Overview

In-memory event store using STM (Software Transactional Memory) for fast, concurrent access. Ideal for testing, development, and scenarios where events don't need to survive process restarts.

⚠️ Data is lost on process termination - not suitable for production use.

Quick Start

import Hindsight.Store.Memory (newMemoryStore)
import Hindsight

main :: IO ()
main = do
  -- Create store
  store <- newMemoryStore

  -- Insert events (see Hindsight.Store for details)
  streamId <- StreamId <$> UUID.nextRandom
  let event = mkEvent MyEvent myData
  result <- insertEvents store Nothing $ singleEvent streamId NoStream event

  -- Subscribe to events
  handle <- subscribe store matcher (EventSelector AllStreams FromBeginning)
  -- ... process events ...

Use Cases

When to use Memory store:

• Unit and integration tests (fast, isolated)
• Development and prototyping
• Temporary event processing pipelines
• Scenarios where persistence isn't required

When NOT to use Memory store:

• Production systems requiring durability
• Multi-process applications (each process has separate state)
• Long-running services that can't afford data loss

Trade-offs

Advantages:

• Fastest performance (no I/O)
• No external dependencies
• Simple setup (single function call)
• Thread-safe via STM

Limitations:

• Data lost on process termination or crash
• Memory usage grows with event count
• Single-process only (no sharing between instances)
• No built-in persistence or snapshots

Implementation

Events and stream metadata stored in-memory using STM TVars. Subscriptions use STM retry mechanism for efficient event notification.