wayr

Wayland-first windowing toolkit for Rust.

wayr is a minimal, opinionated alternative to winit aimed at kryptic-sh apps — buffr, pikr, and future GUI work.

What it is

• Wayland only. No X11, macOS, Windows, mobile, or web. No XWayland fallback. By construction.
• First-class layer-shell + wl_subsurface embedding. Both are built into the API rather than hidden behind raw FFI, because both are the reason wayr exists.
• Fractional scaling + multi-output + text-input-v3 + cursor-shape + high-res scroll. Wayland-native through and through.
• wgpu-friendly. raw-window-handle 0.6 implementations, no version coupling to any specific GPU crate.

What it isn't

• A drop-in winit replacement. The API shape mirrors winit's ApplicationHandler so consumer code ports mechanically, but wayr has no cross-platform pretensions.
• A general-purpose Wayland client library. Use wayland-client or smithay-client-toolkit if you need raw protocol access.
• A clipboard library. Use hjkl-clipboard — it implements wl_data_device over a raw Wayland socket (orthogonal to wayr's own connection) and supports text + HTML + RTF + image/png MIME types, plus X11 / macOS / Windows / OSC52 fallback for free.
• Production-ready. Pre-alpha. Track umbrella issue #1 for the v0.1 MVP plan.

Not to be confused with

There is an unrelated wayrs-* family of crates on crates.io (wayrs-client, wayrs-protocols, etc.) — an alternative low-level Wayland client implementation by a different author. wayr is not part of that family. The name similarity is coincidence; pin wayr (no trailing s-) in your Cargo.toml.

Status

v0.1 — Wayland-first protocol surface complete (toplevel, layer-shell, subsurface, fractional-scale + multi-output, text-input-v3 IME, cursor-shape, high-res scroll, touch). Pre-alpha; API is locked in shape but consumers (buffr, pikr) are still validating in tree. See umbrella issue #1 for the long-term plan and issues for the live punch list.

Why not fork winit?

Considered and rejected. Winit's cross-platform surface drags in X11, macOS, Windows, and mobile code paths kryptic-sh apps will never use. Floem's floem-winit fork was the leading candidate base, but its restructured per-platform subcrate layout and calver versioning make rebases against upstream painful. Building a focused Wayland-only toolkit from scratch — ~4-6k LOC of glue on top of wayland-client + wayland-protocols + xkbcommon — is comparable scope to a fork maintained over a year, with the bonus of an API shaped around the embedder use cases (subsurface, layer-shell) we actually need.

License

MIT. See LICENSE.