core/tests/components/recorder/test_purge.py

1111 lines
39 KiB
Python

"""Test data purging."""
from datetime import datetime, timedelta
import json
import sqlite3
from unittest.mock import MagicMock, patch
from sqlalchemy.exc import DatabaseError, OperationalError
from sqlalchemy.orm.session import Session
from homeassistant.components import recorder
from homeassistant.components.recorder import PurgeTask
from homeassistant.components.recorder.const import MAX_ROWS_TO_PURGE
from homeassistant.components.recorder.models import (
Events,
RecorderRuns,
States,
StatisticsRuns,
StatisticsShortTerm,
)
from homeassistant.components.recorder.purge import purge_old_data
from homeassistant.components.recorder.util import session_scope
from homeassistant.const import EVENT_STATE_CHANGED
from homeassistant.core import HomeAssistant
from homeassistant.helpers.typing import ConfigType
from homeassistant.util import dt as dt_util
from .common import (
async_recorder_block_till_done,
async_wait_purge_done,
async_wait_recording_done,
async_wait_recording_done_without_instance,
)
from .conftest import SetupRecorderInstanceT
async def test_purge_old_states(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test deleting old states."""
instance = await async_setup_recorder_instance(hass)
await _add_test_states(hass, instance)
# make sure we start with 6 states
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 6
assert states[0].old_state_id is None
assert states[-1].old_state_id == states[-2].state_id
events = session.query(Events).filter(Events.event_type == "state_changed")
assert events.count() == 6
assert "test.recorder2" in instance._old_states
purge_before = dt_util.utcnow() - timedelta(days=4)
# run purge_old_data()
finished = purge_old_data(instance, purge_before, repack=False)
assert not finished
assert states.count() == 2
assert "test.recorder2" in instance._old_states
states_after_purge = session.query(States)
assert states_after_purge[1].old_state_id == states_after_purge[0].state_id
assert states_after_purge[0].old_state_id is None
finished = purge_old_data(instance, purge_before, repack=False)
assert finished
assert states.count() == 2
assert "test.recorder2" in instance._old_states
# run purge_old_data again
purge_before = dt_util.utcnow()
finished = purge_old_data(instance, purge_before, repack=False)
assert not finished
assert states.count() == 0
assert "test.recorder2" not in instance._old_states
# Add some more states
await _add_test_states(hass, instance)
# make sure we start with 6 states
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 6
assert states[0].old_state_id is None
assert states[-1].old_state_id == states[-2].state_id
events = session.query(Events).filter(Events.event_type == "state_changed")
assert events.count() == 6
assert "test.recorder2" in instance._old_states
async def test_purge_old_states_encouters_database_corruption(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test database image image is malformed while deleting old states."""
instance = await async_setup_recorder_instance(hass)
await _add_test_states(hass, instance)
await async_wait_recording_done_without_instance(hass)
sqlite3_exception = DatabaseError("statement", {}, [])
sqlite3_exception.__cause__ = sqlite3.DatabaseError()
with patch(
"homeassistant.components.recorder.move_away_broken_database"
) as move_away, patch(
"homeassistant.components.recorder.purge.purge_old_data",
side_effect=sqlite3_exception,
):
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, {"keep_days": 0}
)
await hass.async_block_till_done()
await async_wait_recording_done_without_instance(hass)
assert move_away.called
# Ensure the whole database was reset due to the database error
with session_scope(hass=hass) as session:
states_after_purge = session.query(States)
assert states_after_purge.count() == 0
async def test_purge_old_states_encounters_temporary_mysql_error(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
caplog,
):
"""Test retry on specific mysql operational errors."""
instance = await async_setup_recorder_instance(hass)
await _add_test_states(hass, instance)
await async_wait_recording_done_without_instance(hass)
mysql_exception = OperationalError("statement", {}, [])
mysql_exception.orig = MagicMock(args=(1205, "retryable"))
with patch(
"homeassistant.components.recorder.util.time.sleep"
) as sleep_mock, patch(
"homeassistant.components.recorder.purge._purge_old_recorder_runs",
side_effect=[mysql_exception, None],
), patch.object(
instance.engine.dialect, "name", "mysql"
):
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, {"keep_days": 0}
)
await hass.async_block_till_done()
await async_wait_recording_done_without_instance(hass)
await async_wait_recording_done_without_instance(hass)
assert "retrying" in caplog.text
assert sleep_mock.called
async def test_purge_old_states_encounters_operational_error(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
caplog,
):
"""Test error on operational errors that are not mysql does not retry."""
instance = await async_setup_recorder_instance(hass)
await _add_test_states(hass, instance)
await async_wait_recording_done_without_instance(hass)
exception = OperationalError("statement", {}, [])
with patch(
"homeassistant.components.recorder.purge._purge_old_recorder_runs",
side_effect=exception,
):
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, {"keep_days": 0}
)
await hass.async_block_till_done()
await async_wait_recording_done_without_instance(hass)
await async_wait_recording_done_without_instance(hass)
assert "retrying" not in caplog.text
assert "Error executing purge" in caplog.text
async def test_purge_old_events(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test deleting old events."""
instance = await async_setup_recorder_instance(hass)
await _add_test_events(hass, instance)
with session_scope(hass=hass) as session:
events = session.query(Events).filter(Events.event_type.like("EVENT_TEST%"))
assert events.count() == 6
purge_before = dt_util.utcnow() - timedelta(days=4)
# run purge_old_data()
finished = purge_old_data(instance, purge_before, repack=False)
assert not finished
assert events.count() == 2
# we should only have 2 events left
finished = purge_old_data(instance, purge_before, repack=False)
assert finished
assert events.count() == 2
async def test_purge_old_recorder_runs(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test deleting old recorder runs keeps current run."""
instance = await async_setup_recorder_instance(hass)
await _add_test_recorder_runs(hass, instance)
# make sure we start with 7 recorder runs
with session_scope(hass=hass) as session:
recorder_runs = session.query(RecorderRuns)
assert recorder_runs.count() == 7
purge_before = dt_util.utcnow()
# run purge_old_data()
finished = purge_old_data(instance, purge_before, repack=False)
assert not finished
finished = purge_old_data(instance, purge_before, repack=False)
assert finished
assert recorder_runs.count() == 1
async def test_purge_old_statistics_runs(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test deleting old statistics runs keeps the latest run."""
instance = await async_setup_recorder_instance(hass)
await _add_test_statistics_runs(hass, instance)
# make sure we start with 7 statistics runs
with session_scope(hass=hass) as session:
statistics_runs = session.query(StatisticsRuns)
assert statistics_runs.count() == 7
purge_before = dt_util.utcnow()
# run purge_old_data()
finished = purge_old_data(instance, purge_before, repack=False)
assert not finished
finished = purge_old_data(instance, purge_before, repack=False)
assert finished
assert statistics_runs.count() == 1
async def test_purge_method(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
caplog,
):
"""Test purge method."""
instance = await async_setup_recorder_instance(hass)
service_data = {"keep_days": 4}
await _add_test_events(hass, instance)
await _add_test_states(hass, instance)
await _add_test_statistics(hass, instance)
await _add_test_recorder_runs(hass, instance)
await _add_test_statistics_runs(hass, instance)
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
# make sure we start with 6 states
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 6
events = session.query(Events).filter(Events.event_type.like("EVENT_TEST%"))
assert events.count() == 6
statistics = session.query(StatisticsShortTerm)
assert statistics.count() == 6
recorder_runs = session.query(RecorderRuns)
assert recorder_runs.count() == 7
runs_before_purge = recorder_runs.all()
statistics_runs = session.query(StatisticsRuns)
assert statistics_runs.count() == 7
statistic_runs_before_purge = statistics_runs.all()
await hass.async_block_till_done()
await async_wait_purge_done(hass, instance)
# run purge method - no service data, use defaults
await hass.services.async_call("recorder", "purge")
await hass.async_block_till_done()
# Small wait for recorder thread
await async_wait_purge_done(hass, instance)
# only purged old states, events and statistics
assert states.count() == 4
assert events.count() == 4
assert statistics.count() == 4
# run purge method - correct service data
await hass.services.async_call("recorder", "purge", service_data=service_data)
await hass.async_block_till_done()
# Small wait for recorder thread
await async_wait_purge_done(hass, instance)
# we should only have 2 states, events and statistics left after purging
assert states.count() == 2
assert events.count() == 2
assert statistics.count() == 2
# now we should only have 3 recorder runs left
runs = recorder_runs.all()
assert runs[0] == runs_before_purge[0]
assert runs[1] == runs_before_purge[5]
assert runs[2] == runs_before_purge[6]
# now we should only have 3 statistics runs left
runs = statistics_runs.all()
assert runs[0] == statistic_runs_before_purge[0]
assert runs[1] == statistic_runs_before_purge[5]
assert runs[2] == statistic_runs_before_purge[6]
assert "EVENT_TEST_PURGE" not in (event.event_type for event in events.all())
# run purge method - correct service data, with repack
service_data["repack"] = True
await hass.services.async_call("recorder", "purge", service_data=service_data)
await hass.async_block_till_done()
await async_wait_purge_done(hass, instance)
assert "Vacuuming SQL DB to free space" in caplog.text
async def test_purge_edge_case(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
):
"""Test states and events are purged even if they occurred shortly before purge_before."""
async def _add_db_entries(hass: HomeAssistant, timestamp: datetime) -> None:
with recorder.session_scope(hass=hass) as session:
session.add(
Events(
event_id=1001,
event_type="EVENT_TEST_PURGE",
event_data="{}",
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)
session.add(
States(
entity_id="test.recorder2",
domain="sensor",
state="purgeme",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
event_id=1001,
)
)
instance = await async_setup_recorder_instance(hass, None)
await async_wait_purge_done(hass, instance)
service_data = {"keep_days": 2}
timestamp = dt_util.utcnow() - timedelta(days=2, minutes=1)
await _add_db_entries(hass, timestamp)
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 1
events = session.query(Events).filter(Events.event_type == "EVENT_TEST_PURGE")
assert events.count() == 1
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert states.count() == 0
assert events.count() == 0
async def test_purge_cutoff_date(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
):
"""Test states and events are purged only if they occurred before "now() - keep_days"."""
async def _add_db_entries(hass: HomeAssistant, cutoff: datetime, rows: int) -> None:
timestamp_keep = cutoff
timestamp_purge = cutoff - timedelta(microseconds=1)
with recorder.session_scope(hass=hass) as session:
session.add(
Events(
event_id=1000,
event_type="KEEP",
event_data="{}",
origin="LOCAL",
created=timestamp_keep,
time_fired=timestamp_keep,
)
)
session.add(
States(
entity_id="test.cutoff",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp_keep,
last_updated=timestamp_keep,
created=timestamp_keep,
event_id=1000,
)
)
for row in range(1, rows):
session.add(
Events(
event_id=1000 + row,
event_type="PURGE",
event_data="{}",
origin="LOCAL",
created=timestamp_purge,
time_fired=timestamp_purge,
)
)
session.add(
States(
entity_id="test.cutoff",
domain="sensor",
state="purge",
attributes="{}",
last_changed=timestamp_purge,
last_updated=timestamp_purge,
created=timestamp_purge,
event_id=1000 + row,
)
)
instance = await async_setup_recorder_instance(hass, None)
await async_wait_purge_done(hass, instance)
service_data = {"keep_days": 2}
# Force multiple purge batches to be run
rows = MAX_ROWS_TO_PURGE + 1
cutoff = dt_util.utcnow() - timedelta(days=service_data["keep_days"])
await _add_db_entries(hass, cutoff, rows)
with session_scope(hass=hass) as session:
states = session.query(States)
events = session.query(Events)
assert states.filter(States.state == "purge").count() == rows - 1
assert states.filter(States.state == "keep").count() == 1
assert events.filter(Events.event_type == "PURGE").count() == rows - 1
assert events.filter(Events.event_type == "KEEP").count() == 1
instance.queue.put(PurgeTask(cutoff, repack=False, apply_filter=False))
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
states = session.query(States)
events = session.query(Events)
assert states.filter(States.state == "purge").count() == 0
assert states.filter(States.state == "keep").count() == 1
assert events.filter(Events.event_type == "PURGE").count() == 0
assert events.filter(Events.event_type == "KEEP").count() == 1
async def test_purge_filtered_states(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
):
"""Test filtered states are purged."""
config: ConfigType = {"exclude": {"entities": ["sensor.excluded"]}}
instance = await async_setup_recorder_instance(hass, config)
assert instance.entity_filter("sensor.excluded") is False
def _add_db_entries(hass: HomeAssistant) -> None:
with recorder.session_scope(hass=hass) as session:
# Add states and state_changed events that should be purged
for days in range(1, 4):
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(1000, 1020):
_add_state_and_state_changed_event(
session,
"sensor.excluded",
"purgeme",
timestamp,
event_id * days,
)
# Add state **without** state_changed event that should be purged
timestamp = dt_util.utcnow() - timedelta(days=1)
session.add(
States(
entity_id="sensor.excluded",
domain="sensor",
state="purgeme",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
)
)
# Add states and state_changed events that should be keeped
timestamp = dt_util.utcnow() - timedelta(days=2)
for event_id in range(200, 210):
_add_state_and_state_changed_event(
session,
"sensor.keep",
"keep",
timestamp,
event_id,
)
# Add states with linked old_state_ids that need to be handled
timestamp = dt_util.utcnow() - timedelta(days=0)
state_1 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=1,
)
timestamp = dt_util.utcnow() - timedelta(days=4)
state_2 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=2,
)
state_3 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=62, # keep
)
session.add_all((state_1, state_2, state_3))
# Add event that should be keeped
session.add(
Events(
event_id=100,
event_type="EVENT_KEEP",
event_data="{}",
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)
service_data = {"keep_days": 10}
_add_db_entries(hass)
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 74
events_state_changed = session.query(Events).filter(
Events.event_type == EVENT_STATE_CHANGED
)
events_keep = session.query(Events).filter(Events.event_type == "EVENT_KEEP")
assert events_state_changed.count() == 70
assert events_keep.count() == 1
# Normal purge doesn't remove excluded entities
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert states.count() == 74
assert events_state_changed.count() == 70
assert events_keep.count() == 1
# Test with 'apply_filter' = True
service_data["apply_filter"] = True
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert states.count() == 13
assert events_state_changed.count() == 10
assert events_keep.count() == 1
states_sensor_excluded = session.query(States).filter(
States.entity_id == "sensor.excluded"
)
assert states_sensor_excluded.count() == 0
assert session.query(States).get(72).old_state_id is None
assert session.query(States).get(73).old_state_id is None
assert session.query(States).get(74).old_state_id == 62 # should have been kept
async def test_purge_filtered_events(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
):
"""Test filtered events are purged."""
config: ConfigType = {"exclude": {"event_types": ["EVENT_PURGE"]}}
instance = await async_setup_recorder_instance(hass, config)
def _add_db_entries(hass: HomeAssistant) -> None:
with recorder.session_scope(hass=hass) as session:
# Add events that should be purged
for days in range(1, 4):
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(1000, 1020):
session.add(
Events(
event_id=event_id * days,
event_type="EVENT_PURGE",
event_data="{}",
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)
# Add states and state_changed events that should be keeped
timestamp = dt_util.utcnow() - timedelta(days=1)
for event_id in range(200, 210):
_add_state_and_state_changed_event(
session,
"sensor.keep",
"keep",
timestamp,
event_id,
)
service_data = {"keep_days": 10}
_add_db_entries(hass)
with session_scope(hass=hass) as session:
events_purge = session.query(Events).filter(Events.event_type == "EVENT_PURGE")
events_keep = session.query(Events).filter(
Events.event_type == EVENT_STATE_CHANGED
)
states = session.query(States)
assert events_purge.count() == 60
assert events_keep.count() == 10
assert states.count() == 10
# Normal purge doesn't remove excluded events
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert events_purge.count() == 60
assert events_keep.count() == 10
assert states.count() == 10
# Test with 'apply_filter' = True
service_data["apply_filter"] = True
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert events_purge.count() == 0
assert events_keep.count() == 10
assert states.count() == 10
async def test_purge_filtered_events_state_changed(
hass: HomeAssistant,
async_setup_recorder_instance: SetupRecorderInstanceT,
):
"""Test filtered state_changed events are purged. This should also remove all states."""
config: ConfigType = {"exclude": {"event_types": [EVENT_STATE_CHANGED]}}
instance = await async_setup_recorder_instance(hass, config)
# Assert entity_id is NOT excluded
assert instance.entity_filter("sensor.excluded") is True
def _add_db_entries(hass: HomeAssistant) -> None:
with recorder.session_scope(hass=hass) as session:
# Add states and state_changed events that should be purged
for days in range(1, 4):
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(1000, 1020):
_add_state_and_state_changed_event(
session,
"sensor.excluded",
"purgeme",
timestamp,
event_id * days,
)
# Add events that should be keeped
timestamp = dt_util.utcnow() - timedelta(days=1)
for event_id in range(200, 210):
session.add(
Events(
event_id=event_id,
event_type="EVENT_KEEP",
event_data="{}",
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)
# Add states with linked old_state_ids that need to be handled
timestamp = dt_util.utcnow() - timedelta(days=0)
state_1 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=1,
)
timestamp = dt_util.utcnow() - timedelta(days=4)
state_2 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=2,
)
state_3 = States(
entity_id="sensor.linked_old_state_id",
domain="sensor",
state="keep",
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
old_state_id=62, # keep
)
session.add_all((state_1, state_2, state_3))
service_data = {"keep_days": 10, "apply_filter": True}
_add_db_entries(hass)
with session_scope(hass=hass) as session:
events_keep = session.query(Events).filter(Events.event_type == "EVENT_KEEP")
events_purge = session.query(Events).filter(
Events.event_type == EVENT_STATE_CHANGED
)
states = session.query(States)
assert events_keep.count() == 10
assert events_purge.count() == 60
assert states.count() == 63
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
assert events_keep.count() == 10
assert events_purge.count() == 0
assert states.count() == 3
assert session.query(States).get(61).old_state_id is None
assert session.query(States).get(62).old_state_id is None
assert session.query(States).get(63).old_state_id == 62 # should have been kept
async def test_purge_entities(
hass: HomeAssistant, async_setup_recorder_instance: SetupRecorderInstanceT
):
"""Test purging of specific entities."""
instance = await async_setup_recorder_instance(hass)
async def _purge_entities(hass, entity_ids, domains, entity_globs):
service_data = {
"entity_id": entity_ids,
"domains": domains,
"entity_globs": entity_globs,
}
await hass.services.async_call(
recorder.DOMAIN, recorder.SERVICE_PURGE_ENTITIES, service_data
)
await hass.async_block_till_done()
await async_recorder_block_till_done(hass, instance)
await async_wait_purge_done(hass, instance)
def _add_purge_records(hass: HomeAssistant) -> None:
with recorder.session_scope(hass=hass) as session:
# Add states and state_changed events that should be purged
for days in range(1, 4):
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(1000, 1020):
_add_state_and_state_changed_event(
session,
"sensor.purge_entity",
"purgeme",
timestamp,
event_id * days,
)
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(10000, 10020):
_add_state_and_state_changed_event(
session,
"purge_domain.entity",
"purgeme",
timestamp,
event_id * days,
)
timestamp = dt_util.utcnow() - timedelta(days=days)
for event_id in range(100000, 100020):
_add_state_and_state_changed_event(
session,
"binary_sensor.purge_glob",
"purgeme",
timestamp,
event_id * days,
)
def _add_keep_records(hass: HomeAssistant) -> None:
with recorder.session_scope(hass=hass) as session:
# Add states and state_changed events that should be kept
timestamp = dt_util.utcnow() - timedelta(days=2)
for event_id in range(200, 210):
_add_state_and_state_changed_event(
session,
"sensor.keep",
"keep",
timestamp,
event_id,
)
_add_purge_records(hass)
_add_keep_records(hass)
# Confirm standard service call
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 190
await _purge_entities(
hass, "sensor.purge_entity", "purge_domain", "*purge_glob"
)
assert states.count() == 10
states_sensor_kept = session.query(States).filter(
States.entity_id == "sensor.keep"
)
assert states_sensor_kept.count() == 10
_add_purge_records(hass)
# Confirm each parameter purges only the associated records
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 190
await _purge_entities(hass, "sensor.purge_entity", [], [])
assert states.count() == 130
await _purge_entities(hass, [], "purge_domain", [])
assert states.count() == 70
await _purge_entities(hass, [], [], "*purge_glob")
assert states.count() == 10
states_sensor_kept = session.query(States).filter(
States.entity_id == "sensor.keep"
)
assert states_sensor_kept.count() == 10
_add_purge_records(hass)
# Confirm calling service without arguments matches all records (default filter behaviour)
with session_scope(hass=hass) as session:
states = session.query(States)
assert states.count() == 190
await _purge_entities(hass, [], [], [])
assert states.count() == 0
async def _add_test_states(hass: HomeAssistant, instance: recorder.Recorder):
"""Add multiple states to the db for testing."""
utcnow = dt_util.utcnow()
five_days_ago = utcnow - timedelta(days=5)
eleven_days_ago = utcnow - timedelta(days=11)
attributes = {"test_attr": 5, "test_attr_10": "nice"}
async def set_state(entity_id, state, **kwargs):
"""Set the state."""
hass.states.async_set(entity_id, state, **kwargs)
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
for event_id in range(6):
if event_id < 2:
timestamp = eleven_days_ago
state = f"autopurgeme_{event_id}"
elif event_id < 4:
timestamp = five_days_ago
state = f"purgeme_{event_id}"
else:
timestamp = utcnow
state = f"dontpurgeme_{event_id}"
with patch(
"homeassistant.components.recorder.dt_util.utcnow", return_value=timestamp
):
await set_state("test.recorder2", state, attributes=attributes)
async def _add_test_events(hass: HomeAssistant, instance: recorder.Recorder):
"""Add a few events for testing."""
utcnow = dt_util.utcnow()
five_days_ago = utcnow - timedelta(days=5)
eleven_days_ago = utcnow - timedelta(days=11)
event_data = {"test_attr": 5, "test_attr_10": "nice"}
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
with recorder.session_scope(hass=hass) as session:
for event_id in range(6):
if event_id < 2:
timestamp = eleven_days_ago
event_type = "EVENT_TEST_AUTOPURGE"
elif event_id < 4:
timestamp = five_days_ago
event_type = "EVENT_TEST_PURGE"
else:
timestamp = utcnow
event_type = "EVENT_TEST"
session.add(
Events(
event_type=event_type,
event_data=json.dumps(event_data),
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)
async def _add_test_statistics(hass: HomeAssistant, instance: recorder.Recorder):
"""Add multiple statistics to the db for testing."""
utcnow = dt_util.utcnow()
five_days_ago = utcnow - timedelta(days=5)
eleven_days_ago = utcnow - timedelta(days=11)
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
with recorder.session_scope(hass=hass) as session:
for event_id in range(6):
if event_id < 2:
timestamp = eleven_days_ago
state = "-11"
elif event_id < 4:
timestamp = five_days_ago
state = "-5"
else:
timestamp = utcnow
state = "0"
session.add(
StatisticsShortTerm(
start=timestamp,
state=state,
)
)
async def _add_test_recorder_runs(hass: HomeAssistant, instance: recorder.Recorder):
"""Add a few recorder_runs for testing."""
utcnow = dt_util.utcnow()
five_days_ago = utcnow - timedelta(days=5)
eleven_days_ago = utcnow - timedelta(days=11)
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
with recorder.session_scope(hass=hass) as session:
for rec_id in range(6):
if rec_id < 2:
timestamp = eleven_days_ago
elif rec_id < 4:
timestamp = five_days_ago
else:
timestamp = utcnow
session.add(
RecorderRuns(
start=timestamp,
created=dt_util.utcnow(),
end=timestamp + timedelta(days=1),
)
)
async def _add_test_statistics_runs(hass: HomeAssistant, instance: recorder.Recorder):
"""Add a few recorder_runs for testing."""
utcnow = dt_util.utcnow()
five_days_ago = utcnow - timedelta(days=5)
eleven_days_ago = utcnow - timedelta(days=11)
await hass.async_block_till_done()
await async_wait_recording_done(hass, instance)
with recorder.session_scope(hass=hass) as session:
for rec_id in range(6):
if rec_id < 2:
timestamp = eleven_days_ago
elif rec_id < 4:
timestamp = five_days_ago
else:
timestamp = utcnow
session.add(
StatisticsRuns(
start=timestamp,
)
)
def _add_state_and_state_changed_event(
session: Session,
entity_id: str,
state: str,
timestamp: datetime,
event_id: int,
) -> None:
"""Add state and state_changed event to database for testing."""
session.add(
States(
entity_id=entity_id,
domain="sensor",
state=state,
attributes="{}",
last_changed=timestamp,
last_updated=timestamp,
created=timestamp,
event_id=event_id,
)
)
session.add(
Events(
event_id=event_id,
event_type=EVENT_STATE_CHANGED,
event_data="{}",
origin="LOCAL",
created=timestamp,
time_fired=timestamp,
)
)