KEFCore: roadmap

The roadmap can be synthetized in the following points:

• [x] Create a first working provider starting from the code of InMemory provider
• [x] Extends the first provider with new features able to create Apache Kafka™ Streams topology to retrieve information
• [x] Add external package to manage data serialization
• [x] Add Avro external package to manage data serialization
• [x] Add Protobuf external package to manage data serialization
• [x] Add Templates package
• [x] Separate singleton-scoped from context-scoped options — correct EF Core Service Provider cache behavior (#438)
• [x] Topic naming convention and attributes — KEFCoreTopicAttribute, KEFCoreTopicPrefixAttribute, fluent API (#417)
• [x] Event management convention — per-entity opt-out via KEFCoreIgnoreEventsAttribute (#501)
• [x] ComplexType equality enforcement convention (#494)
• [x] ComplexType converter registration convention (#417)
• [x] StreamsManager factory — lifecycle managed per ApplicationId via IKEFCoreCluster (#501)
• [x] ComplexType LINQ binding — TryBindMember extension, KEFCoreComplexTypeProjectionExpression (#491)
• [x] TranslateRightJoin implementation — EF Core 10 requirement (#514)
• [ ] Remove deprecated options: UseCompactedReplicator, UseGlobalTable, ManageEvents, ConsumerConfig, DefaultConsumerInstances
• [ ] Full documentation update for new conventions and options architecture

Query engine improvements

Store-level query optimizations (#496, #497)

These optimizations bypass full table scans by delegating work directly to the underlying Kafka Streams state store (RocksDB). Each optimization is independently controlled via a KafkaOptionsExtension flag.

• [x] Single key lookup (UseStoreSingleKeyLookup, default: true) — PK equality predicate → direct store.get(key) instead of full scan + Where (#497)
• [x] Key range scan (UseStoreKeyRange, default: true) — PK range predicate (>=, <=) → store.range(from, to) iterator (#497)
• [x] Reverse iteration (UseStoreReverse, default: true) — Last()/LastOrDefault() without range → native store.reverseAll() iterator (#497)
• [x] Reverse range scan (UseStoreReverseKeyRange, default: true) — Last()/LastOrDefault() with PK range → store.reverseRange(from, to) with inverted bounds (#497)
• [ ] Prefix scan (UseStorePrefixScan, default: false, experimental) — LIKE 'prefix%' on PK string or partial composite PK equality → store.prefixScan(prefix, serializer). Requires string-serialized keys for correct lexicographic behavior. (stub introduced in #524, implementation tracked in #541)

KTable-level query optimizations (#496)

These optimizations require Kafka Streams topology awareness. KTable joins operate only on record keys (PK), require co-partitioned topics, and must be explicitly enabled.

• [ ] KTable inner/left/right join (UseStoreKTableJoin, default: false) — EF Core Join/LeftJoin/RightJoin on PK-to-PK conditions → KTable.join() / KTable.leftJoin() native API (#537)

Future / topology-dependent

These features require pre-configured Kafka Streams topology elements (declared before Streams startup) and are not implementable on-the-fly:

• [ ] Secondary index support via KGroupedTable — opt-in per property via fluent API / attribute; enables optimized Where on non-PK properties (#538)
• [ ] Aggregate push-down (Count, Sum, Min, Max, Average) via KGroupedTable.aggregate() / reduce() (#539)
• [ ] GroupBy push-down via KTable.groupBy() → KGroupedTable (#540)
• [ ] Subquery-to-KTable join — use pre-built KTable results as subquery sources in LINQ