You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
SoulSync/tests/automation/test_schedule.py

520 lines
22 KiB

"""Tests for ``core/automation/schedule.py:next_run_at``.
Pure function over (trigger_type, trigger_config, now_utc, default_tz)
so each case can pin a single rule without monkeypatching the system
clock. Covers the existing engine behaviour (interval, daily, weekly)
plus the new ``monthly_time`` shape PR 1 introduces.
"""
from __future__ import annotations
from datetime import datetime, timedelta, timezone
from zoneinfo import ZoneInfo
import pytest
from core.automation.schedule import next_run_at
# ---------------------------------------------------------------------------
# Helper — clear, timezone-aware datetime construction in test bodies.
# ---------------------------------------------------------------------------
def _utc(year: int, month: int, day: int, hour: int = 0, minute: int = 0) -> datetime:
"""Aware UTC datetime — every ``now_utc`` injection in tests
flows through this so a stray timezone bug is impossible."""
return datetime(year, month, day, hour, minute, tzinfo=timezone.utc)
# ---------------------------------------------------------------------------
# Dispatcher: trigger_type routing.
# ---------------------------------------------------------------------------
def test_returns_none_for_unrecognised_trigger_type():
"""Event-based / unknown trigger types are not scheduled — the
caller should NOT write a next_run for them."""
now = _utc(2026, 5, 27, 12, 0)
assert next_run_at('event', {}, now) is None
assert next_run_at('garbage', {'interval': 1}, now) is None
assert next_run_at('', {}, now) is None
def test_returns_none_for_non_dict_config():
"""Defensive — callers may pass through whatever ``json.loads``
returned. Non-dict configs trigger the fallback path which is
'treat as empty dict + use defaults'."""
now = _utc(2026, 5, 27, 12, 0)
# Interval-typed with garbage config falls back to defaults
# (interval=1, unit='hours') rather than crashing.
result = next_run_at('schedule', None, now)
assert result == now + timedelta(hours=1)
# ---------------------------------------------------------------------------
# Interval (``trigger_type='schedule'``)
# ---------------------------------------------------------------------------
@pytest.mark.parametrize('unit,seconds_per_unit', [
('minutes', 60),
('hours', 3600),
('days', 86400),
])
def test_interval_units(unit, seconds_per_unit):
"""Every supported unit scales the interval correctly. Kept in
lockstep with the engine's existing ``_calc_delay_seconds`` map
— see _INTERVAL_MULTIPLIERS docstring."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('schedule', {'interval': 3, 'unit': unit}, now)
assert result == now + timedelta(seconds=3 * seconds_per_unit)
def test_interval_weeks_unit_falls_back_to_hours_matching_engine():
"""Engine's ``_calc_delay_seconds`` only recognises minutes / hours
/ days — anything else defaults to hours. Drift between this helper
and the engine would silently mis-schedule rows whose config snuck
through with an unsupported unit. Pin the alignment until PR 2
collapses both paths through this function."""
now = _utc(2026, 5, 27, 12, 0)
# 'weeks' is not in our map; falls back to hours.
assert next_run_at('schedule', {'interval': 2, 'unit': 'weeks'}, now) == now + timedelta(hours=2)
def test_interval_unknown_unit_defaults_to_hours():
"""Backward compat with DB rows whose ``unit`` field is missing
or an unrecognised value — engine's historic behaviour was to
treat as hours, and we preserve that."""
now = _utc(2026, 5, 27, 12, 0)
assert next_run_at('schedule', {'interval': 2, 'unit': 'fortnights'}, now) == now + timedelta(hours=2)
assert next_run_at('schedule', {'interval': 2}, now) == now + timedelta(hours=2)
def test_interval_clamps_zero_and_negative_to_one():
"""Without a floor a zero/negative interval would schedule for
the past or fire instantly in a loop. Engine clamped to >=1 via
``max(int(interval), 1)``; we preserve that contract."""
now = _utc(2026, 5, 27, 12, 0)
assert next_run_at('schedule', {'interval': 0, 'unit': 'hours'}, now) == now + timedelta(hours=1)
assert next_run_at('schedule', {'interval': -5, 'unit': 'hours'}, now) == now + timedelta(hours=1)
def test_interval_garbage_interval_falls_back_to_one():
"""Non-numeric ``interval`` → default of 1. Survives a JSON column
where the field was typed as a string by an old admin script."""
now = _utc(2026, 5, 27, 12, 0)
assert next_run_at('schedule', {'interval': 'oops', 'unit': 'hours'}, now) == now + timedelta(hours=1)
def test_interval_ignores_tz_field():
"""Interval scheduling is wall-clock-independent — adding 6 hours
is the same in every timezone. The ``tz`` field is ignored even
if a caller mistakenly sets it."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('schedule',
{'interval': 6, 'unit': 'hours', 'tz': 'America/Los_Angeles'},
now)
assert result == now + timedelta(hours=6)
# ---------------------------------------------------------------------------
# Daily (``trigger_type='daily_time'``)
# ---------------------------------------------------------------------------
def test_daily_today_at_future_time_runs_today():
"""It's 12:00 UTC and the schedule says 18:00 UTC — next run is
today at 18:00, not tomorrow."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('daily_time', {'time': '18:00', 'tz': 'UTC'}, now)
assert result == _utc(2026, 5, 27, 18, 0)
def test_daily_today_at_past_time_runs_tomorrow():
"""It's 18:00 UTC and the schedule says 09:00 UTC — next run is
tomorrow at 09:00."""
now = _utc(2026, 5, 27, 18, 0)
result = next_run_at('daily_time', {'time': '09:00', 'tz': 'UTC'}, now)
assert result == _utc(2026, 5, 28, 9, 0)
def test_daily_at_exact_target_time_runs_tomorrow():
"""Edge case: schedule fires at exactly 09:00, and ``now`` is
exactly 09:00. ``<=`` check pushes to tomorrow — otherwise we'd
immediately reschedule for the present moment and the engine
would run again in 0s."""
now = _utc(2026, 5, 27, 9, 0)
result = next_run_at('daily_time', {'time': '09:00', 'tz': 'UTC'}, now)
assert result == _utc(2026, 5, 28, 9, 0)
def test_daily_respects_user_timezone_not_server_local():
"""User on Pacific time, schedule says ``09:00 America/Los_Angeles``.
Server is UTC. At 12:00 UTC = 05:00 LA local, next run is 09:00 LA
today = 16:00 UTC. Pre-fix the engine used naive ``datetime.now()``
and read 12:00 as if it were the user's tz, mis-scheduling by the
server-vs-user tz offset."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('daily_time',
{'time': '09:00', 'tz': 'America/Los_Angeles'},
now)
# 09:00 LA on 2026-05-27 → 16:00 UTC (PDT, UTC-7).
assert result == _utc(2026, 5, 27, 16, 0)
def test_daily_falls_back_to_default_tz_when_config_missing():
"""``tz`` field absent on the config — pulls from ``default_tz``
(typically the app-level setting)."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('daily_time', {'time': '09:00'}, now,
default_tz='America/Los_Angeles')
assert result == _utc(2026, 5, 27, 16, 0)
def test_daily_garbage_time_string_defaults_to_midnight():
"""Bad ``time`` string → defaults to 00:00 (engine's existing
behaviour). Better than crashing the scheduler when a row's
config was hand-edited."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('daily_time', {'time': 'garbage', 'tz': 'UTC'}, now)
# 00:00 today already passed → tomorrow at 00:00.
assert result == _utc(2026, 5, 28, 0, 0)
def test_daily_unknown_tz_falls_back_to_utc():
"""Unknown IANA tz string → fall back to UTC rather than crash."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at('daily_time',
{'time': '15:00', 'tz': 'Imaginary/Place'},
now)
# Treated as UTC → next run today at 15:00 UTC.
assert result == _utc(2026, 5, 27, 15, 0)
def test_unknown_tz_logs_warning_once(caplog):
"""Silent fallback to UTC was a bug — user configures
'America/Los_Angeles' but tzdata is missing → schedule runs at the
wrong hour with no log line. Log once per unknown name so the
misconfiguration is debuggable from a single grep, and don't spam
the log on every poll cycle for the same row."""
import logging
from core.automation import schedule
schedule._UNKNOWN_TZ_WARNED.clear() # fresh state for the test
now = _utc(2026, 5, 27, 12, 0)
with caplog.at_level(logging.WARNING, logger='soulsync.automation.schedule'):
# Two calls with the same bad name — only ONE warning emitted.
next_run_at('daily_time', {'time': '09:00', 'tz': 'Bogus/Tz'}, now)
next_run_at('daily_time', {'time': '09:00', 'tz': 'Bogus/Tz'}, now)
matching = [r for r in caplog.records if 'Bogus/Tz' in r.getMessage()]
assert len(matching) == 1
assert 'tzdata' in matching[0].getMessage().lower()
def test_unknown_tz_warning_includes_helpful_hint():
"""Log line must point the user at the two real causes: typo in
the IANA name, or missing tzdata on the host. Without that hint
the symptom (schedule running at UTC offset) is bewildering."""
import logging
from core.automation import schedule
schedule._UNKNOWN_TZ_WARNED.clear()
caplog_records = []
class _Capture(logging.Handler):
def emit(self, record):
caplog_records.append(record.getMessage())
handler = _Capture()
logger_obj = logging.getLogger('soulsync.automation.schedule')
logger_obj.addHandler(handler)
try:
next_run_at('daily_time', {'time': '09:00', 'tz': 'Made/Up'},
_utc(2026, 5, 27, 12, 0))
finally:
logger_obj.removeHandler(handler)
assert any("'Made/Up'" in m for m in caplog_records)
assert any('IANA' in m for m in caplog_records)
# ---------------------------------------------------------------------------
# DST edge cases — pin that ``zoneinfo``'s default resolution handles
# spring-forward gap + fall-back ambiguity sensibly. Both transitions
# happen in the user's local tz, NOT in UTC, so the result UTC offset
# changes across the boundary.
# ---------------------------------------------------------------------------
def test_dst_spring_forward_lands_after_the_gap():
"""In Los Angeles, 2026-03-08 02:30 doesn't exist — clocks jump
from 02:00 PST directly to 03:00 PDT. A schedule for 02:30 daily
that fires through this transition must NOT raise and must land
on a real instant. ``zoneinfo``'s default resolution maps the
gap to the post-jump side (treating 02:30 as 03:30 PDT), so the
UTC equivalent shifts by an hour relative to non-DST days."""
# 2026-03-08 00:00 UTC = 2026-03-07 16:00 PST (still PST).
# Schedule fires 02:30 LA daily. Next run on 03-07 was 02:30 PST
# = 10:30 UTC. We're querying after that → next run is 03-08
# 02:30 LA, which falls in the spring-forward gap. zoneinfo
# resolves to 03:30 PDT = 10:30 UTC (offset already shifted to
# PDT for the rest of the day post-transition).
now = _utc(2026, 3, 8, 0, 0)
result = next_run_at('daily_time',
{'time': '02:30', 'tz': 'America/Los_Angeles'},
now)
# Must be aware UTC, must NOT crash on the gap.
assert result is not None
assert result.tzinfo is not None
# Result is somewhere on 03-08 — exact time depends on zoneinfo's
# gap-resolution policy, but it must be on the right day and
# past ``now``.
assert result > now
assert result.date() == datetime(2026, 3, 8).date()
def test_dst_fall_back_handles_ambiguous_local_time():
"""2026-11-01 01:30 in Los Angeles happens TWICE (once at PDT
UTC-7, once at PST UTC-8 after the fall-back). A daily schedule
for 01:30 must resolve to ONE instant — ``zoneinfo``'s default
picks the first occurrence (PDT), so the UTC time is 08:30."""
# 2026-11-01 00:00 UTC = 2026-10-31 17:00 PDT.
# Next 01:30 LA is 2026-11-01 — ambiguous, zoneinfo defaults to
# the earlier (PDT) instant: 08:30 UTC.
now = _utc(2026, 11, 1, 0, 0)
result = next_run_at('daily_time',
{'time': '01:30', 'tz': 'America/Los_Angeles'},
now)
assert result is not None
# 01:30 PDT (UTC-7) → 08:30 UTC.
assert result == _utc(2026, 11, 1, 8, 30)
def test_weekly_across_dst_boundary_keeps_local_wall_clock():
"""User schedules "every Sunday at 09:00 LA". Crossing the
spring-forward DST boundary, the LOCAL wall clock stays at 09:00
even though the UTC equivalent shifts by an hour. Pre-DST Sunday
09:00 PST = 17:00 UTC; post-DST Sunday 09:00 PDT = 16:00 UTC."""
# Pre-DST Sunday: 2026-03-01.
pre_dst = _utc(2026, 3, 1, 10, 0) # Sunday 02:00 PST already past 09:00? No — 02:00 < 09:00, so today still qualifies.
result_pre = next_run_at('weekly_time',
{'time': '09:00', 'days': ['sun'],
'tz': 'America/Los_Angeles'},
pre_dst)
# 09:00 PST = 17:00 UTC.
assert result_pre == _utc(2026, 3, 1, 17, 0)
# Post-DST Sunday: 2026-03-15 (the 8th was DST switch day).
post_dst = _utc(2026, 3, 15, 10, 0) # 03:00 PDT — before 09:00.
result_post = next_run_at('weekly_time',
{'time': '09:00', 'days': ['sun'],
'tz': 'America/Los_Angeles'},
post_dst)
# 09:00 PDT = 16:00 UTC.
assert result_post == _utc(2026, 3, 15, 16, 0)
# Same local wall clock, different UTC — the kind of bug that
# caused the May 2026 "next in 8h" tz mismatch.
assert result_pre.hour == 17
assert result_post.hour == 16
# ---------------------------------------------------------------------------
# Weekly (``trigger_type='weekly_time'``)
# ---------------------------------------------------------------------------
def test_weekly_picks_next_matching_weekday():
"""It's Wednesday and the schedule wants Mon/Wed/Fri — same day
qualifies if the time is still in the future."""
# 2026-05-27 is a Wednesday.
now = _utc(2026, 5, 27, 8, 0)
result = next_run_at('weekly_time',
{'time': '14:00', 'days': ['mon', 'wed', 'fri'], 'tz': 'UTC'},
now)
assert result == _utc(2026, 5, 27, 14, 0)
def test_weekly_rolls_to_next_allowed_day_when_today_passed():
"""Wednesday 18:00 UTC, schedule wants Mon/Wed/Fri at 14:00 —
Wed 14:00 already passed today, next match is Friday at 14:00."""
now = _utc(2026, 5, 27, 18, 0) # Wed
result = next_run_at('weekly_time',
{'time': '14:00', 'days': ['mon', 'wed', 'fri'], 'tz': 'UTC'},
now)
assert result == _utc(2026, 5, 29, 14, 0) # Fri
def test_weekly_wraps_to_next_week():
"""Sunday past the time, schedule wants only Monday — next match
is the very next day."""
# 2026-05-31 is a Sunday.
now = _utc(2026, 5, 31, 15, 0)
result = next_run_at('weekly_time',
{'time': '09:00', 'days': ['mon'], 'tz': 'UTC'},
now)
assert result == _utc(2026, 6, 1, 9, 0) # next Monday
def test_weekly_empty_days_means_every_day():
"""Empty ``days`` list → treat as every weekday. Matches the
engine's existing fallback in ``_next_weekly_occurrence``."""
now = _utc(2026, 5, 27, 8, 0)
result = next_run_at('weekly_time',
{'time': '14:00', 'days': [], 'tz': 'UTC'},
now)
# Today (Wed) qualifies since 14:00 is still future.
assert result == _utc(2026, 5, 27, 14, 0)
def test_weekly_unrecognised_day_abbreviations_dropped():
"""``'mond'`` / ``'frid'`` are not in the map — silently drop.
If ALL listed days are invalid, fall through to the every-day
default (matches the empty-list behaviour)."""
now = _utc(2026, 5, 27, 8, 0)
result = next_run_at('weekly_time',
{'time': '14:00', 'days': ['mond', 'frid'], 'tz': 'UTC'},
now)
# All garbage → every day → today (Wed) qualifies.
assert result == _utc(2026, 5, 27, 14, 0)
def test_weekly_day_abbreviations_case_insensitive():
"""``MON`` / ``Mon`` / ``mon`` all parse to weekday 0."""
now = _utc(2026, 5, 27, 8, 0) # Wed
result = next_run_at('weekly_time',
{'time': '14:00', 'days': ['MON', 'WED'], 'tz': 'UTC'},
now)
assert result == _utc(2026, 5, 27, 14, 0)
def test_weekly_respects_user_tz_across_day_boundary():
"""It's 23:30 UTC on Wednesday → 16:30 LA local (still Wed).
Schedule fires Mon/Wed/Fri at 18:00 LA. Next run is 18:00 LA
today (Wed in LA, but Thursday in UTC because of the 7h offset)."""
now = _utc(2026, 5, 27, 23, 30) # Wed 23:30 UTC / Wed 16:30 LA
result = next_run_at('weekly_time',
{'time': '18:00', 'days': ['mon', 'wed', 'fri'],
'tz': 'America/Los_Angeles'},
now)
# 2026-05-27 18:00 LA → 2026-05-28 01:00 UTC.
assert result == _utc(2026, 5, 28, 1, 0)
# ---------------------------------------------------------------------------
# Monthly (``trigger_type='monthly_time'`` — NEW in PR 1)
# ---------------------------------------------------------------------------
def test_monthly_picks_target_day_this_month_when_future():
"""It's the 5th, schedule fires on the 15th — next run is the
15th of the current month."""
now = _utc(2026, 5, 5, 12, 0)
result = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 15, 'tz': 'UTC'},
now)
assert result == _utc(2026, 5, 15, 9, 0)
def test_monthly_rolls_to_next_month_when_target_day_passed():
"""It's the 20th, schedule fires on the 15th — already past in
May, next run is June 15."""
now = _utc(2026, 5, 20, 12, 0)
result = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 15, 'tz': 'UTC'},
now)
assert result == _utc(2026, 6, 15, 9, 0)
def test_monthly_clamps_to_last_day_when_month_too_short():
"""Schedule wants day 31; February has 28 (or 29). Clamp to the
LAST valid day of that month — running a day or two early in
short months is less surprising than silently skipping a month
entirely. Standard cron convention."""
now = _utc(2026, 2, 1, 12, 0) # 2026 is not a leap year
result = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 31, 'tz': 'UTC'},
now)
# 2026 Feb has 28 days → run on the 28th instead.
assert result == _utc(2026, 2, 28, 9, 0)
def test_monthly_handles_leap_year_february():
"""2024 was a leap year — February has 29 days, so day-31 clamps
to the 29th, not the 28th."""
now = _utc(2024, 2, 1, 12, 0)
result = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 31, 'tz': 'UTC'},
now)
assert result == _utc(2024, 2, 29, 9, 0)
def test_monthly_clamps_day_above_31_and_below_1():
"""Defensive — config values outside [1, 31] clamp to the nearest
valid bound rather than crashing the scheduler."""
now = _utc(2026, 5, 5, 12, 0)
high = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 99, 'tz': 'UTC'},
now)
low = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': -5, 'tz': 'UTC'},
now)
# 99 → clamped to 31 → May has 31 days → May 31st.
assert high == _utc(2026, 5, 31, 9, 0)
# -5 → clamped to 1 → next 1st is June 1 (May 1 already passed).
assert low == _utc(2026, 6, 1, 9, 0)
def test_monthly_rolls_year_at_december_to_january():
"""December 20, schedule fires on the 5th — next run is January 5
of the FOLLOWING year, not month 13 of the current year."""
now = _utc(2026, 12, 20, 12, 0)
result = next_run_at('monthly_time',
{'time': '09:00', 'day_of_month': 5, 'tz': 'UTC'},
now)
assert result == _utc(2027, 1, 5, 9, 0)
def test_monthly_respects_user_tz():
"""Schedule wants the 1st of each month at 02:00 LA. ``now`` is
May 1 at 06:00 UTC = April 30 at 23:00 LA. So locally we haven't
hit May 1 02:00 LA yet → next run is May 1 02:00 LA = May 1 09:00
UTC (PDT, UTC-7)."""
now = _utc(2026, 5, 1, 6, 0)
result = next_run_at('monthly_time',
{'time': '02:00', 'day_of_month': 1,
'tz': 'America/Los_Angeles'},
now)
assert result == _utc(2026, 5, 1, 9, 0)
# ---------------------------------------------------------------------------
# Result shape — every returned datetime must be aware UTC so the engine
# can serialise it to the DB ``next_run`` column without ambiguity.
# ---------------------------------------------------------------------------
@pytest.mark.parametrize('trigger_type,config', [
('schedule', {'interval': 1, 'unit': 'hours'}),
('daily_time', {'time': '09:00', 'tz': 'America/Los_Angeles'}),
('weekly_time', {'time': '09:00', 'days': ['mon'], 'tz': 'America/Los_Angeles'}),
('monthly_time', {'time': '09:00', 'day_of_month': 15, 'tz': 'America/Los_Angeles'}),
])
def test_result_is_always_aware_utc(trigger_type, config):
"""Engine writes the result as a naive string to the DB but the
convention is "stored as UTC". Returning a naive datetime would
leak the caller's local tz into the column. Pin the contract:
every result has ``tzinfo`` and is at UTC offset zero."""
now = _utc(2026, 5, 27, 12, 0)
result = next_run_at(trigger_type, config, now)
assert result is not None
assert result.tzinfo is not None
assert result.utcoffset() == timedelta(0)
def test_naive_now_utc_is_coerced_to_aware_utc():
"""Defensive — naive ``now_utc`` inputs are assumed UTC and the
result is still aware UTC. Matches the engine's convention
when parsing the DB ``next_run`` column."""
naive_now = datetime(2026, 5, 27, 12, 0)
result = next_run_at('schedule', {'interval': 1, 'unit': 'hours'}, naive_now)
assert result == _utc(2026, 5, 27, 13, 0)
assert result.tzinfo is not None