You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Parity apparatus efficiency: shard R cache, close automation loop, dedupe CI, pin dev env
Cache sharding (schema v2): tests/_r_cache.json (15.7MB monolith, opaque hash
keys) becomes tests/_r_cache/ with one JSON shard per R function label and
keys of the form "<label>|<sha256>". Regeneration diffs now name the function
they touch, git delta-compresses the shards, and a live-R run rewrites only
the shard it changed. All 2314 entries were migrated by replaying tests/parity
against the old monolith (the sha256 payload is unchanged, so legacy keys are
the digest part of new keys); zero orphans, full suite green cache-only.
regenerate_r_cache.py, run_live_r_parity_for_changed_api.py, the sync
manifest/scripts, and docs follow the new layout.
Automation loop: the regen workflow now records r_commit/r_version in
sync/nns_source.json itself and posts a per-function changed-entry table in
the PR body, so a cache PR is self-consistent and reviewable without forensic
scripts.
CI dedupe: native-backend-ci ran the identical Python 3.11-3.14 matrix twice
per PR push (push + pull_request events). push is now main-only, with a
concurrency group cancelling superseded runs.
Dev env pinning: pre-commit hooks run ruff/mypy via `python -m` in the
project environment, so local results match CI instead of a foreign mypy
reporting spurious numpy import errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AxBV5HmXHY5nCUvW9SwDvb
The exact R source package was installed and the committed parity cache was regenerated. Any remaining Python/R parity mismatches are retained in the workflow diagnostics and should be repaired against this R-authored baseline.
345
+
The exact R source package was installed and the committed parity cache was regenerated. `sync/nns_source.json` records this R commit as the behavioral-truth provenance. Any remaining Python/R parity mismatches are retained in the workflow diagnostics and should be repaired against this R-authored baseline.
0 commit comments