…rate each table completely without exceeding scratch file size limit
…tate of remaining scheduled reduction
…aded just notify the already running task about new items available for processing. This way we avoid task cancellations because of thread pool exhaustion.
…m based on index storage free drive space
…xists or not in it, using a sparse, bit array based set
… used on both RavenDB and Voron.
…t is still the default.
…p to 1000 chars
…of the websockets.
…he size of slices can vary it is better to lose a bit (10%) on smaller slices (less than 20 bytes) and to win big on the bigger ones. Before 16 bytes the gain is -10% After 24 bytes the gain is 10% After 64 bytes the gain is 2x After 128 bytes the gain is 4x. We should control the distribution of typical workloads over time.
…ndexing to allow the query operation to properly resolve conflicts if there is any registered client side conflict listener. All conflicts are indexed but saved with the same '__document_id' field. This way the client always receives the document with ids of conflicts and it is able to resolve it. Also the client can query by using the value from a conflict document.
…way the synchronization is also open for additional extensions.
…letions to SynchronizationBehavior