Repository navigation
Conversation
vahid-ahmadi
left a comment
There was a problem hiding this comment.
Two rebase notes. This conflicts with main (14181d7) in chronicle/source_package.py and tests/test_chronicle_bundle.py.
1. The Ostbelgien package duplicates main's ostbelgien-kindergeld-2025 (#389). packages/ostbelgien/child_benefit_caseload_2025_12/source_package.yaml registers the December 2025 caseload of 15,533 children and 8,302 recipients (db/data/ostbelgien/child_benefit_caseload_2025_12/ostbelgien_child_benefit_caseload_2025_12.csv:2-3). Main already registers the same two figures from the same ostbelgienstatistik.be article in packages/ostbelgien/kindergeld_2025/source_package.yaml:63 and :269, using a pinned capture of that article. The keys differ:
| This PR | Main (#389) | |
|---|---|---|
| concept | ostbelgien.child_benefit_paid_children / ostbelgien.child_benefit_payment_recipients (:92, :173) |
ostbelgien.kindergeld.children.count / ostbelgien.kindergeld.recipients.count (:256, :467) |
| geography | BE-DG (:32) |
ostbelgien:deutschsprachige_gemeinschaft (:77) |
| period | month 2025-12 (:24-25) |
calendar_year 2025, with period_coverage 2025-12-01 to 2025-12-31 (:69-73) |
After a rebase, the same observation would appear twice under keys that are not linked. Main's copy has better custody: an article capture, where this PR has a two-row transcription because the publisher returned 403. So this package can be dropped, or folded into #389's package.
2. The new Belgian packages need adding to main's BE bundle list. Main's curated suite (chronicle/bundle.py:563-590) is checked against every alias under BE_BUNDLE_SOURCE_PREFIXES (:529-541) by assert_be_bundle_sources_match_aliases (:730). tests/test_source_package_alias_drift.py:425-428 also pins len(BE_BUNDLE_SOURCES) == 26. Three of the new aliases (chronicle/source_package.py:127-142 at this head) sit under the sfpd, opgroeien and ostbelgien prefixes. Unless they are added to BE_BUNDLE_SOURCES, the drift guard fails, and with it build-bundle --suite be. If the Ostbelgien package is dropped as in point 1, that leaves two. iriscare and parlement_wallonie are not BE prefixes, so those two packages stay out of the BE suite unless their prefixes are added on purpose.
Review drafted with AI assistance and checked against the code at 0f75a2b.
Summary
These are publisher facts with provenance only. They do not assert legal eligibility, participation rates, take-up behavior, calibration mappings, or simulator concepts.
Scope semantics
Opgroeien, Iriscare, Ostbelgien, and the Walloon response describe administrative scheme populations. They are not relabeled as Belgian or NUTS1 population totals. Downstream model layers must supply an explicit typed scheme-to-population mapping before using them as calibration targets.
The Opgroeien and Iriscare dashboard captures are pinned to the exact source bytes used for extraction. The SFPD and Walloon records retain their official source documents and selectors.
Validation
uv run pytest -q tests/test_chronicle_bundle.pypasses: 9 passedCompanion
PolicyEngine/microcosm#824 consumes only the GRAPA reference and intentionally leaves the child-scheme facts unbound until typed scheme-population support exists.