Repository navigation
Bind Brussels and Flemish post-18 education flags to enrolment - #1611
Merged
Merged
Conversation
MaxGhenis
force-pushed
the
be/pr4a2-post18-enrolment
branch
3 times, most recently
from
November 15, 2026 15:24
2d8247b to
5aab210
Compare
Brussels article 25 §2(b) and Flemish article 8 §2 item 3 pay family benefits to a child in education up to 25. PR 4a defaulted both flags to false, a lower bound that paid nothing from 18 on. This binds each flag, like Wallonia's and the German-speaking Community's, to a person_role predicate over fact:person.education_enrollment's six enrolled bands, in its scheme's eligibility module and in the regional routing bridge. The closure goes from 35 encoded and 242 defaulted roots to 39 and 238. Six defaults keep their values and get reasons that say what the new bindings change; the Flemish residence-right reason names the age groups before its exception. A new test pins all four regions' post-18 flags to one enrolment predicate. The coverage and country-spec goldens are regenerated, and the transport ADR now says encode runs Belgium's person_role bindings and defers only its 11 other group bindings. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MaxGhenis
force-pushed
the
be/pr4a2-post18-enrolment
branch
from
November 15, 2026 19:19
5aab210 to
941b8e3
Compare
MaxGhenis
marked this pull request as ready for review
November 15, 2026 23:01
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Ready. #1602 (PR 4a) merged as e555b4d, and this PR targets
main: its one commit, 941b8e3, is 5aab210 moved onto main, with the same patch (git patch-id). It changes what Belgium's closure feeds, so it needed a Subfleet hard-tier review before merge; that review approved it (see Verification). CI onmain: run 43751784209 passed all 12 jobs after one rerun; engine-free (3.13) had first failed on a wall-clock assert intest_precalibration_ancestry_composed_stage_scan_is_linear, a synthetic test this PR doesn't touch (2.13 s against 1 s; the test's exact-count assert passed). NZ's #1629 removes that timer.What this changes
Brussels and Flanders pay family benefits past 18 to a child in education, up to 25:
brussels_family_benefits_child_in_qualifying_education_training_or_stage.flanders_family_benefits_child_is_pupil_student_intern_or_school_leaver.PR 4a defaulted both flags to false. That was a lower bound that paid nothing to anyone aged 18 or over in either region. Wallonia's and the German-speaking Community's flags were already bound to
fact:person.education_enrollment. This PR binds the Brussels and Flemish flags the same way. Each flag gets aperson_rolepredicate binding over the six enrolled ISCED bands (early_childhoodtotertiary), with relationapproximate. Each flag is bound in its scheme's eligibility module and inbe/statutes/family_benefits/regional_routing.yaml, which reads the regional roots through its imports.axiom:bebindingsDefaults whose reasons changed
No default changes value. Six reasons change because the new bindings made them stale or incomplete. The formulas at rulespec-be
632e1da1are in be-brueligibility.yaml(the article 25 gate and the article 4 exclusion) and be-vlgeligibility.yaml(the article 8 age and status windows and the two exclusions).flanders_family_benefits_child_in_month_turning_25(false). Its old reason said the flag changed nothing while the student flag stayed false. The student flag now reads enrolment, so false ends an enrolled child's window at the age the frame reports, before the end of the birth month. This is a lower bound.brussels_family_benefits_child_has_foreign_study_training_volunteer_or_au_pair_stay_authorization,flanders_family_benefits_child_has_excluded_study_training_volunteer_or_au_pair_residence_rightandflanders_family_benefits_child_raised_or_studies_outside_belgium(all false). Each exclusion still removes nobody, so each default overstates entitlement for the children it should exclude. That overstatement used to cover children under 18. It now also covers enrolled children aged 18 to 24. Each reason names the saving clause or exemption that keeps some of those children entitled: a Brussels right that was open on 31 August 2025, a Flemish child of a study, training, volunteer or au-pair permit holder, or a Flemish Government exemption. The routing bridge's copy offlanders_family_benefits_child_raised_or_studies_outside_belgiumkeeps Add Belgium's Child rules bindings and Axiom input closure (PR 4a), still inactive #1602's wording, which says the bridge's region default (99) routes no child, so this PR changes only that name's eligibility-module entry.brussels_family_benefits_child_is_apprenticeandbrussels_family_benefits_child_prepares_higher_education_thesis(false). Each lower bound now misses only the children whom the enrolment fact doesn't count, as the Walloon apprentice reason already says.The Flemish binding note says the enrolment fact doesn't record where a child studies, so a child taught abroad reads as enrolled. Article 8 §3's exclusion (
flanders_family_benefits_outside_belgium_exclusion_applies) reads a separate input. That input stays uncovered and the closure defaults it to false, so nothing in the model applies the exclusion.Also regenerated
docs/concept-coverage/axiom-be.mdandpackages/microcosm-frame/tests/golden/concept-coverage/axiom-be.json, viatools/refresh_concept_coverage.py --engine axiom-be. Concept-fed inputs go from 105 to 109 and uncovered inputs from 1,762 to 1,758.packages/microcosm-build/tests/golden/be_country_spec.json(the closure's hash and the fingerprint).changelog.d/be-post18-enrolment-bindings.changed.md.docs/concept-schema-transport-adr.mdcounted all of Belgium's group bindings (46 of 105) as deferred.ConceptMapping.encoderunsperson_rolebindings on the person table, though (is_executablereports them as executable). The ADR now says thatencoderuns Belgium's 39person_rolebindings and that its 11 other group bindings, out of 109, stay deferred.Tests
test_be_axiom_input_closure.pytakes the new counts. It also gainstest_the_post_18_education_flags_read_one_enrolment_predicate, which pins three things for all four regions' post-18 education flags:person_role,approximatebinding that reads the same predicate over the same six bands;Verification
Independent review (Subfleet hard tier, job 20261113-190022-be-pr4a2-review-sf, at 8890611): approve, with nothing at P0 to P2 and three P3 notes. Two are taken: the Flemish residence-right reason now names the age groups before its exception, and this body credits
encode, notis_executable, with running the bindings. The third (the new test's closure check compares a set of pairs, so it can't see one missing entry) needs no change: two existing tests catch that case.Rebased on c90f5c9. #1602's round-2 fixes (5fd86ad and before) move three defaults (the routing region code to 99, and the German-speaking extension's months used and its Article 9 flag), and c90f5c9, which answers its second review, rewords 22 reasons, the closure description and its judgment policy. The only one of those entries that carries one of this PR's names is the routing bridge's copy of
flanders_family_benefits_child_raised_or_studies_outside_belgium, which now keeps #1602's bridge wording. The closure was rebuilt from this PR's rulings plus #1602's round-2 and round-3 scripts, not merged by text. Against c90f5c9 it changes the same eight names in 15 entries (two each, but one for that name), and the closure description only in its counts (35 to 39 encoded, 242 to 238 defaulted). The other six files' changes are line for line as before.Locally, on
b81ac72c6plus this change (as first pushed, 8890611), with Python 3.14.7:tools/refresh_concept_coverage.py --engine axiom-be --check, a regeneration check ofbe_country_spec.json, and a rebuild of the closure from its rulings all report no drift.test_be_axiom_input_closure.py(19)test_axiom_concept_mapping.py(15)test_concept_mapping.py(287)test_axiom_adapter.py(20, 25 skipped)test_axiom_graph_adapter.py(165, 5 skipped)test_country_spec.py(144)test_spec_source_citations.py(138)test_spec_engine_country_bundles.py(22)test_spec_only_country_packages.py(37)test_gate_battery.py(30)test_be_target_references.py(156, 32 skipped)test_transport_holdout_separation.py(51)test_spec_engine_reemission.pypasses 7 of 8 tests. The eighth stops onduplicate installed distribution identity: microcosm-emitter-interface, before any hash check, because the local path carried three copies of that package. PR 4a's local run hit the same error, and CI passed the file for Add Belgium's Child rules bindings and Axiom input closure (PR 4a), still inactive #1602 at this PR's base. CI only runs on pull requests intomain, so this one got its first run, in a clean environment, when it retargeted.At 941b8e3 (on
mainat e555b4d, whose Belgian files equal c90f5c9's): the closure rebuilt from the rulings, the regeneratedbe_country_spec.jsonand the coverage check all show no drift. The earlier rebase, 2d8247b, passed the citation guards intest_spec_source_citations.pyand, at 0a08144, which differs from it only by 15 commit pins, the closure test's 19 other tests. Local test runs wait until host load drops below 100; CI onmainis the test run.Two in-session review passes checked the rows, each reason against its formula, and the counts. They don't replace the independent review this PR still needs. The first pass found stale reasons, a missing
relationassertion and the ADR's deferred count. The second found only wording fixes. Every upheld finding is in this diff.What this can't check yet
Nothing runs these modules yet. The engine pins are null, no node produces a selector column, and Belgium has no transport graph (see #1602). The bindings are
approximate. The enrolment fact doesn't observe the minimum study intensity, programme list or registration conditions that each region's government sets.🤖 Generated with Claude Code