CompletableFuture异步编排实战

# CompletableFuture异步编排实战 最近优化商品详情页接口，原来是串行查询4个服务，800ms。用CompletableFuture改成并行后，直接降到200ms。 ## 串行 vs 并行 ```java // 改造前：串行调用，慢得要死 public ProductDetail getDetail(Long id) { ProductInfo info = productService.getInfo(id); // 200ms List skus = skuService.list(id); // 200ms List reviews = reviewService.list(id); // 200ms ShopInfo shop = shopService.getInfo(info.getShopId()); // 200ms // 总计约800ms return new ProductDetail(info, skus, reviews, shop); } // 改造后：并行调用 public ProductDetail getDetailAsync(Long id) { CompletableFuture infoFuture = CompletableFuture.supplyAsync(() -> productService.getInfo(id), executor); CompletableFuture> skuFuture = CompletableFuture.supplyAsync(() -> skuService.list(id), executor); CompletableFuture> reviewFuture = CompletableFuture.supplyAsync(() -> reviewService.list(id), executor); // shop依赖info的结果，用thenCompose串联 CompletableFuture shopFuture = infoFuture .thenComposeAsync(info -> CompletableFuture.supplyAsync( () -> shopService.getInfo(info.getShopId()), executor )); // 等所有任务完成 CompletableFuture.allOf(infoFuture, skuFuture, reviewFuture, shopFuture).join(); // 总计约400ms（info+shop串行200+200，其他并行） return new ProductDetail(infoFuture.join(), skuFuture.join(), reviewFuture.join(), shopFuture.join()); } ``` **注意事项**：一定要自定义线程池，别用默认的ForkJoinPool.commonPool()，否则一个慢任务会拖垮所有异步调用。