Repository navigation
Conversation
Contributor
Author
|
Independent merge-delta re-review APPROVE at 9f7dead: upstream mandatory explicit RuleSpec roots and canonical fixtures are preserved; NZ period, Decimal, and relationship checks remain intact. Independent focused checks: 79 passed / 4 optional skips. PolicyEngine's downstream adapter also passes all 30 focused tests against this exact Microcosm head and rulespec-nz 3b663b3 using the real HDF5 codec and native Axiom runtime. These are integration fixtures, not a calibrated NZ population or published budget score. CI remains the merge gate. |
This was referenced Oct 7, 2026
Contributor
Author
|
Closing in favour of the graph-native Microcosm NZ v0 (Max's ruling d962; plan: NZ hub notes/1007/microcosm-nz-graph-plan.md). This PR predates packages/microcosm-graph, and its operation lists in build/nz/source_stages.json belong to the stage runtime the graph replaced. The live parts are split out:
|
MaxGhenis
added a commit
that referenced
this pull request
Oct 7, 2026
Finding 1. The hold-out check is now a validator, holdout_leaks(cal, precal, holdout), in test_nz_spec_package.py. A calibration or pre-calibration reference leaks when it: - has a hold-out name; - shares a Ledger identifier with a hold-out reference; - resolves through a selector that one fact could satisfy together with a hold-out selector; - belongs to a hold-out family. Identifiers are compared across the fact-key and source-record-id fields, as ledger_targets._ledger_fact_index resolves them. Selectors include scaled_by_ratio operands (reference_fact_selectors). As in _fact_matches_selector, a list matches by membership and an empty value is a wildcard. A Hypothesis test plants a fresh name, identifier, selector or family in either upstream set and requires a refusal. It uses cross-field identifiers, five overlapping selector shapes, and ratio operands on either side. The example test on the committed data stays, and each row is now asserted to be identity with no operands. Finding 2. Every refusal branch of crosswalk_errors, scenario_errors and the modal_area tie check has a Hypothesis negative case. The new cases found two holes, now closed: - a crosswalk ta_code of None or a non-string passed; - a rent_stock_factor of inf or True passed as a positive number. Finding 3. The rulespec-commit scanner now treats a 7-40 hex run as a mention when it is: - anywhere in a string that contains "rulespec" (any case); - any key or value below a key containing "rulespec"; - in a record whose repository contains "rulespec", in any spelling; - under a key containing "commit" outside a foreign record; - an all-digit SHA that YAML reads as an integer, or a hex mapping key. The string rule is wider than the requested 40-character window, because the package's own "git diff 6fe181fc 8dc2507" sentence sat 102 characters from "rulespec". A test plants five phrasings into the package payloads: the two the review found and three found by the review of this fix-up. On the committed package the scanner still finds only the two reviewed commits. The docstring names both reviewed commits instead of "the one pin". Finding 4 and further wording: - axiom_rules_bindings.json, as_rate_bridge.json, benefit_unit_rule.json, currency_bridge.json, export_contract.json, source_stages.json and spec/sources.yaml no longer describe unwritten code (packages G1, G4, G5a, G6 and G7) as running. - MC11a reads "proposed (awaiting D1)". - benefit_unit_rule.json cites executor _structural_columns and _validate_population_declaration: family_household_id is not structural. The unreachable parent_2 branch is replaced by the parent-1 rule and the parent_order check. - docs/nz-take-up-engine-plan.md: the "nz appears only in" line, the crosswalk row (a placeholder with rows: []), the existing geography kernels and the MC11a wording. - docs/nz-calibration-targets.md drops "only" from the TAWA WFF sentences and marks DistributionExplorer as a comparator, not a calibration target. - release_contract.json no longer credits reform_validation.json with #821's official-score comparisons. source_stages.json is re-receipted in sources.yaml, and the NZ golden is regenerated. Tests: 322 passed across the PR's nine targeted test files. test_nz_spec_package.py has 70 tests, up from 44. Before the round-2 review, all 57 mutants of the validators turned a test red. On the code round 2 changed, 10 of 10 did; an 11th survived only because its clause was redundant, and that clause is removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Oct 10, 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
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.
Fixes #820. Related to #343; the broader NZ population, calibration, publication, and official-score replication work remains open.
Scope
Verification
The fixture reproduces the upstream entitlement deltas 568.50 and 438.50 NZD and uses Frame-resolved family weights [2, 7, 7]; weighted fixture totals are 1,137 and 3,069.50 NZD. These are test results, not NZ population estimates.
Not a release
No donor population was downloaded, no calibration ran, and no Hugging Face release or dashboard data was fabricated. The country kernels remain contract-only and the planner refuses missing implementations.
Before the first solvable candidate: authenticate the donor; implement shared donor/family/geography operators; bind the 11 evidenced NZD/family/period fields and receipt states; activate cell-level facts with reviewed period/geography bridges and country thresholds. Treasury source activation remains pending immutable corpus publication and protected validation. Before an official fiscal-score replication, bridge annual entitlement to the published fiscal-year cash/accrual quantity, including impairment and payment tails.