import logging from datetime import datetime from threading import Event import pytz from core.data_store import DATA_STORE logger = logging.getLogger(__name__) class SIGRefDataProvider: """Generic SIG reference data provider class. Subclasses of this query the individual URLs or files for data.""" def __init__(self, sig_name, provider_config): """Constructor""" self.sig_name = sig_name self.enabled = provider_config["enabled"] self.last_update_time = datetime.min.replace(tzinfo=pytz.UTC) self.status = "Not Started" if self.enabled else "Disabled" self.reference_count = 0 self._stop_event = Event() def start(self): """Start the provider. This should return immediately after spawning threads to access the remote resources""" raise NotImplementedError("Subclasses must implement this method") def stop(self): """Stop any threads and prepare for application shutdown. Subclasses should implement this method and call super().""" self._stop_event.set() def _add_data(self, new_data): """Add all the provided reference data objects to the data store.""" # with transact() batches all writes together to save making thousands of individual sqlite writes. However, # that means that each provider holds the lock while it writes, and the default behaviour for other attempted # transact()s is to fail if they can't get the lock (?!). This behaviour is fixed by retry=True. with DATA_STORE.sigrefs.transact(retry=True): for d in new_data: DATA_STORE.sigrefs.set(f"{self.sig_name}:{d.id}", d) # For the big data sources, loading will take a few minutes. If we want to shut down the software neatly # within the first few minutes of startup, we need a way to abort this expensive process of filling up the # disk cache. if self._stop_event.is_set(): break self.reference_count = len(new_data) logger.info(f"Loaded {self.reference_count} references for {self.sig_name} into the data store.")