🪨 diskrune

Fast parallel disk space analyzer for Windows, macOS and Linux. Built with Rust + Tauri 2, powered by Rayon work-stealing parallelism.

---

Features

• Ultra-fast scanning — parallel recursive traversal across all CPU cores via Rayon
• Live scan progress — animated shimmer bar + real-time directory counter while scanning
• Dynamic drive detection — drive buttons populated automatically at startup (Windows: A–Z probe; macOS: /Volumes; Linux: /, /mnt, /media)
• Interactive tree view — expandable directory tree with size bars and percentages
• Top 15 biggest directories — instant ranking with colour-coded usage bars
• Drill-down navigation — click ⤵ on any folder to re-scan it as the new root, with breadcrumb history to navigate back
• Full-text search — find any folder by name or path in real time
• Folder picker — browse to any directory with the native OS dialog (📂)
• Bounded IPC payload — tree is trimmed to 5 levels before sending to the UI (stats preserved at every cut-off), preventing OOM on large drives like C:\
• Syscall-efficient — uses file_type() from readdir cache instead of a separate stat() per entry
• Symlink-safe — symlinks are skipped to avoid infinite loops and double-counting
• Single-pass stats — size, directory count and file count computed in one fold

---

Screenshots

Coming soon — run locally with cargo tauri dev

---

Build

Prerequisites

• Rust 1.77+
• Tauri v2 CLI: cargo install tauri-cli --version "^2"
• Windows — WebView2 (included with Windows 11 / available as redistributable)
• Linux — GTK + WebKit system libraries:

  sudo apt install libgtk-3-dev libwebkit2gtk-4.1-dev \
    libappindicator3-dev librsvg2-dev patchelf

Development

git clone https://github.com/cyberlife-coder/diskrune.git
cd diskrune
cargo tauri dev

Release build

cargo tauri build

The installer is generated in src-tauri/target/release/bundle/.

---

Architecture

src-tauri/src/
├── main.rs       Entry point (hides console window in release)
├── lib.rs        Tauri builder — registers commands and plugins
├── commands.rs   Tauri commands: start_scan, open_path, list_drives
├── scanner.rs    Parallel recursive scanner (Rayon) + AtomicU64 progress counter
└── node.rs       DirNode model, aggregate(), sort_by_size_desc(), trim_to_depth()

Performance design

Technique	Impact
rayon::par_iter() recursive descent	All CPU cores used automatically
file_type() instead of metadata() for dirs	~50% fewer syscalls on dir-heavy trees
Single-pass aggregate() fold	3× fewer iterations over children
AtomicU64 progress counter	Lock-free dir count across rayon threads
trim_to_depth(node, 5) before IPC emit	Prevents OOM on 700 k+ node trees
sort_unstable_by	Fastest in-place sort, no extra allocation
Release: lto=true, codegen-units=1, opt-level=3	Maximum compiler optimisation

IPC payload strategy

diskrune sends only the first 5 levels of the directory tree to the frontend. Aggregated stats (total size, dir count, file count) are preserved at every trimmed node so all numbers remain accurate. Clicking ⤵ on any directory triggers a fresh start_scan on that path, loading its subtree on demand.

---

Ecosystem

diskrune is part of the Veles ecosystem by Wiscale France.

---

License

MIT