Skip to main content

Overview

There is no KeyLayout/schema type and no ready-to-use keyboard component in this library — a keyboard is a hand-written View/Text/TouchableOpacity component you write yourself, the same file running live on both Android and iOS (real React Native, Hermes + JSI + Fabric). Beyond the bridge/settings functions, this package also supplies the Key/KeyRow data shape used by the gesture-matching feature, word prediction, and iOS's setup CLI.

Key / KeyRow / KeyAction​

interface Key {
  id: string;             // unique within the row, e.g. "q", "shift"
  label: string;          // rendered text
  width: number;          // multiples of one standard key
  action: KeyAction;
  value?: string;         // committed text for insertChar/commitText; falls back to label
  secondaryLabel?: string;
}

interface KeyRow {
  keys: Key[];
}

type KeyAction =
  | 'insertChar' | 'commitText' | 'deleteBackward'
  | 'toggleShift' | 'toggleSymbols' | 'switchLanguage' | 'space' | 'enter';

This is data modeling for the gesture-matching feature below (keysAlongPath returns Key[]) — it isn't rendered by anything and doesn't restrict script or language: label/value are plain Unicode strings, so a Key can represent any language/script (Arabic, Devanagari, CJK, Cyrillic, Bangla, etc.) the same way.

Geometry​

computeGeometry(rows, rowHeight = 1): KeyGeometry[]​

Lays out a set of rows on a grid (row-major, left-to-right) in "standard key" units. Pure/ deterministic. Used by the gesture matcher below.

interface KeyGeometry { key: Key; x0: number; x1: number; y0: number; y1: number; }

Gestures​

keysAlongPath(rows, path, rowHeight = 1): Key[]​

Given a set of rows and a raw touch path (TouchPoint[], same units as geometry), returns the ordered, de-duplicated sequence of keys the path passed over. Points outside every key are ignored. Pure function — no I/O.

interface TouchPoint { x: number; y: number; }

Word prediction​

WordPredictor (interface)​

interface WordPredictor {
  suggest(prefix: string, options?: SuggestOptions): string[];
}
interface SuggestOptions { limit?: number; } // default 3

Pluggable — apps may supply frequency-weighted or ML-backed implementations.

StaticDictionaryPredictor​

Minimal WordPredictor backed by an in-memory word list. Case-insensitive prefix match, ranked shortest-then-alphabetical.