每日新聞摘要

📝 今日焦點 📝 Gradle 已成為專案中隱性且影響深遠的基礎設施：量測與精簡建置外掛、優化任務圖與快取能直接提升開發效率與維護性，忽視會導致慢速建置與隱性技術債。 📝 應用「衰退」（app rot）非偶發現象：持續監控崩潰與使用者指標、定期更新相依項目與跨版本自動化測試可防止品質與評分逐步下滑。 📝 Android 14/15 前景服務規則改動嚴格：必須在 manifest 指定 foregroundServiceType 並在每次會話取得 mediaProjection 授權；需調整服務生命週期與權限流程以避免例外與隱私問題。 📝 高階 Android 面試重點在系統性思考而非語法：能夠闡述設計決策、預見失效模式與衡量營運成本，對面試結果影響大。 🔧 效能最佳化 🔧 OCR 優化案例提供可復用策略：降解析度、以 ROI 先行偵測、依腳本選擇辨識器、以 Kotlin 協程並行處理並以感知雜湊快取結果，實際將處理時間從 ~4s 降到 400–800ms，示範如何針對瓶頸做系統性優化。 🏗 架構設計 🏗 React Native Android bridge：釐清 bridge 模式、方法暴露與雙向通訊能降低與原生互通時的複雜度，改善跨層偵錯與相容性。 🏗 OkHttp Interceptor 的角色：把 HTTP 的認證、日誌、快取與度量等橫切關注點集中管理，可提升可維護性與一致性，並簡化請求/回應相關的除錯。 🏗 Multiplatform 的「隱形稅」提醒：採用前需量化跨團隊協調、測試與建置複雜度的長期成本，避免被短期代碼共用誘惑而忽視營運負擔。 🏗 Server Driven UI（SDUI）：將介面描述伺服器化能加速迭代與實驗，但會把渲染與 API 設計的複雜度移回後端與客戶端，需在靈活性與一致性間做取捨。 🏗 Clean Architecture 實證（jOOQ + DGS）：以 repository 介面作為合約、呈現層作為適配器與協議中立的錯誤處理，示範如何在不改變業務邏輯下替換基礎設施。 ✅ 測試與品質 ✅ Compose 預覽為「無聲的信任破壞者」：預覽可能隱藏動態狀態與裝置差異，應以真機/模擬器驗證並在工作流程中明確預覽的適用範圍以降低返工風險。 🚀 語言演進 🚀 KoreDB（純 Kotlin 的 NoSQL）提出 Kotlin 原生資料層的可能性：作為實驗性替代方案值得關注，但在採用前需評估成熟度與生態支援。 🚀 Compose Multiplatform + Conveyor 的打包教學：為桌面應用產生可發佈工件提供實用流程，對多平台發佈工作流有直接參考價值。

📝 Today's highlights 📝 Treat Gradle as first‑class infra: measure build impact, audit plugins/scripts, and optimize task graph and caching to avoid slow builds and hidden technical debt that degrade developer productivity. 📝 App rot is gradual but predictable: continuous crash/metric monitoring, regular dependency updates, and cross‑version automated tests prevent silent declines in performance and store ratings. 📝 Android 14/15 foreground service changes are breaking: manifests must declare foregroundServiceType and mediaProjection consent must be reobtained per session—update lifecycle, permission flows, and testing to avoid runtime exceptions and privacy regressions. 📝 Senior Android interviews prioritize systems thinking over API trivia: articulate trade‑offs, failure modes, and operational costs rather than reciting syntax. 🔧 Performance optimizations 🔧 OCR pipeline case study: downscale screenshots, run ROI text detection, pick script‑specific ML recognizers, parallelize with Kotlin coroutines, and cache via perceptual hashing—reduced end‑to‑end latency from ~4s to 400–800ms, illustrating targeted bottleneck mitigation. 🏗 Architecture & design 🏗 React Native Android bridge: document bridge patterns, method exposure and two‑way comms to reduce native interop complexity and improve cross‑platform compatibility and debugging. 🏗 OkHttp Interceptor: centralize authentication, logging, caching, and metrics in interceptors to enforce consistent HTTP behavior and simplify debugging. 🏗 Multiplatform "invisible tax": assess long‑term coordination, testing, and build complexity costs before committing—short‑term code reuse can incur sustained operational overhead. 🏗 Server‑Driven UI (SDUI): enables rapid iteration and experimentation but shifts complexity to API and client rendering—design APIs and rendering strategy with maintainability in mind. 🏗 Clean Architecture proof: by keeping a repository interface contract and adapter presentation layer (with protocol‑neutral error handling), infrastructure like jOOQ or GraphQL can be swapped without touching business logic. ✅ Testing & quality ✅ Compose previews are a "silent trust killer": they can hide dynamic states and device differences—validate UIs on real devices/emulators and embed preview limits into team workflows to avoid late surprises. 🚀 Language & ecosystem 🚀 KoreDB (Kotlin NoSQL) is an experiment in Kotlin‑native data layers: worth following for alternative mobile storage patterns but evaluate maturity and ecosystem support before adoption. 🚀 Compose Multiplatform + Conveyor packaging guide: documents a path to produce distributable desktop artifacts, providing a practical workflow for shipping multiplatform desktop apps.