feat: 술어 alias 시 원본 술어 기록·추적 — relation-aliases.md + provenance 역추적 (engine-free, read-only)#201
Merged
Merged
Conversation
…ion guard
Add relation_aliases(root) to parse policy/relation-aliases.md into a
{raw: canonical} dict, surface_variants(canonical, aliases) for reverse
lookup, and fail-loud collision detection (self-map, conflicting targets,
alias chains). Absent file returns {} — byte-identical for KBs without
the file. Mirrors sync_ignore_patterns() style and _assert_no_alias_collision
discipline.
When the RELATION term in `factlog provenance` is a declared canonical (surface_variants is non-empty), expand the match to include rows stored under any surface predicate, grouping them separately with a [surface: <raw>] label. When the term is itself a surface predicate, emit a `canonical: <name>` context line. Absent policy/relation-aliases.md -> zero aliases -> no expansion -> output byte-identical to today (regression-guarded by two TestProvenanceNoAliasFile tests and the full existing shell+golden suite).
justinjoy
force-pushed
the
feat/provenance-alias-trace
branch
from
c2baf75 to
d5367b6
Compare
This was referenced Jul 2, 2026
feat(#188 대안): 개선된 A — canonical 질의는 쿼리레이어 확장, 규칙은 별도 canonical/3 emit (relation/3 verbatim 유지)
#227
Closed
justinjoy
added a commit
that referenced
this pull request
Jul 2, 2026
…ema_context A relation query whose relation argument is a declared canonical name (i.e. surface_variants(name, relation_aliases()) is non-empty) now validates as engine-askable in classify_query, even though the stored facts carry surface variants, not the canonical. _relation_match_count is extended to count surface-variant rows when the relation arg is a canonical name, so the validator returns QUERY_OK (not QUERY_FACT_ABSENT) when matching rows exist under any alias. schema_context() gains a "Canonical relation names (prefer these):" section listing each canonical and its sorted variants; absent alias file -> section is empty -> output is byte-identical to today. No engine changes, no accepted.dl writes, no compile_facts changes. Reuses existing relation_aliases()/surface_variants() (#201).
justinjoy
added a commit
that referenced
this pull request
Jul 2, 2026
…evaluate_relation evaluate_relation now detects when the relation argument is a declared canonical name (surface_variants non-empty) and matches fact rows whose relation field equals the canonical OR is in the variant set. This lets a single canonical query (e.g. published_year) return all surface-variant rows (게재연도, publication_year) with their real stored triples and real provenance — no [no extraction backing] markers. Pure query-layer change: no engine writes, no accepted.dl mutation, no compile_facts changes. relation_aliases/surface_variants are imported from common (single source of truth, #201). Absent alias file -> all behavior byte-identical to before (opt-in no-op), pinned by a test. Integration tests added to tests/test_ask_router.sh: canonical validate routes engine with code=ok, evaluate returns both variant rows, render shows real stored triples, no-alias-file behavior unchanged.
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.
무엇을
술어를 canonical로 alias할 때 문서의 원 술어(원본)가 무엇이었는지 기록·추적되게 합니다. 사람이 선언하는
policy/relation-aliases.md+factlog provenance의 역추적으로, canonical 조회 시 그 canonical을 이루는 모든 원 술어 행을[surface: <원술어>]라벨 + source + note와 함께 보여줍니다.이슈 #188(술어 canonical-alias) 검토의 후속 — 술어 위치엔 compound term이 막혔고, alias 레이어의 하드 제약 #2("원 술어·provenance verbatim 저장 불변")를 만족하는 기록/추적 절반을 엔진 없이 먼저 구현합니다.
왜 엔진 없이 성립하나 (3-레이어 원본 보존)
candidates.csv에 원 술어+source가 verbatim 저장 — 무접촉.policy/relation-aliases.md(git-tracked·리뷰).provenance역추적 — alias 테이블 + candidates.csv만으로 host-side 계산.canonical/3엔진 파생 불필요.동작 예
커밋 (아토믹 3개)
feat(common): relation_aliases() parser + surface_variants() + collision guardfeat(provenance): trace canonical predicate back to surface variantsfix(common): harden relation_aliases — NFC-normalize tokens, require -> separator설계 원칙 / 하드 불변식
테스트 / 무회귀
test_provenance.sh19, golden 5/5 byte-identical.리뷰 경위 (다단계 합의)
구현(아토믹) → 코드리뷰에서 🔴 CRITICAL NFC/NFD 불일치(macOS NFD 한글 alias 무증상 실패 — 정확히 타깃 use case)와 🟡 arrow 미검증 적발 → 구현자 수정 + 회귀 테스트 핀 고정 → **아키텍트+비평가 독립 검증 합의(CONCUR/CONCUR)**로 성립 확정.
알려진 비블로킹 후속
cli.py:1148-1160동일 if/else grouping 분기 중복(behavior-neutral) — cosmetic dedup, 별도 정리.비목표
엔진
canonical/3파생·규칙,/factlog askrender의 canonical 병기(arity-3 query 표면 의존), 입력측 정규화, corroboration-join → 후속 슬라이스.