@graffiticode/learnosity-cqt

The Learnosity custom question type runtime shared by Graffiticode languages.

Learnosity loads two bundles per question type — a browser question.js and a scorer.js that also runs server-side — and instantiates new Question(init, lrnUtils) and new Scorer(question, responseValue). This package supplies both classes; a language supplies only its renderer and its scoring functions.

Usage

import { createQuestion, createScorer } from "@graffiticode/learnosity-cqt";
import { Form, scoreCells, getCellsValidation } from "@graffiticode/l0166";

const defaultData = {
  validation:  { points: 0, ranges: {}, cells: {} },
  interaction: { type: "table", v: "0.0.1", cells: {}, columns: {} },
};

export const Question = createQuestion({ Form, scoreCells, getCellsValidation, defaultData });
export const Scorer   = createScorer({ scoreCells, defaultData });

defaultData is a parameter rather than a constant because the fallback envelope is language-specific — the shape above is the spreadsheet one.

Ship the stylesheet and the authoring layout alongside the bundles:

import "@graffiticode/learnosity-cqt/styles.css";

@graffiticode/learnosity-cqt/authoring-layout.html is the Question Editor layout: a 348-line postMessage bridge that opens the Graffiticode editor in a popup and writes the authored item back through editor.getJson() / setJson(). It is generic apart from a single __GC_LANG__ placeholder, which the consuming build substitutes with the four-digit language id.

Scoring contract

This package implements the cell-scoring contract: scoreCells({ cells, validation }) returns the cells with a score attached, and the question's total is the sum of score.points across cells. That yields partial credit.

Languages that predate scoreCells — l0152, l0153, l0154, l0155, l0157 — use an all-or-nothing contract instead, comparing the response against question.valid_response.value. That path is not implemented here yet. Adding it means a sibling pair of factories (createSimpleQuestion / createSimpleScorer) rather than branching inside these: the two differ in their state reducer, their suggested-answer derivation, and their scoring, so a shared implementation would be mostly if.

Notes

• Built with Babel, not the repo's tsconfig.base.json — this is browser JSX, and that config targets Node (lib: ["ES2022"], no DOM, no jsx). Output is ESM so consumers can keep the usual exclude: /node_modules/ on their own babel-loader.
• react and react-dom are peer dependencies.