Files
spothole/core/data_store.py
T

80 lines
3.8 KiB
Python

import logging
from pathlib import Path
import diskcache
from core.config import MAX_SPOT_AGE, MAX_ALERT_AGE
from core.constants import SIGS
from core.live_data_cache import LiveDataCache
from data.solar_conditions import SolarConditions
class DataStore:
"""Data caching/storage object. Handles storage of spots, alerts, solar conditions, SIG reference data, and callsign
lookup data using different caching strategies for each."""
def __init__(self):
self._CACHE_DIR = "./cache"
self._MAX_SPOT_COUNT = 100000
self._MAX_ALERT_COUNT = 100000
self._SPOT_ALERT_SNAPSHOT_INTERVAL_SEC = 300
self._CALLSIGN_DATA_TTL_SEC = 30 * 24 * 60 * 60
self.alerts = None
self.spots = None
self.callsigns = None
self.sigrefs = None
self.status_data = None
self._status = None
self.solar_conditions = None
self._solar = None
def setup(self):
Path(self._CACHE_DIR).mkdir(parents=True, exist_ok=True)
# Standard disk cache for solar data and status data, but each cache contains only a single object which we
# expose to the wider application
self._solar = diskcache.Cache(self._CACHE_DIR + "/solar")
if "solar_conditions" not in self._solar:
self._solar.add("solar_conditions", SolarConditions())
self.solar_conditions = self._solar.get("solar_conditions")
self._status = diskcache.Cache(self._CACHE_DIR + "/status")
if "status_data" not in self._status:
self._status.add("status_data", {})
self.status_data = self._status.get("status_data")
# Standard disk cache for SIG ref data. Separate provider threads will repopulate theis on a regular basis
# but there's no need for a TTL since old data is better than no data. This is a two-layer dict, keys are SIG
# name and then reference ID, with the final value being a SIGRef object.
self.sigrefs = diskcache.Cache(self._CACHE_DIR + "/sigrefs")
for k in list(self.sigrefs.iterkeys()):
logging.info(f"Loaded data for %d references in %s SIG.", len(self.sigrefs[k]), k)
# Standard disk cache for callsign data. This data does have a TTL to trigger an occasional re-lookup.
# Old data *is* better than no data, but we can't have a background thread re-looking-up every callsign
# we've seen, so we rely on them timing out and this triggering another lookup.
self.callsigns = diskcache.Cache(self._CACHE_DIR + "/callsigns")
logging.info(f"Loaded data for %d callsigns.", len(self.callsigns))
# Special caches for spots and alerts, which have TTL and write snapshots to disk at an interval. We
# specifically load these caches *last* so that any sigref and callsign data is already loaded from disk cache
# before the spots and alerts are live in the system.
self.spots = LiveDataCache(maxsize=self._MAX_SPOT_COUNT, ttl=MAX_SPOT_AGE,
snapshot_dir=self._CACHE_DIR + "/spots",
snapshot_interval_sec=self._SPOT_ALERT_SNAPSHOT_INTERVAL_SEC)
logging.info(f"Loaded %d spots from a previous run.", len(self.spots.keys()))
self.alerts = LiveDataCache(maxsize=self._MAX_ALERT_COUNT, ttl=MAX_ALERT_AGE,
snapshot_dir=self._CACHE_DIR + "/alerts",
snapshot_interval_sec=self._SPOT_ALERT_SNAPSHOT_INTERVAL_SEC)
logging.info(f"Loaded %d alerts from a previous run.", len(self.alerts.keys()))
def close(self):
self.spots.close()
self.alerts.close()
self._solar.close()
self._status.close()
self.sigrefs.close()
self.callsigns.close()
# Global object
DATA_STORE = DataStore()