Freegle / Iznik / 21646

71%

push

circleci

fix(test): seed closestPostcode sentinels in open sea so they survive a populated KNN index

The closestPostcode / findClosestPostcodeId integration tests seed a sentinel
postcode into the live spatial 'postcodes' index and assert it's returned as
nearest. They seeded at real UK cities (Aberdeen, Nottingham, Edinburgh, London),
which only works when the index is otherwise empty.

Until PR #767 the postcodes loader aborted on prod-shaped data (ST_X on a
non-point ourgeometry → "unexpected type LINESTRING"), so in CI the index built
empty and the sentinels always won. With the loader fixed, CI's spatial server
now populates 'postcodes' from a real DB, so a real postcode out-competes the
city-located sentinel:
  - LocationIdTest offset test got real id 1687432, not sentinel 99000011;
  - LocationTest got a real id absent from the *test* DB, so the by-id enrich
    returned null (assertNotNull failed).

Move every seeded sentinel + its query point into empty open sea (central North
Sea, ~lat 56.5–56.9 lng +3.0) — >300km and far past the KNN's largest
0.32°/~35km buffer from any real UK postcode. The sentinel is then unambiguously
nearest whether the index is empty or full. Verified against the live populated
index: all five points return no real postcode; seed→exact/offset KNN→remove
returns the sentinel id then clears.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

10980 of 14580 branches covered (75.31%)

Branch coverage included in aggregate %.

118915 of 168850 relevant lines covered (70.43%)

36.25 hits per line