You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add support for integrated suggest-as-you-type suggestions.
This patch aims to remove the govarnam dependency for search suggestions.
- Introduces a new trie-based in-memory word search mechanism which
is populated using all words in the DB during load. Configurable with
the `[suggestions]` config block. This works for all languages that
are entered in the input box in their native Unicode forms.
If no suggestions are found, it falls back to fulltext search on the DB.
- For languages (eg: Malayalam, Kannada) where the input can be Latin
characters denoting phonetic strings (eg: "amma"), search + suggestions
still work thanks to the new transliteration algorithm shipped in the
bundled luka tokenizers. This works for any language. As long as
tokenize() in the Lua tokenizers return tokens, whatever the input maybe.
So, this patch introduces transliteration mapping for generating IndicPhone
hashes for "Manglish" and "Kanglish" latin char phonetic inputs.
0 commit comments