mirror of
https://git.ianrenton.com/ian/spothole.git
synced 2026-08-06 02:21:42 +00:00
Improve use of URL cache #118
This commit is contained in:
+9
-9
@@ -4,17 +4,17 @@ 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
|
||||
|
||||
CACHE_DIR = "./cache/"
|
||||
|
||||
|
||||
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
|
||||
@@ -29,15 +29,15 @@ class DataStore:
|
||||
self._solar = None
|
||||
|
||||
def setup(self):
|
||||
Path(self._CACHE_DIR).mkdir(parents=True, exist_ok=True)
|
||||
Path(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")
|
||||
self._solar = diskcache.Cache(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")
|
||||
self._status = diskcache.Cache(CACHE_DIR + "status")
|
||||
if "status_data" not in self._status:
|
||||
self._status.add("status_data", {})
|
||||
self.status_data = self._status.get("status_data")
|
||||
@@ -46,25 +46,25 @@ class DataStore:
|
||||
# but there's no need for a TTL since old data is better than no data. We need to key on both SIG and reference,
|
||||
# and trying to do two layers of dict in diskcache absolutely destroys performance with unpickling huge dicts,
|
||||
# so we have an ugly "SIG:ref" syntax for keys to keep it a single level.
|
||||
self.sigrefs = diskcache.Cache(self._CACHE_DIR + "/sigrefs")
|
||||
self.sigrefs = diskcache.Cache(CACHE_DIR + "sigrefs")
|
||||
logging.info(f"Loaded data for %d SIG references.", len(self.sigrefs))
|
||||
|
||||
# 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")
|
||||
self.callsigns = diskcache.Cache(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_dir=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_dir=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()))
|
||||
|
||||
|
||||
Reference in New Issue
Block a user