Files
eezd c79522ee7c
Development CI / check-and-build (push) Failing after 2m46s
Development CI / deploy-development (push) Skipped
GitHub Git mirror / mirror (push) Successful in 6s
Fix Jin10 calendar backfill, retries and withdrawn schedules
Refresh historical previews after the local day ends and isolate retries by date without resetting browser scheduling progress.

Preserve withdrawal and restoration revisions, exclude withdrawn fallback values, and cover the behavior with isolated regression checks.
2026-10-31 13:40:38 +08:00

156 lines
9.8 KiB
Python

"""Exercise calendar persistence with synthetic credentials in the isolated CI database."""
import asyncio
from copy import deepcopy
from datetime import UTC, date, datetime, timedelta
import json
import os
from pathlib import Path
from sqlalchemy import func, select
from alphapulse.calendar.jin10 import normalize
from alphapulse.calendar.jin10_history import calendar_history
from alphapulse.calendar.jin10_store import next_history_day, save_entry
from alphapulse.calendar.jin10_view import jin10_item, merge_items
from alphapulse.calendar.jin10_worker import ConnectionChanged, persist_day
from alphapulse.calendar.parsers import Scheduled
from alphapulse.calendar.service import release_rows, save_raw, save_result, save_schedule
from alphapulse.core.crypto import encrypt
from alphapulse.core.db import SessionLocal, engine
from alphapulse.models import EconomicRelease, EconomicRevision, RawResponse, Source
from alphapulse.models.jin10 import Jin10Connection, Jin10SyncDay
async def verify():
assert os.environ["DATABASE_URL"].startswith("postgresql+asyncpg://ci:ci-only@"), "Calendar acceptance requires the isolated CI database"
fixture = json.loads((Path("tests/fixtures/calendar/jin10.json")).read_text())
now = datetime(2026, 11, 5, tzinfo=UTC)
connected = now - timedelta(days=1)
encrypted = encrypt("isolated-ci-calendar-account")
async with SessionLocal() as session:
row = await session.get(Jin10Connection, 1)
row.token_encrypted, row.connected_at = encrypted, connected
assert (await session.get(Source, "jin10")).collector == "calendar"
await session.commit()
day = date(2026, 11, 4)
responses = {"data": fixture[str(day)], "event": fixture[f"event:{day}"], "holiday": fixture[f"holiday:{day}"]}
await persist_day(day, responses, encrypted, connected, now)
async with SessionLocal() as session:
rows = list(await session.scalars(select(EconomicRelease).where(EconomicRelease.source == "jin10")))
assert len(rows) == len(responses["data"]["list"]) + len(responses["event"]["list"])
adp = next(r for r in rows if "jin10_78" in r.values)
assert adp.first_received_at is None
assert not adp.values["jin10_78"]["forecast_before_release"]
assert adp.values["jin10_78"]["observed_at"]["forecast"] == now.isoformat()
assert (await session.get(Jin10SyncDay, day)).counts == {"data": len(responses["data"]["list"]), "event": 8, "holiday": 0}
before = await session.scalar(select(func.count()).select_from(EconomicRevision))
assert await next_history_day(session, date(2025, 1, 1), day) == date(2025, 1, 1)
await persist_day(day, responses, encrypted, connected, now + timedelta(minutes=1))
async with SessionLocal() as session:
assert await session.scalar(select(func.count()).select_from(EconomicRevision)) == before
try:
await persist_day(day, responses, b"obsolete-credential", connected, now)
raise AssertionError("Obsolete connection was accepted")
except ConnectionChanged:
pass
invalid = deepcopy(responses)
invalid["event"]["date"] = "2026-11-05"
async with SessionLocal() as session:
raw_before = await session.scalar(select(func.count()).select_from(RawResponse))
coverage_before = (await session.get(Jin10SyncDay, day)).captured_at
try:
await persist_day(day, invalid, encrypted, connected, now + timedelta(hours=1))
raise AssertionError("Mismatched response date was accepted")
except ValueError:
pass
async with SessionLocal() as session:
assert await session.scalar(select(func.count()).select_from(RawResponse)) == raw_before + 3
assert (await session.get(Jin10SyncDay, day)).captured_at == coverage_before
# Existing preview checkpoints must be revisited after downtime, without a schema migration.
async with SessionLocal() as session:
checkpoint = await session.get(Jin10SyncDay, day)
checkpoint.captured_at = now - timedelta(days=2)
await session.commit()
assert await next_history_day(session, day, day) == day
assert await next_history_day(session, day, day, excluded={day}) is None
await persist_day(day, responses, encrypted, connected, now + timedelta(minutes=2))
async with SessionLocal() as session:
assert await next_history_day(session, day, day) is None
other_day_raw = await save_raw(session, "jin10", "https://rili.jin10.com/day/2026-11-03/", 200, {}, b"isolated-other-dates", now)
other_entries = [normalize(fixture["2025-01-08"], "data", date(2025, 1, 8))[0],
*normalize(fixture["holiday:2026-11-03"], "holiday", date(2026, 11, 3))]
for entry in other_entries:
await save_entry(session, entry, other_day_raw, now)
await session.commit()
revision_count = await session.scalar(select(func.count()).select_from(EconomicRevision))
# Removing an economic release and event withdraws only those identities, once per change.
removed = deepcopy(responses)
adp_identity = adp.external_id
removed["data"]["list"] = [r for r in removed["data"]["list"] if f"data:{r['data_id']}" != adp_identity]
removed_event = removed["event"]["list"].pop(0)
removed_ids = {adp_identity, f"event:{removed_event['id']}"}
await persist_day(day, removed, encrypted, connected, now + timedelta(minutes=3))
async with SessionLocal() as session:
rows = list(await session.scalars(select(EconomicRelease).where(EconomicRelease.source == "jin10")))
withdrawn = [r for r in rows if r.external_id in removed_ids]
assert len(withdrawn) == 2 and all(r.schedule_status == "withdrawn" and r.revision_no == 2 for r in withdrawn)
assert all(jin10_item(r, now)["status"] == "withdrawn" for r in withdrawn)
assert all(r.schedule_status == "scheduled" and r.revision_no == 1 for r in rows if r.external_id not in removed_ids)
evidence = list((await session.execute(select(EconomicRevision, RawResponse).join(
RawResponse, RawResponse.id == EconomicRevision.raw_response_id,
).where(EconomicRevision.release_id.in_([r.id for r in withdrawn]), EconomicRevision.revision_no == 2))).all())
assert len(evidence) == 2
assert all(revision.payload["schedule_status"] == "withdrawn" and raw.url.endswith("#" + owner.external_id.split(":")[0])
for revision, raw in evidence for owner in withdrawn if owner.id == revision.release_id)
assert await session.scalar(select(func.count()).select_from(EconomicRevision)) == revision_count + 2
await persist_day(day, removed, encrypted, connected, now + timedelta(minutes=4))
async with SessionLocal() as session:
assert await session.scalar(select(func.count()).select_from(EconomicRevision)) == revision_count + 2
await persist_day(day, responses, encrypted, connected, now + timedelta(minutes=5))
async with SessionLocal() as session:
restored = list(await session.scalars(select(EconomicRelease).where(
EconomicRelease.source == "jin10", EconomicRelease.external_id.in_(removed_ids),
)))
assert all(r.schedule_status == "scheduled" and r.revision_no == 3 for r in restored)
assert next(r for r in restored if r.external_id == adp_identity).id == adp.id
assert await session.scalar(select(func.count()).select_from(EconomicRevision)) == revision_count + 4
# Join an official weekly release to the exact provider series, including its revised previous value.
claims_day = date(2026, 10, 29)
entries = normalize(fixture[str(claims_day)], "data", claims_day)
claims = next(e for e in entries if "jin10_79" in e.values)
async with SessionLocal() as session:
raw = await save_raw(session, "jin10", "https://rili.jin10.com/day/2026-10-29/", 200, {}, b"isolated-claims", now)
for entry in entries:
await save_entry(session, entry, raw, now)
official_raw = await save_raw(session, "dol", "https://www.dol.gov/claims.pdf", 200, {}, b"isolated-official", now)
event = Scheduled("isolated-claims", "initial_claims", "Official initial claims", claims.at, "scheduled", "2026-10-24")
await save_schedule(session, "dol", "https://oui.doleta.gov/unemploy/archive.asp", [event], official_raw, now, withdraw_missing=False)
official = await session.scalar(select(EconomicRelease).where(EconomicRelease.source == "dol", EconomicRelease.external_id == event.external_id))
save_result(session, official, {"initial_claims": {"actual": "19.40", "previous": "19.5"}}, "https://www.dol.gov/claims.pdf", official_raw, now)
await session.commit()
extra = next(r for r in await session.scalars(select(EconomicRelease).where(EconomicRelease.source == "jin10")) if "jin10_79" in r.values)
item, = merge_items([(official, release_rows(official, now)[0])], [extra], now)
assert item["actual"] == "19.40" and item["forecast"] == "19.9"
assert item["previous_original"] == "19.4" and item["previous_revised"] == "19.5"
assert item["field_sources"]["actual"]["source"] == "dol"
assert item["field_sources"]["forecast"]["source"] == "jin10"
assert item["forecast_before_release"] is False
history = await calendar_history(session, official, "initial_claims", 36)
assert len(history) == 1 and history[0]["actual"] == "19.40" and history[0]["forecast"] == "19.9"
assert history[0]["first_recorded_actual"] == "19.40"
assert history[0]["first_received_at"] == now
print("Calendar acceptance passed: three categories, exact official joins, revisions, raw evidence, late forecasts, post-day coverage, scoped withdrawal/restoration and obsolete connection rejection")
async def main():
try:
await verify()
finally:
await engine.dispose()
asyncio.run(main())