feat(historical): B2/#126 — frame-native historical (pandas out of the construction) - #129
Merged
Merged
Conversation
…es the construction, the artifact stays reader-identical _read_historical_data dispatches on declared_data_format (pipeline-core's contract fn — the one gate, never re-implemented): feature_frame descriptors fetch via get_feature_frame → FeatureFrame, making the FAO product the frame path's first production consumer (C-40 gate lifted). Frame-native month clip (frame_extraction.drop_months_above; same producer-sourced boundary, degrade-open). unfao/historical.py builds the SAME artifact faoapi's deployed reader ingests — month_id/priogrid_id/targets/9-GAUL-cols plain parquet — pandas-free (searchsorted gather from the injected ADR-011 lookup; sidecar cast discipline: plain strings, float64 codes); null-gate + unmapped-count carried over fail-loud. Deliberate divergences, both reader-safe and pinned by tests: junk row/col columns DROPPED (the legacy chain would have shipped them as bogus auto-detected targets) and values ship as SCALARS, not the C-40 single-element lists (faoapi's _validate_feature_samples normalizes both identically — verified in their code). Contract-mode gap closed: _save_contract now stages+uploads the historical artifact alongside the wire, behind the same §11.4 interlock (S7 had scoped it to the wire only — run 0 would have shipped forecasts with no actuals). Reader-level parity with the B1 characterization golden: green (same slice, frame chain vs legacy chain, compared through faoapi's reader semantics). Closes #126 (with #113-era amendments). Refs #92, #89. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This was referenced Jul 27, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part B of the run-0 OOM fix — the root cause. The historical path drops from could-not-finish-in-8-minutes/23.8 GB-OOM to a ~0.5 GB FeatureFrame + pyarrow gather.
What
managers/unfao.py:_read_historical_datadispatches ondeclared_data_format(pipeline-core's contract fn); frame branch →get_feature_frame(first production consumer of the frame path — C-40's gate has lifted). Contract-mode_savenow also ships the historical artifact (an S7 gap: run 0 would have had forecasts but no actuals), behind the same interlock.unfao/historical.py— the artifact builder, pandas-free, reader-identical; null-gate + unmapped-count preserved fail-loud.frame_extraction.drop_months_above— frame-native clip, same degrade-open policy.Deliberate reader-safe divergences (both verified in faoapi's code + pinned by tests)
row/colcolumns dropped — legacy would have served them as bogus targets._validate_feature_samples(handlers.py:374) normalizes both to the same array.AC
Closes #126. Refs #92 #89. 🤖 Generated with Claude Code