Ahoo Wang

1. 01github.com/Ahoo-Wang/Wow★ 305

   Wow

   Put the domain model at the heart of the system.

   SourceDocs

   CartSpec.ktkotlin

   class CartSpec : AggregateSpec<Cart, CartState>(
       {
           on {
               val ownerId = generateGlobalId()
               val addCartItem = AddCartItem(
                   productId = "productId",
                   quantity = 1,
               )
               givenOwnerId(ownerId)
               whenCommand(addCartItem) {
                   expectNoError()
                   expectEventType(CartItemAdded::class)
                   expectState {
                       items.assert().hasSize(1)
                   }
                   // …

   Problem

   Complex business rules need traceable state changes and coordination across aggregates.

   Decision

   Model every write as command → domain decision → event → projection, and verify behavior with Given → When → Expect specifications.

   Trade-off

   Worth it where history and rules justify Event Sourcing; plain CRUD may not need the cost.

   Domain examples & testsKaiCode’26 Excellent Award

2. 02github.com/Ahoo-Wang/CosId★ 651

   CosId

   Unique IDs without a coordination bottleneck.

   SourceDocs

   SegmentChainId · segment chainlock-free

   SnowflakeId bit layout64 bit

   A 64-bit ID: 1 unused sign bit, a 41-bit millisecond timestamp (about 69 years), a 10-bit machine ID (1,024 nodes) and a 12-bit sequence (4,096 IDs per millisecond).

   timestamp41 bit

   ≈ 69 years

   machineId10 bit

   1,024 nodes

   sequence12 bit

   4,096 / ms

   ops/s · SegmentChainId

   127M+

   ops/s · CosIdGenerator

   15.6M

   IDs/s per node · Snowflake

   4.096M

   JMH benchmarks, CosId docs

   Problem

   Distributed ID generation must handle machine identity, clock rollback and coordination overhead.

   Decision

   Two families of generators. SegmentChainId links ID segments into a lock-free chain; a PrefetchWorker keeps a safe distance of segments ready, doubling it when the generator runs hungry and halving it when the chain is full. SnowflakeId adds automatic machine-ID allocation and clock-rollback handling.

   Trade-off

   Pick the algorithm and coordination store per deployment; component benchmarks are not application throughput guarantees.

   Algorithm designJMH benchmarks

3. 03github.com/Ahoo-Wang/CoSky★ 96

   CoSky

   Service governance without a heavy footprint.

   SourceDocs

   CoSky read path0 RTT

   ops/s · config read, local cache

   706.7M

   ops/s · config read, Redis

   58.8k

   faster on the hot path

   ≈12,000×

   JMH, 8 threads, Apple M4 Pro, CoSky 6.0.0

   Problem

   Services need discovery and shared configuration without another heavy cluster to operate.

   Decision

   Build on Redis: in-process caches kept fresh by Pub/Sub, integrated with Spring Cloud.

   Trade-off

   A focused choice where Redis already fits; availability follows the Redis deployment you choose.

   Architecture docsService examples

1. 01

   Make intent visible

   Start from the domain. Express business intent in explicit, testable models.

2. 02

   Give change a boundary

   Make responsibilities and dependencies explicit, so the system can evolve without losing integrity.

3. 03

   Let evidence lead

   Explain a design with code, tests and benchmarks — and state its costs and limits.

All repositories on GitHub