asxelot/fastdict.koplugin

FastDict is a plugin for KOReader, the open source e-reader software many Kindle owners install to replace the stock reading app. Its job is to make tap-on-word dictionary lookups faster. Normally, KOReader looks up definitions by launching a separate program called sdcv every time you tap a word, and that program rereads the entire dictionary file from scratch on each launch. With large dictionaries, such as a 46 megabyte file holding 1.7 million word forms, that rescan can take several seconds, which feels slow on an e-reader. FastDict fixes this by handling exact word lookups itself, inside KOReader's own process, instead of starting a new program each time. It builds small helper files, called cache files, once per dictionary, and then uses those caches to jump straight to the right entry with a fast search method rather than scanning the whole file again. When a dictionary or query type is not supported by this faster method, or if something goes wrong, FastDict automatically falls back to the original sdcv approach, so lookups should never fail, only sometimes run at the older speed. The plugin also tries a few simple English word shape guesses when an exact match is not found, such as turning streets into street or walked into walk, so common inflected words still resolve quickly. It only accepts a guess if it matches a real dictionary entry, so it will not invent answers, though irregular words like went still rely on the slower fallback search. Installing FastDict on a Kindle means copying a handful of Lua files into KOReader's plugins folder and restarting the app. There is also an optional step to pre build the lookup caches right away instead of waiting for the first tap. The project includes a companion Python tool for merging dictionary index files offline, plus test scripts to check that results match the original sdcv output exactly and that the plugin's behavior is correct on a desktop computer before using it on a device.