SAME format (FCC standard) is PSSCCC: P=area-type indicator, SS=state FIPS,
CCC=county FIPS. Pre-v0.10.7 _build_regions read code[:2] (PS) as the state,
so the leading P=0 collapsed every SAME code to state 01 = Alabama. The
dumped CAP envelope from CENTRAL_WX#968 showed Bannock County Idaho alerts
flagged as US-AL-FIPS016005 + primary_region=US-AL-FIPS016005, which then
routed to central.wx.alert.us.al.county.fips016005 instead of
central.wx.alert.us.id.county.fips16005.
Fix:
- Slice code[1:3] for the state FIPS, code[1:] for the 5-digit ANSI county
FIPS (SSCCC -- standard interoperable form). Drops the P padding from
emitted region/subject; P stays preserved verbatim in data.geocode.SAME
for any power user that needs it.
- Length guard tightened: ==6 + isdigit + isinstance str (was >= 2). Now
malformed entries (too short, too long, non-digit, None) are silently
skipped with no crash.
- Deleted dead _extract_states_from_codes (defined but never called; same
bug, removed rather than fixed).
Tests:
- New TestSameStateParse parametrized over 4 distinct-state cases per spec:
016005 -> US-ID-FIPS16005 (Bannock area), 001005 -> US-AL-FIPS01005
(Autauga area), 056005 -> US-WY-FIPS56005 (Carbon area), 049005 ->
US-UT-FIPS49005 (Cache UT).
- Area-subset (P>=1) and unknown-state-FIPS coverage.
- Malformed-input parametrize: empty, too short (2 forms), too long (7
digits), non-digit char, all-alpha, None -- each silently skipped.
- Existing SAMPLE_FEATURE_* fixtures updated from constructed-to-match-bug
values (160001/410051/060037/530033) to proper 0SSCCC format
(016001/041051/006037/053033); existing TestBuildRegions assertions
updated to expect 5-digit ANSI form.
Followup ticket (NOT v0.10.7 scope, recorded in PR body):
(a) Null-geometry alerts with valid Idaho UGC zones are silently dropped
by _geometry_intersects_region (line 297-298): needs UGC-fallback or
geometry-or-UGC check. NWS issues many Special Weather Statements
without GeoJSON polygons but with UGC IDZ* zones that should pass.
(b) Configured monitoring bbox north=44.5 only covers the southern third
of Idaho; Idaho extends to 49.0N, so Coeur d'Alene / Lewiston / etc.
are out of scope. Verify whether the narrow bbox was an intentional
dev limit or accidental.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
NWSAdapter.sweep_old_ids ran an unscoped global DELETE FROM published_ids
(no adapter = ? filter), so each per-adapter sweep cycle purged EVERY
adapter dedup rows older than 8 days -- capping dedup memory at ~8d for the
8 adapters whose window exceeds it (eonet/gdacs/nwis 30d; swpc_kindex/
protons/alerts + wfigs_incidents/perimeters 14d). Events that went silent
past 8d then re-listed were re-published as downstream duplicates.
Extract the 3 dedup methods onto the SourceAdapter base (finishing v0.9.19;
is_published/mark_published were byte-identical) and preserve the 8-day
window via dedup_sweep_days = 8, so the inherited sweep scopes to adapter=?.
Effect: each adapter retains dedup to its configured window -> fewer
downstream duplicates. Retention-window scoping, NOT a dedup-forward-only
change (no event-shape change, no migration/backfill).
Adds a regression guard asserting a foreign adapter row survives an nws
sweep. Suite 900->901/1.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Add stream_name, subject_for(), and settings_schema() to SourceAdapter ABC
- Implement all three methods in NWSAdapter, FIRMSAdapter, USGSQuakeAdapter
- Replace manual _ADAPTER_REGISTRY with pkgutil.iter_modules auto-discovery
- Remove subject_for_event from models.py (each adapter owns its subject logic)
- Update supervisor to use adapter.subject_for(event) instead of helper
- Fix quake events going to wrong stream (was publishing to CENTRAL_WX)
- Update test files to use adapter methods
This fixes the quake stream bug where events were published to
central.wx.alert.us.unknown instead of central.quake.event.<tier>.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* test(bootstrap): isolate env vars in test_reads_from_env_file
The test was failing on CT104 because live CENTRAL_DB_DSN
environment variable overrode the test .env file content.
Fix: use monkeypatch.delenv to clear all CENTRAL_* env vars
before creating the Settings object, ensuring the test env
file is the only source of configuration values.
Also add CENTRAL_CSRF_SECRET to test env file since it's
now a required field.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* test(models): remove stale test_custom_prefix test
The test called subject_for_event(event, prefix="myapp.events")
but the prefix parameter was removed from the API.
The prefix functionality was intentionally removed - subjects
now always use the "central." prefix hardcoded in the function.
Delete the test rather than re-add the parameter.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* test(nws): update fixtures for new adapter signature and region filtering
NWSAdapter.__init__ signature changed from (config, cursor_db_path)
to (config, config_store, cursor_db_path) with config now being
AdapterConfig with a settings dict instead of NWSAdapterConfig.
Also adapts tests to region-based bbox filtering:
- TestStateFilter now uses region bbox to accept PNW, reject CA
- Add geometry to SAMPLE_FEATURE_OR so it passes region filter
- Other test fixtures use region=None to skip filtering
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Matt Johnson <mj@k7zvx.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
