Make locale failures and discovery first-class

.github/workflows/ci.yml

@@ -34,6 +34,8 @@ jobs:
           sudo apt-get install -y \
             libhunspell-dev \
             hunspell-en-us \
+            hunspell-de-de \
+            hunspell-fr \
             libclang-dev
 
       - name: Test
@@ -46,10 +48,11 @@ jobs:
       - uses: actions/checkout@v4
       - uses: dtolnay/rust-toolchain@stable
         with:
-          components: clippy
+          components: rustfmt, clippy
       - uses: Swatinem/rust-cache@v2
       - name: Install hunspell
         run: |
           sudo apt-get update
           sudo apt-get install -y libhunspell-dev hunspell-en-us libclang-dev
+      - run: cargo fmt --all -- --check
       - run: cargo clippy --all-targets -- -D warnings

CHANGELOG.md

@@ -6,6 +6,12 @@ This project follows [Semantic Versioning](https://semver.org/).
 
 ## [Unreleased]
 
+### Breaking
+- `Error` is now `InvalidLocale` / `UnsupportedLocale` / `DictionaryNotFound` / `InitializationFailed` (removed `Unavailable`)
+- macOS `with_locale` returns `UnsupportedLocale` if the language is not installed
+- Added `Checker::locale` and `Checker::available_locales`
+- Removed the crate binary (`src/main.rs`); use `examples/`
+
 ### Changed
 - Document platform defaults for `Checker::new()`, locale failure, suggestions cap, and UTF-8 error ranges in rustdoc
 - README now states that macOS `Checker::new()` uses the system language

Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "spellkit"
-description = "Bindings to your friendly neighborhood spellchecker."
-keywords = ["spellcheck", "spellchecker", "hunspell"]
+description = "Cross-platform native spell checking for Rust (NSSpellChecker, Windows Spell Checker, Hunspell)."
+keywords = ["spellcheck", "spellchecker", "hunspell", "nsspellchecker"]
 categories = ["os", "text-processing"]
 documentation = "https://docs.rs/spellkit"
 repository = "https://github.com/rtmongold/spellkit"

README.md

@@ -1,35 +1,47 @@
-# spellkit
+# Spellkit
+
+Cross-platform native spell checking for Rust.
 
 [![On crates.io](https://img.shields.io/crates/v/spellkit.svg)](https://crates.io/crates/spellkit)
 ![Downloads](https://img.shields.io/crates/d/spellkit?style=flat-square)
 [![CI](https://github.com/rtmongold/spellkit/actions/workflows/ci.yml/badge.svg)](https://github.com/rtmongold/spellkit/actions/workflows/ci.yml)
 [![Docs](https://docs.rs/spellkit/badge.svg)](https://docs.rs/spellkit)
 
-Native spell checking with a small Rust API.
+## Why Spellkit?
 
-This project is **based on** [euclio/spellbound](https://github.com/euclio/spellbound)
-(last upstream commit 2020).
+Use the spell-checking facilities already on the user's system.
 
-| Platform | API                |
-| -------- | ------------------ |
+| Platform | Backend |
+| -------- | ------- |
 | macOS    | [`NSSpellChecker`] |
-| Windows  | [`ISpellChecker`]  |
-| *nix     | [`hunspell`] |
+| Windows  | [`ISpellChecker`] (Windows Spell Checker) |
+| Linux / other Unix | [`Hunspell`] with system dictionaries |
 
 [`ISpellChecker`]: https://docs.microsoft.com/en-us/windows/desktop/api/spellcheck/nn-spellcheck-ispellchecker
 [`NSSpellChecker`]: https://developer.apple.com/documentation/appkit/nsspellchecker
-[`hunspell`]: https://hunspell.github.io/
+[`Hunspell`]: https://hunspell.github.io/
+
+Applications should not reimplement macOS, Windows, and Hunspell separately. Spellkit is one small API over those backends.
+
+This project is **based on** [euclio/spellbound](https://github.com/euclio/spellbound) (last upstream commit 2020).
+
+## What Spellkit is not
+
+Spellkit does **not** bundle dictionaries or implement its own spelling algorithm. It wraps the platform backend and uses system / installed dictionaries. Behavior can differ across operating systems where the APIs differ.
 
-## Example
+That is the distinction from crates that ship an engine and word lists (for example Spellbook).
+
+## Quick start
+
+    cargo add spellkit
 
 ```rust
 use spellkit::Checker;
 
 fn main() -> Result<(), spellkit::Error> {
     let checker = Checker::new()?;
-    // Or: Checker::with_locale("en-US")?;
 
-    for err in checker.check("I beleeve I can fly") {
+    for err in checker.check("I havv a spelling error.") {
         println!("{} @ {}..{}", err.text(), err.start(), err.end());
         for suggestion in checker.suggest(err.text()) {
             println!("  → {suggestion}");
@@ -39,30 +51,97 @@ fn main() -> Result<(), spellkit::Error> {
 }
 ```
 
-`Checker::new()` uses a platform default: system language on macOS, the user locale on Windows (falling back to `en-US`), and `LC_ALL` / `LC_MESSAGES` / `LANG` on Linux when a dictionary exists (otherwise `en_US` / `en_GB`). Use `with_locale` for another language.
+`Checker::locale()` is the language this instance is using. `Checker::available_locales()` lists what the OS can check.
 
-Unknown or unsupported locales behave differently by platform:
+```rust
+use spellkit::Checker;
 
-- **Linux:** missing dictionary / unknown locale → `Error::Unavailable`
-- **macOS:** empty locale → `Error::Unavailable`; unknown tags may still create a checker (system fallback)
-- **Windows:** unsupported language tag → `Error::Unavailable`
+fn main() -> Result<(), spellkit::Error> {
+    println!("available: {:?}", Checker::available_locales());
+    let checker = Checker::new()?;
+    println!("using: {}", checker.locale());
+    Ok(())
+}
+```
 
-## Threading
+## Features
 
-`Checker` is not `Send` or `Sync`. Do not share it across threads. macOS also serializes access to the shared `NSSpellChecker`.
+- Cross-platform: macOS, Windows, Linux
+- System dictionaries (no files shipped in the crate)
+- Suggestions (up to 10)
+- Locale via `with_locale` (`en_US` and `en-US` both work)
+- Temporary ignored words (`ignore` is per checker, not global)
+- UTF-8 byte ranges (`start` / `end` / `range`)
+- Small API
+
+## Platform support
+
+| | Linux | macOS | Windows |
+| --- | --- | --- | --- |
+| Backend | Hunspell | NSSpellChecker | ISpellChecker |
+| `Checker::new()` | `LC_ALL` / `LC_MESSAGES` / `LANG` if a dict exists, else `en_US` / `en_GB` | system language | user locale, else `en-US` |
+| Unknown `with_locale` | `Error::DictionaryNotFound` (paths searched) | `Error::UnsupportedLocale` | `Error::UnsupportedLocale` |
+| Empty locale | `Error::InvalidLocale` | `Error::InvalidLocale` | `Error::InvalidLocale` |
+| Suggestions | yes | yes | yes |
+| `ignore` | yes (this handle only) | yes (this document tag only) | yes (this checker only) |
+| `available_locales` | `*.dic` stems on disk (`DICPATH` then system dirs) | `availableLanguages` | `SupportedLanguages` |
+| `Send` / `Sync` | no | no | no |
+
+Linux also honors `DICPATH` (colon-separated directories) before `/usr/share/hunspell` and the other built-in paths.
 
-## Linux
+Word breaks are **not** identical: Linux tokenizes alphanumeric / `'` runs; macOS and Windows use the OS checker.
 
-Needs a hunspell dictionary on disk (default search includes `/usr/share/hunspell`). Example:
+## How it works
+
+`Checker` is a thin wrapper. On each OS it calls the native API, then converts misspelling ranges to UTF-8 byte offsets into the original `&str`.
+
+## Spellkit vs other approaches
+
+**Why not Spellbook?** Use Spellbook when you want a portable engine and bundled (or app-shipped) dictionaries. Use Spellkit when you want the OS dictionaries, native suggestions, and minimal integration.
+
+**Why not Hunspell directly?** You would own dictionary discovery, FFI, and a second implementation for macOS and Windows. Spellkit is that integration.
+
+**Why not ispell?** You would own an external process, its lifetime, and the command protocol. Spellkit stays in-process.
+
+## Errors
+
+- empty locale → `Error::InvalidLocale`
+- Linux missing `.aff`/`.dic` → `Error::DictionaryNotFound` (includes search paths)
+- macOS / Windows language not installed → `Error::UnsupportedLocale`
+- backend failed to start (null Hunspell handle, COM factory, empty macOS language) → `Error::InitializationFailed`
+
+## Linux packages
 
 - Arch: `pacman -S hunspell hunspell-en_us`
 - Debian/Ubuntu: `apt install libhunspell-dev hunspell-en-us`
+- Extra languages used in CI: `hunspell-de-de`, `hunspell-fr`
+
+Without a dictionary, `Checker::new()` returns `Error::DictionaryNotFound`.
+
+## Examples
 
-Without a dictionary, `Checker::new()` returns `Error::Unavailable`.
+    cargo run --example check -- "I havv a spelling error."
+    cargo run --example suggestions -- "I beleeve I can fly"
+    cargo run --example locale
+    cargo run --example highlight
+
+## Threading
+
+`Checker` is not `Send` or `Sync`. Do not share it across threads. macOS also serializes access to the shared `NSSpellChecker`.
+
+## Documentation
+
+- [docs.rs/spellkit](https://docs.rs/spellkit)
+- [CHANGELOG.md](CHANGELOG.md)
+
+## Contributing
+
+Issues and PRs: [github.com/rtmongold/spellkit](https://github.com/rtmongold/spellkit)
 
 ## License
+
 MIT OR Apache-2.0
 
 ## Credits
 
-Originally by [Andy Russell](https://github.com/euclio). Maintained as `spellkit` by Robert Mongold.
+Originally by [Andy Russell](https://github.com/euclio). Maintained as `spellkit` by Robert Mongold.

examples/check.rs

@@ -0,0 +1,11 @@
+use spellkit::Checker;
+use std::env;
+
+fn main() -> Result<(), spellkit::Error> {
+    let text = env::args().skip(1).collect::<Vec<_>>().join(" ");
+    let checker = Checker::new()?;
+    for error in checker.check(&text) {
+        println!("{}", error.text());
+    }
+    Ok(())
+}

examples/highlight.rs

@@ -0,0 +1,12 @@
+use spellkit::Checker;
+
+fn main() -> Result<(), spellkit::Error> {
+    let text = "I beleeve I can fly";
+    let checker = Checker::new()?;
+    for err in checker.check(text) {
+        let range = err.range();
+        println!("{} @ {range:?}", err.text());
+        println!("  slice: {}", &text[range]);
+    }
+    Ok(())
+}

examples/locale.rs

@@ -0,0 +1,6 @@
+fn main() -> Result<(), spellkit::Error> {
+    println!("available: {:?}", spellkit::Checker::available_locales());
+    let c = spellkit::Checker::new()?;
+    println!("locale: {}", c.locale());
+    Ok(())
+}

examples/suggestions.rs

@@ -0,0 +1,14 @@
+use spellkit::Checker;
+use std::env;
+
+fn main() -> Result<(), spellkit::Error> {
+    let text = env::args().skip(1).collect::<Vec<_>>().join(" ");
+    let checker = Checker::new()?;
+    for error in checker.check(&text) {
+        println!("{}", error.text());
+        for s in checker.suggest(error.text()) {
+            println!("  {s}");
+        }
+    }
+    Ok(())
+}