fix(analyze): cad_exists_in_db requires non-NULL geometry #1034

Merged
bot-backend merged 1 commit from fix/cad-exists-null-geom into main 2026-06-03 15:00:07 +00:00

1 commit

Author SHA1 Message Date
85c43ff68b fix(analyze): cad_exists_in_db must require non-NULL geometry
All checks were successful
Deploy / build-backend (push) Successful in 2m34s
Deploy / deploy (push) Successful in 2m11s
Deploy / changes (push) Successful in 6s
Deploy / build-frontend (push) Has been skipped
Deploy / build-worker (push) Successful in 3m10s
Complements the NULL-geom 500 fix. cad_exists_in_db (docstring: "is there
GEOMETRY") checked only row existence, not geom IS NOT NULL — so for the ~964
meta-but-NULL-geom parcels it returned True. Consequence after the 500 fix:
such a parcel fell into the analyze fallback, find_or_enqueue_fetch step 2 saw
cad_exists_in_db=True → returned ("ready", None) → NO NSPD fetch enqueued →
analyze looped to a 202 with job_id=null and the parcel was stuck "fetching"
forever (never pulled real geometry, never resolved).

Fix: add `AND geom IS NOT NULL` to all three EXISTS branches (aligns the
function with its docstring). Now a NULL-geom parcel → cad_exists_in_db=False →
a real NSPD fetch is enqueued (202 + real job_id) → geometry populates →
re-poll → analyze succeeds (or 404 not_in_nspd if NSPD lacks it). No more
stuck-202. Valid-geom parcels unaffected. All 3 callers want geometry-presence
semantics. 37 analyze/fetch/by-bbox tests green. Refs #944.
2026-06-03 19:59:59 +05:00