fix(projection): rank competing claims instead of overwriting
All checks were successful
CI Smoke / host-smoke (push) Successful in 0s
CI Smoke / container-smoke (push) Successful in 1s
Build and Publish Multi-Context Image / build-and-push (push) Successful in 24s

Two rows can claim one record. `cust-wp-0010` claimed CUST-WP-0010 by its own
identifier while `workstream-lifecycle-documentation` claimed it by path, both
pointing at that workplan's file; `kont-wp-0013` and
`KONT-WP-0013-blob-storage-content-streaming` did the same. Assigning into
`matched` unconditionally let whichever row came last win and dropped the other
silently — never matched, so never stale, so never reported by any pass. It was
also order-dependent, so which row survived depended on row iteration order.

Claims are now ranked: derived UUID, then the record's own identifier, then
backing path, then prefix heuristic. The strongest wins and the loser is
displaced into `stale`, which makes it a retirement candidate rather than
invisible. Ranking is order-independent, verified both ways round.

Under ADR-007 the identifier is the identity, so the row naming the record wins
and its stale title and path are repaired from the file; the hub-first row that
never held the identifier retires with its history intact.

751 pass.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 2583210@bnt-lap001
Assistant-Session: f2bff2d5-e9b2-4338-92ca-10282a927006
This commit is contained in:
tegwick 2026-08-28 15:01:58 +02:00
parent 1af9ec2f0c
commit 218f7fd8b5
2 changed files with 63 additions and 7 deletions

View file

@ -720,17 +720,41 @@ async def reset_repository_projection(
r.slug = _tombstone_slug(r.slug or str(r.id), r.projection_retired_at)
outcome.released.append(before)
# How a row came to claim a key, strongest first. Two rows can claim the
# same record — `cust-wp-0010` by its own identifier and
# `workstream-lifecycle-documentation` by path, both pointing at
# CUST-WP-0010's file. Assigning into `matched` unconditionally let the
# later one win and dropped the other silently: never matched, so never
# stale, so never reported by any pass.
UUID_MATCH, SLUG_MATCH, PATH_MATCH, PREFIX_MATCH = 0, 1, 2, 3
matched: dict[str, Any] = {}
claim: dict[str, int] = {}
displaced: list[Any] = []
def _claim(key: str, row: Any, strength: int) -> None:
held = claim.get(key)
if held is None:
matched[key], claim[key] = row, strength
return
if strength < held:
# The new claim is stronger; the incumbent loses the record and
# becomes a retirement candidate rather than disappearing.
displaced.append(matched[key])
matched[key], claim[key] = row, strength
else:
displaced.append(row)
for row in rows:
key = (row.slug or "").strip().lower()
by_uuid = want_by_uuid.get(str(row.id))
if by_uuid is not None:
matched[by_uuid] = row
_claim(by_uuid, row, UUID_MATCH)
continue
if key in want:
# The identifier itself derives again — including for a retired row,
# which is then deliberately un-retired below.
matched[key] = row
_claim(key, row, SLUG_MATCH)
continue
if row.projection_retired_at is None and RETIRED_SLUG_MARK in (row.slug or ""):
# A stamped slug with a cleared flag is a resurrected row: the mark
@ -782,10 +806,12 @@ async def reset_repository_projection(
continue
bp = row.backing_relative_path
if bp and _path_key(bp) in want_paths:
key = want_paths[_path_key(bp)]
else:
cand = [k for k in want if key.startswith(k + "-")]
key = cand[0] if len(cand) == 1 else key
_claim(want_paths[_path_key(bp)], row, PATH_MATCH)
continue
cand = [k for k in want if key.startswith(k + "-")]
if len(cand) == 1:
_claim(cand[0], row, PREFIX_MATCH)
continue
matched[key] = row
# An identifier this repository would create may already belong to another
@ -866,7 +892,7 @@ async def reset_repository_projection(
stale = [
r for k, r in matched.items()
if k not in want and r.projection_retired_at is None
]
] + [r for r in displaced if r.projection_retired_at is None]
if stale and not derived.retirement_eligible:
# Acknowledgement cannot authorise this. The caller is confirming that
# records which stopped deriving should be retired; here nothing has

View file

@ -949,3 +949,33 @@ class TestRetirementIsNotUndoneByPath:
zombie = src.index("projection_retired_at is None and RETIRED_SLUG_MARK in")
tail = src[zombie:zombie + 700]
assert "matched[key] = row" in tail, "a resurrected row must enter `matched`"
class TestCompetingClaims:
"""Two rows can claim one record; the weaker must not vanish.
`cust-wp-0010` claimed CUST-WP-0010 by its own identifier while
`workstream-lifecycle-documentation` claimed it by path. Assigning into
`matched` unconditionally let the later one win and dropped the other
silently never matched, so never stale, so never reported by any pass.
"""
def test_claims_are_ranked_not_overwritten(self):
import inspect
src = inspect.getsource(fp.reset_repository_projection)
assert "def _claim(" in src
assert "displaced.append" in src
def test_a_displaced_row_becomes_stale(self):
import inspect
src = inspect.getsource(fp.reset_repository_projection)
stale_at = src.index("stale = [")
assert "displaced" in src[stale_at:stale_at + 400]
def test_identifier_beats_path(self):
"""A row naming the record outranks one that merely shares its file."""
import inspect
src = inspect.getsource(fp.reset_repository_projection)
assert src.index("UUID_MATCH, SLUG_MATCH, PATH_MATCH, PREFIX_MATCH = 0, 1, 2, 3") > 0
# Strength is ordered strongest-first, so a lower number wins.
assert "if strength < held:" in src