Exposed JDBC Caffeine Cache

This adapter combines blocking JDBC or suspended JDBC with in-process Caffeine. It implements the common cache repository contract without hiding lifecycle or durability choices.

Problem

Cache and database state can diverge when a miss, write, invalidation, timeout, cancellation, or shutdown completes only partly. The adapter supplies the integration; the application still defines acceptable staleness and failure behavior.

When to use it

Use this module when in-process Caffeine is the chosen backend and the persistence path is blocking JDBC or suspended JDBC. Compare all six adapters in the cache selection guide before adding infrastructure.

Coordinates

dependencies {
    implementation(platform("io.github.bluetape4k:bluetape4k-dependencies:<version>"))
    implementation("io.github.bluetape4k.exposed:bluetape4k-exposed-jdbc-caffeine")
}

Users select the central BOM version; this page records the stable 1.11 source line.

Core concepts

get and getAll are read-through operations. Cache-aside is an application-owned DB update followed by invalidation. True write-through requires the configured writer to finish DB persistence before returning. Write-behind accepts a delayed durability window. Ordinary put must not be called write-through without that writer policy. This adapter is local-only; it does not synchronize entries across JVMs.

Quick start

Extend AbstractJdbcCaffeineRepository and provide table, extractId, ResultRow.toEntity, and the update/insert DSL hooks. Construct it with a reviewed LocalCacheConfig.

repository.use { repo ->
    val current = repo.get(id)          // miss -> DB loader -> cache
    current?.let { repo.put(id, it) }   // durability follows write mode
    repo.invalidate(id)                 // cache-only by default
}

API by task

Task	API
Cache-backed read	containsKey, get, getAll
DB bypass	findByIdFromDb, findAllFromDb, countFromDb
Policy-controlled write	put, putAll
Eviction	invalidate, invalidateAll, clear
Entity mapping	ResultRow.toEntity, extractId, update/insert hooks
Lifecycle	close

Recommended patterns

Keep one repository per cache namespace and close it with the application. For cache-aside, commit the DB transaction before invalidating. Treat write-through failure as an incomplete write. Before write-behind, expose queue/dead-letter state and define the shutdown drain budget. Keep key prefixes and serialized value formats stable.

Integrations

The module joins Exposed JDBC, the shared cache foundation, and in-process Caffeine. Loader/writer adapters own the cache-to-table bridge; the application owns the surrounding service transaction and client lifecycle.

Configuration

LocalCacheConfig controls namespace, TTL/expiry, cache/write mode, and backend limits. Values remain in-process, so no Redis wire codec is used. Validate positive durations and bounded batch/queue sizes at startup.

Failure modes

A full write-behind channel rejects the operation before updating the cache. Process failure can still lose accepted entries not yet flushed. A cache hit can be stale, a backend success can be followed by DB failure, and DB commit can be followed by failed invalidation. Record these as distinct states.

Operations

The repository owns the Caffeine cache and bounded write-behind scope. close() drains queued writes within its shutdown boundary. Monitor hit/miss, backend latency, retries/timeouts, queue depth, rejected or dead-letter writes, invalidation lag, and shutdown drain. Set separate SLOs for cache and database paths.

Testing

Use isolated cache and database fixtures with unique namespaces. Prove miss loading, partial getAll misses, each enabled write mode, cache-only invalidation, TTL, partial failure, and lifecycle cleanup.

./gradlew :bluetape4k-exposed-jdbc-caffeine:test

Workshops and learning path

1. Read Exposed cache foundation.
2. Choose a backend in cache selection.
3. Implement read-through plus cache-only invalidation first.
4. Add partial-failure and shutdown tests before near cache or write-behind.
5. Continue with exposed-workshop.

Limitations

The adapter does not create a distributed transaction, provision the backend, migrate stored cache values, or decide whether stale data is safe. It is local-only; it does not synchronize entries across JVMs.

Release diagrams

These diagrams are loaded directly from README assets published with the 1.11.0 release and pinned to its immutable commit. They describe this manual’s released structure and runtime flows, not later Snapshot changes. Select a preview to open the SVG at the same release commit.

JDBC Caffeine local cache architecture diagram

Release README: exposed/jdbc-caffeine/README.md

Write Strategy Flows diagram

Release README: exposed/jdbc-caffeine/README.md

Sources

• Module README
• Abstract repository
• Module build