Yorùbá Code Spell Checker

Adds a Yorùbá dictionary to Code Spell Checker so it stops flagging Yorùbá words in comments, strings, and identifiers as misspelled. Both toned (yorùbá) and plain-ASCII (yoruba) spellings are accepted, regardless of which form you actually type.

How this works

This extension depends on Code Spell Checker (extensionDependencies: ["streetsidesoftware.code-spell-checker"]). src/extension.js activates on startup, resolves an absolute path to the bundled cspell-ext.json via context.asAbsolutePath(), and registers it with Code Spell Checker's exported registerConfig(path) API. cspell-ext.json in turn declares dictionaryDefinitions/dictionaries pointing at dictionaries/yoruba.txt.

This replaced an earlier v1 approach that tried to register the dictionary via a static configurationDefaults.cSpell.customDictionaries entry with a relative path. That doesn't work: Code Spell Checker's path resolver only recognizes ./~/${workspaceFolder}/${root} as leading tokens, all of which resolve against the user's workspace folder, never the contributing extension's own install directory - confirmed by reading Code Spell Checker's source directly (WorkspacePathResolver.mts), and by comparing against streetsidesoftware/vscode-cspell-dict-extensions, the official reference implementation every real published dictionary extension (French, Medical Terms, etc.) is built from. There's no static-JSON way to reference "this extension's own directory" - context.asAbsolutePath() only exists at runtime, which is why this needs real extension-host code, not just a different path string.

The dictionary

dictionaries/yoruba.txt (~7,300 words as of DICT-1) is generated by npm run generate:wordlist, which runs scripts/build-real-wordlist.mjs. It combines the ~400 snippet-vocabulary words with a Yorùbá stopword list and translated sentiment lexicon sourced from NaijaSenti (CC-BY 4.0, hausanlp/NaijaSenti - vendored in scripts/corpus-sources/naijasenti/, attribution and full sourcing notes in docs/DATA_SOURCES.md). Both the as-sourced spelling and its tone-mark-stripped ascii form are included for every word.

Unlike data/vocab-snippets.csv's hand-curated 220 entries, these corpus-derived words don't carry individual confirmed/draft confidence tags - there are thousands of them, not twenty. Native-speaker review is still valuable here (see VOCAB-1 in docs/BACKLOG.md and CONTRIBUTING.md), just at the level of spot-checking and flagging bad entries rather than reviewing each one individually before anything ships.

Try it locally

1. Make sure Code Spell Checker is installed in your VS Code.
2. Open this folder (packages/yoruba-code-spell-checker) in VS Code.
3. Press F5 to launch an Extension Development Host.
4. Open a file and type a Yorùbá word from the dictionary in a comment - confirm the red squiggle doesn't appear. Try both a toned and a plain-ASCII spelling of the same word.
5. If it's still flagged: open the Output panel, "Code Spell Checker" channel, and check whether the yoruba dictionary shows up as loaded - that'll tell you whether registerConfig fired at all versus the dictionary loading but not covering that particular word.