_revocation_outcome_metrics_tests` - `aes_phase1_figures_tests` - `aes_phase1_figures_smoke_test` - plus regression slice: - `aes_event_log_reader_tests` - `aes_event_log_replay_tests` - `aes_revocation_strategy_tests` - `aes_strategy_equivalence_tests` - `aes_invalid_event_determinism_tests` ## Notes This branch history was rewritten to align commit subjects with **Issue #79**.