Improved performance of aggregated stream projection.
Fixed bug in return of AggregatedStream.projector.run which would always re-run it in the executor right away.
BREAKING CHANGE: Extended the RunOnNotification protocol
Added interested_in_notification(dict) -> bool
Added take_notification_hint(dict) -> bool
BREAKING CHANGE: Changed DB schema of aggregated streams
Added columns; you will have to do a downtime deployment for these changes
Use the following command to generate a migration script:
python -m depeche_db generate-migration-script <PREV-VERSION> 0.11 --message-store=<NAME> --aggregated-stream=<STREAM-NAME> --aggregated-stream=<ANOTHER...>
0.10.3
Use one insert operation per batch in aggregated stream update instead of one per message
0.10.2
Make exception on Message ID mismatch in MessageStore.synchronize more specific.
0.10.1
Fix performance of aggregated stream projection
0.10.0
Add batched acknowledgement strategy
Performance improvements
Use less connections on load in subscription
Removed overly defensive check in acknowledgement operation
Pre-calculate hot SqlA queries
Improved Performance in pydantic (de)serialization
0.9.1
Fix passing of batch size in stream/subscription factory
0.9.0
Add a stateful reader to aggregated streams
Make RunOnNotification implementations fair by exiting after time budget is used.
0.8.3
Logging fix in LogAndIgnore error handler
0.8.2 (never released)
Compatibility with psycopg 2 and 3
0.8.1
Remove direct dependency on psycopg2 -> allow using psycopg2-binary as well
0.8.0
Allow ack in client transaction: This allows for exactly once delivery semantics