{"slug": "swift-1-0-stable-apis-rust-1-95-language-features", "title": "Swift 1.0 Stable APIs + Rust 1.95 Language Features", "summary": "Swift's server ecosystem reached two 1.0 milestones with the release of Swift System Metrics, providing stable APIs for CPU, memory, and file descriptor monitoring. Rust 1.95 introduced new zeroed allocation APIs and iterator optimizations, while a new CLI tool called add-mcp solves MCP configuration fragmentation across multiple AI agents. Deno 2.7 stabilized the Temporal API and added native Windows ARM support.", "body_md": "This week's releases are less about flashy new capabilities and more about the quieter, more valuable kind of progress: stable APIs that let you stop maintaining workarounds. Swift's server ecosystem crossed two 1.0 milestones simultaneously, Rust 1.95 landed language features that eliminate entire dependency categories, and a small CLI tool solved a genuinely annoying MCP configuration problem. Here's what's worth your attention.\n\nSwift System Metrics gives you CPU usage, memory consumption, and file descriptor counts out of the box for any Swift service, wired into the backend-agnostic Swift Metrics API. That means it routes to Prometheus, OpenTelemetry, or whatever metrics sink you're already running—no vendor lock-in, no custom `/proc`\n\nparsing.\n\nBefore this, Swift backend observability meant writing your own instrumentation layer or vendoring fragmented packages with no stability guarantees. The 1.0 tag changes that calculus: you get a stable, maintained API, a Grafana dashboard template included in the repo, and lifecycle integration via ServiceLifecycle so metrics collection starts and stops cleanly with your service.\n\n**Verdict: Ship now.** Requires Swift 5.9+ and the Swift Metrics ecosystem. Drop it into any Swift backend and get production observability from first deploy. The Grafana template alone saves an hour of dashboard setup.\n\nTemporal moves out of unstable in Deno 2.7, which means date arithmetic, timezone handling, and calendar operations that don't involve `Date`\n\nhacks are now production-ready without flags. Native Windows ARM builds land as well, removing emulation overhead for that platform. Node.js compatibility gaps in worker threads, `child_process`\n\n, and `zlib`\n\nalso close in this release, which matters if you're migrating existing Node projects.\n\nThe subprocess spawn APIs—`Deno.spawn()`\n\nas a simpler alternative to `Deno.Command()`\n\n—also appear in 2.7 but remain unstable. That's the one thing worth waiting on.\n\n**Verdict: Ship Temporal now, wait on spawn APIs.** If you have any date/timezone logic in a Deno project, replace it with Temporal immediately—it's significantly cleaner and now carries a stability guarantee. Hold off on `spawn()`\n\nuntil it stabilizes; `Deno.Command()`\n\nstill works fine. Run `deno upgrade`\n\nto get 2.7.\n\nIf you're running multiple AI agents—Claude Code, Cursor, VS Code with Copilot, Codex—and you've added even two or three MCP servers, you've already dealt with the configuration fragmentation problem. Each tool has its own config file format, its own location, its own schema. `add-mcp`\n\nis a single CLI that auto-detects which agents you have installed and writes the appropriate config for all of them in one pass.\n\nThe actual command is `npx add-mcp`\n\nfollowed by an MCP server URL or npm package name. It supports 9 agents at v1. There's no daemon, no persistent process—it's a config file writer that knows where everyone keeps their configs.\n\n**Verdict: Adopt immediately if you manage MCPs across multiple tools.** The only requirements are having a supported agent installed and knowing your MCP server's URL or package name. This is the kind of glue tooling that seems minor until you've spent 20 minutes hunting down why Claude Code picked up a server that Cursor didn't.\n\nRust's latest stable adds `Box::new_zeroed()`\n\n, `Rc::new_zeroed()`\n\n, and `Arc::new_zeroed()`\n\nas first-class APIs, replacing the `vec![0; n]`\n\npattern or unsafe manual zeroing that previously handled this case. Iterator specialization for `TrustedLen`\n\nalso lands, letting `Iterator::eq`\n\nskip redundant work when the iterator length is statically known.\n\nThe more important change is a behavior fix: `iter::Repeat::last()`\n\npreviously looped infinitely because `Repeat`\n\nhas no end. That silent hang is now a panic—breaking if you relied on the old behavior, but the old behavior was a correctness trap. If you hit this in production without realizing it, you now get a clear failure instead of a hung process.\n\n**Verdict: Ship now.** These are additive stabilizations with no migration cost for the new APIs. The `Repeat::last`\n\nchange is technically breaking but fixes something that should never have worked. Update to latest stable Rust and grep your codebase for `.last()`\n\ncalls on `Repeat`\n\niterators if you have any doubt.\n\nTwo language features stabilize in Rust 1.95 that reduce real friction in day-to-day code. `cfg_select!`\n\nis a standard library macro for compile-time configuration matching—it replaces the `cfg-if`\n\ncrate that most Rust projects have been pulling in for years to handle conditional compilation across platforms or feature flags. One fewer dependency, same functionality, now in stdlib.\n\nIf-let guards extend match expressions to support `if let`\n\nbindings inside guard position—`match x { Some(v) if let Ok(n) = parse(v) => ... }`\n\nis now valid syntax. This closes a real expressiveness gap where you'd previously need nested matches or awkward intermediate variables to combine pattern matching with fallible bindings.\n\n**Verdict: Ship both now.** `cfg_select!`\n\nis a direct drop-in for `cfg-if`\n\n—remove the crate dependency and update the macro call. If-let guards are opt-in syntax; nothing breaks, you adopt where it simplifies existing match arms. Both are stable, backward-compatible, and available on Rust 1.95.\n\nSwift Configuration 1.0 gives you a single `ConfigReader`\n\nabstraction that reads from environment variables, config files, command-line arguments, or any custom provider—with explicit precedence ordering between them. Libraries can accept a `ConfigReader`\n\nwithout knowing or caring where the config comes from, which is the architectural property that matters here.\n\nThe practical additions on top of that abstraction are useful: hot-reload without refactoring read call sites, secret redaction for logging, and access logging to audit what your service actually reads at startup. The 40+ PRs merged since October give you a sense of how actively this was refined before the 1.0 tag.\n\n**Verdict: Ship now for server apps and CLI tools.** Add the `swift-configuration`\n\npackage and refactor config access to go through `ConfigReader`\n\n. The upfront cost is the refactor; the payoff is never writing environment variable parsing boilerplate again and getting hot-reload essentially for free.\n\nIf this kind of technically grounded coverage of the tools actually shipping in production is useful to you, [Dev Signal publishes it every week at thedevsignal.com](https://thedevsignal.com). Worth subscribing if you'd rather spend your research time building.", "url": "https://wpnews.pro/news/swift-1-0-stable-apis-rust-1-95-language-features", "canonical_source": "https://dev.to/devsignal/swift-10-stable-apis-rust-195-language-features-4e96", "published_at": "2026-06-18 18:16:19+00:00", "updated_at": "2026-06-18 18:29:41.296499+00:00", "lang": "en", "topics": ["developer-tools", "large-language-models", "ai-agents", "artificial-intelligence", "machine-learning"], "entities": ["Swift", "Rust", "Deno", "Temporal", "add-mcp", "Claude Code", "Cursor", "Copilot"], "alternates": {"html": "https://wpnews.pro/news/swift-1-0-stable-apis-rust-1-95-language-features", "markdown": "https://wpnews.pro/news/swift-1-0-stable-apis-rust-1-95-language-features.md", "text": "https://wpnews.pro/news/swift-1-0-stable-apis-rust-1-95-language-features.txt", "jsonld": "https://wpnews.pro/news/swift-1-0-stable-apis-rust-1-95-language-features.jsonld"}}