Fatbobman's Swift Weekly #074
The Cycle of Expectation and Disappointment: Apple's AI Dilemma and Resilience
The Cycle of Expectation and Disappointment: Apple's AI Dilemma and Resilience
A few days ago, Apple announced that its "More Personalized Siri" feature would be delayed until next year. Combined with the continued absence of Swift Assist, it's clear Apple has not fulfilled the AI vision it presented at WWDC 2024. At least in the realm of large language models, Apple has not exhibited its characteristic agility in keeping pace with industry trends.
Yet it's premature to conclude that Apple has lost its capability to respond to this AI wave. Whether by design or not, Apple's transition to unified memory architecture with the introduction of M-series chips has equipped its Mac lineup with remarkable capabilities for local model inference. Furthermore, with the recent Mac Studio capable of supporting up to 512GB of memory, Apple has firmly secured another foothold in this AI transformation.
While NVIDIA remains the preferred choice for model training and inference, Macs have attracted numerous users due to their high integration and energy efficiency. As more AI models become optimized for MLX, Mac performance in large-scale inference tasks is expected to improve further.
Of course, these successes with Mac devices do not fully mitigate Apple's slow response to large language models, but they highlight the inherent strength of diversified tech giants: when one strategy falters, another can thrive. Such strategic diversity provides resilience against rapidly shifting technological waves.
Do large models truly represent the future? Will companies that fail to keep up inevitably be left behind? Though I doubt it, the current media environment clearly pressures enterprises to deliver tangible results in AI development and applications. Thus, even companies with established AI roadmaps may find themselves forced to pivot due to shifting market dynamics and public sentiment.
Consequently, we can expect to see more "PowerPoint-style" visionary announcements, both recently and in coming years. Without setting expectations, the market becomes disappointed; setting expectations but failing to meet them creates further disappointment. This cycle of expectation and disappointment seems to have become the new normal in tech, particularly in the highly uncertain realm of AI.
In the AI domain, striking a balance between maintaining corporate resolve and satisfying market demands grows increasingly difficult. The expectation-disappointment cycle may have already become an industry norm. For developers and enterprises alike, true wisdom lies in embracing innovation without blindly chasing market hype, carefully balancing diligent execution with meeting public expectations. After all, the true value of technology lies not in flashy promises but in the tangible improvements it delivers to users.
Previous Issue｜Newsletter Archive
If you appreciate my work and want to promote your product to the Swift and iOS developer community, sponsoring my platform could be an excellent opportunity for you.
Your support through Patreon, or Buy Me a Coffee helps keep this newsletter and blog running. Thank you for finding this content helpful!
Original
Lazy Initialization @State in SwiftUI: Overcoming Premature Object Creation
The Observation framework has brought native property-level observation to Swift, effectively preventing unnecessary view updates in SwiftUI triggered by unrelated property changes, thereby enhancing application performance. However, since @State does not offer a lazy initialization constructor like @StateObject, it may lead to performance degradation or even logical issues due to the premature construction of instances in certain scenarios. This article explores how to implement a lazy initialization solution for Observable instances using @State.
Recent Recommendations
Dynamically Constructing Generic Types in Swift
Generic types typically require concrete types at compile time, but in certain scenarios, sufficient type information may not be available during compilation. While working on OpenSwiftUI, Kyle Ye discovered that SwiftUI's _ConditionalContent<TrueContent, FalseContent> is exactly such a case. This article presents an implementation approach that dynamically constructs generic types by defining custom Metadata and ProtocolDescriptor types to access and parse type information at runtime.
Designing a Custom Lazy List in SwiftUI with Better Performance
SwiftUI provides lazy containers such as List and LazyVStack, but their performance remains limited when handling large datasets. Additionally, SwiftUI's Layout protocol currently lacks support for lazy loading, making it difficult for developers to build more efficient custom scrolling containers natively. In this article, Matthaus Woolard implements a row-recycling mechanism that dynamically adjusts view positions instead of creating new views, resulting in a high-performance scrolling container optimized for fixed-height, repeating subviews. While this approach has certain limitations in applicable scenarios, its core idea offers valuable insights for improving SwiftUI view performance.
The original code snippets in the article are scattered. To improve readability, I have compiled and organized them into a Gist for easier reference.
Exploring the Long-Standing Issues of Xcode Virtual Directories and My Open Source Tool Solution
For years, Xcode has managed projects using virtual directories, leading to frequent merge conflicts in collaborative development and compatibility issues with CI/CD processes and modern tools. As Apple gradually improves Xcode’s directory management, modern projects have increasingly relied on actual file structures, yet legacy projects still struggle with virtual directories. In this article, ZhgChgLi provides an in-depth analysis of the impact of virtual directories and introduces his open-source tool, XCFolder. This tool automatically converts Xcode's virtual directories into actual file structures, improving project clarity while ensuring compatibility with XcodeGen and Tuist, offering an efficient solution for modern Xcode project management.
SwiftUI Performance - How to Use UIKit
As SwiftUI continues to evolve, more developers are choosing it as the core framework for building applications while incorporating UIKit in performance-critical scenarios for optimization. In this article, Majid Jabrayilov explores how to effectively integrate UIKit within a SwiftUI-based architecture to balance development convenience and runtime efficiency. The article highlights UIHostingConfiguration—a powerful mechanism for efficiently embedding SwiftUI views within UITableViewCell or UICollectionViewCell. This approach allows developers to leverage UIKit’s efficient view recycling while maintaining SwiftUI’s flexibility, enhancing performance when handling large datasets.
How to Clean Xcode Junk and Reclaim Valuable Disk Space on Your Mac
For Apple developers, Xcode often consumes a significant amount of disk space, impacting system performance and development efficiency. In this article, Karin Prater provides a detailed breakdown of the various file types generated by Xcode and offers a range of safe and effective cleaning methods to help developers free up disk space and optimize their development environment.
Building Complex Things
Traditional macOS CI environments, based on physical Mac machines, are susceptible to external interference, leading to environment drift and unstable builds. Paul Samuels explores virtualization as a means to create a more controllable and reusable CI environment. In this article, Paul shares his journey from problem definition and tool research to prototype validation, ultimately achieving a stable and repeatable solution. This article is not just a technical exploration but also an insightful look into the process of building complex systems.
期待与失望的循环：苹果的 AI 困境与韧性
几天前苹果宣布将 “More Personalized Siri” 功能推迟到明年，再结合始终不见踪影的 Swift Assist，很显然，苹果没有达成其在 WWDC 2024 上构建的 AI 愿景。至少在大模型的应用场景上，苹果没有展现出一贯的快速跟随能力。
但就此断定苹果在本次 AI 浪潮中失去了反击的能力还为时尚早。无论是有意还是无意，苹果在转向 M 系列芯片时采用的统一内存机制让其 Mac 系列设备具备了相当不错的本地模型推理能力。而且，随着最新支持 512GB 内存的 Mac Studio 推出，苹果又以另外的身份在本次 AI 变革中取得了一席之地。
虽然英伟达仍是模型训练和推理的首选，但 Mac 设备凭借高集成度与高能效比，已吸引不少用户。随着更多模型优化适配 MLX，其在大模型推理中的表现有望进一步提升。
当然，Mac 设备取得的成绩仍无法掩盖苹果在大模型方向上反应不及时的窘境，但至少表明了一些多管齐下的大企业所具备的天然优势：东边不亮西方亮。这种策略多元性恰恰是科技巨头在面对快速变化的技术浪潮时的一张王牌。
大模型真的代表未来？没能紧跟这股浪潮，企业就一定会被淘汰吗？尽管我并不这样认为，但显然在当下的舆论环境中，消费者、投资者需要企业在大模型开发、应用等领域中交出一份有感的答卷。也就是即便企业有既定的 AI 发展路线，也可能因市场和舆论环境的变化而被迫调整。
于是，我们就在最近的一年以及未来的几年中会看到越来越多的 PPT 式的愿景发布会。不给出期待大家会失望，给出了期待但没有达成大家会继续失望。这种"期待-失望"的循环似乎已成为科技行业的新常态，特别是在 AI 这样充满不确定性的领域。
在 AI 领域，企业既要保持定力，又要让市场满意，这种平衡正变得越来越难，“期待-失望”的循环也许已成为行业的常态。对于开发者和企业来说，或许真正的智慧在于既能拥抱创新浪潮，又不盲目追随市场喧嚣，在踏实做事与满足期待之间找到平衡点。毕竟，技术的真正价值不在于其炫目的承诺，而在于它能为用户带来的实际改变。
如果您发现这份周报或我的博客对您有所帮助，可以考虑通过 爱发电，Buy Me a Coffee 支持我的创作。
原创
让 @State 实现懒加载
Observation 框架为 Swift 带来了原生的属性级观察能力，有效避免了 SwiftUI 中因无关属性变化而引发的多余视图更新，从而提升了应用性能。但由于 @State 并未提供类似 @StateObject 的懒加载构造方式，在某些场景下会因实例过早构建而引起性能损失甚至逻辑问题。本文将探讨如何为 Observable 实例定制一个支持懒加载的 @State 解决方案。
近期推荐
在 Swift 中动态构造泛型类型的方法探索 (Dynamically Constructing Generic Types in Swift)
泛型类型通常需要在编译期确定具体类型，但在部分特殊场景下，编译期无法提供足够的类型信息。Kyle Ye 在开发 OpenSwiftUI 时，发现 SwiftUI 中的 _ConditionalContent<TrueContent, FalseContent> 就属于这种情况。本文介绍了一种通过定义自定义的 Metadata 和 ProtocolDescriptor 类型来访问和解析运行时类型信息，以动态构建泛型类型的实现方案。
巧用视图复用，提升 SwiftUI 滚动列表性能 (Designing a Custom Lazy List in SwiftUI with Better Performance)
SwiftUI 提供了 List、LazyVStack 等惰性容器，但在处理大规模数据时，这些组件的性能仍然存在局限性。此外，SwiftUI 目前的 Layout 协议尚未支持惰性加载，使得开发者难以通过原生方式构建更高效的自定义滚动容器。在本文中，Matthaus Woolard 采用行复用机制，通过动态调整视图位置而非创建新视图，实现了一种适用于固定高度、重复子视图的高性能滚动容器。这种方法虽对适用场景有所限制，但其核心思路对于优化 SwiftUI 视图性能具有重要的借鉴价值。
原文中的代码较为分散，我已将其整理并汇总至 Gist，方便阅读和参考。
XCode 虚拟目录万年问题探究与我的开源工具解决方案
长期以来，Xcode 采用虚拟目录管理项目，不仅导致多人协作时合并冲突频繁，也对 CI/CD 流程和现代工具的兼容性不友好。随着 Apple 逐步优化 Xcode 目录管理，现代 Xcode 项目已更多依赖实际文件结构，但对于历史遗留项目，虚拟目录仍然是一个棘手的问题。在本文中，ZhgChgLi 详细分析了虚拟目录的影响，并介绍了他的开源工具 XCFolder。该工具可自动将 Xcode 项目中的虚拟目录转换为实际文件目录，使项目结构更加清晰，同时兼容 XcodeGen 和 Tuist，为 Xcode 项目的现代化管理提供了一种高效的解决方案。
SwiftUI 性能优化：如何有效结合 UIKit (SwiftUI Performance - How to use UIKit)
随着 SwiftUI 功能的不断丰富，越来越多的开发者选择以 SwiftUI 为核心构建应用，同时在性能关键的场景中引入 UIKit 以优化体验。在本文中，Majid Jabrayilov 探讨了如何在 SwiftUI 架构下合理结合 UIKit，以兼顾开发便捷性和运行效率。文章重点介绍了 UIHostingConfiguration——一个能够在 UITableViewCell 或 UICollectionViewCell 内高效集成 SwiftUI 视图的机制。通过这种方式，开发者可以充分利用 UIKit 的高效视图复用，同时保持 SwiftUI 组件的灵活性，从而提升应用在处理大规模数据时的流畅度。
清理 Xcode 垃圾文件，释放 Mac 磁盘空间 (How to Clean Xcode Junk and Reclaim Valuable Disk Space on Your Mac)
对于苹果开发者来说，Xcode 占用大量磁盘空间是一个常见问题，影响系统性能和开发效率。Karin Prater 在本文中详细解析了 Xcode 生成的各种文件类型，并提供了一系列安全、高效的清理方法，帮助开发者释放磁盘空间，优化开发环境。
从零到一：构建复杂系统的探索之路 (Building Complex Things)
传统的 macOS CI 运行环境（基于 Mac 物理机）易受到外部干扰，导致环境漂移和构建不稳定。Paul Samuels 旨在通过引入虚拟化技术，使 CI 运行环境更加可控和可复用。在这篇文章中，Paul 分享了他的构建过程，从最初的问题定义、工具调研，到原型验证，最终实现了一个稳定且可重复使用的解决方案。这不仅是一次技术探索，更展示了复杂系统构建的全过程。
Apple 开发者官方公众号现已上线
欢迎关注 Apple 开发者的微信官方公众号，获取面向中文开发者社区的最新新闻、公告和活动动态。
