fix: Improve rendering of TOC panel

The previous code could not correctly handle nested XML elements in the TOC text entries.

README.md

@@ -0,0 +1,36 @@
+## High-level architecture
+
+```mermaid
+flowchart TD
+    Cellar[Cellar API]
+    Frontend[React Frontend]
+
+    subgraph "FastAPI backend"
+        client[Cellar Client]
+        conv[Formex -> HTML renderer]
+        FastAPI[REST API]
+    end
+
+    Cellar --> meta@{ shape: docs, label: "XML/RDF Metadata"}
+    Cellar --> docs@{ shape: docs, label: "Formex 4 content"}
+
+    meta --> client
+    docs --> client
+
+    client -- Formex --> FastAPI
+    client -- Formex --> conv
+
+    conv -- HTML --> FastAPI
+
+    FastAPI -- TOC, HTML article text --> Frontend
+```
+
+## Resources
+
+- Formex 4
+  - [XML schema](https://op.europa.eu/documents/3938058/5910419/formex_manual_on_screen_version.html/)
+  - [Physical specifications](https://op.europa.eu/documents/3938058/5896514/physical-specifications/)
+- [Cellar](https://op.europa.eu/en/web/cellar/home)
+  - [Publications API](https://op.europa.eu/en/web/cellar/cellar-data/publications)
+  - [Metadata REST API](https://op.europa.eu/en/web/cellar/cellar-data/metadata/metadata-notices)
+  - [Metadata SPARQL API](https://op.europa.eu/en/web/cellar/cellar-data/metadata/knowledge-graph)

frontend/__mocks__/zustand.ts

@@ -0,0 +1,64 @@
+import { act } from "@testing-library/react";
+import type * as ZustandExportedTypes from "zustand";
+export * from "zustand";
+
+const { create: actualCreate, createStore: actualCreateStore } =
+  jest.requireActual<typeof ZustandExportedTypes>("zustand");
+
+// a variable to hold reset functions for all stores declared in the app
+export const storeResetFns = new Set<() => void>();
+
+const createUncurried = <T>(
+  stateCreator: ZustandExportedTypes.StateCreator<T>
+) => {
+  const store = actualCreate(stateCreator);
+  const initialState = store.getInitialState();
+  storeResetFns.add(() => {
+    store.setState(initialState, true);
+  });
+  return store;
+};
+
+// when creating a store, we get its initial state, create a reset function and add it in the set
+export const create = (<T>(
+  stateCreator: ZustandExportedTypes.StateCreator<T>
+) => {
+  console.log("zustand create mock");
+
+  // to support curried version of create
+  return typeof stateCreator === "function"
+    ? createUncurried(stateCreator)
+    : createUncurried;
+}) as typeof ZustandExportedTypes.create;
+
+const createStoreUncurried = <T>(
+  stateCreator: ZustandExportedTypes.StateCreator<T>
+) => {
+  const store = actualCreateStore(stateCreator);
+  const initialState = store.getInitialState();
+  storeResetFns.add(() => {
+    store.setState(initialState, true);
+  });
+  return store;
+};
+
+// when creating a store, we get its initial state, create a reset function and add it in the set
+export const createStore = (<T>(
+  stateCreator: ZustandExportedTypes.StateCreator<T>
+) => {
+  console.log("zustand createStore mock");
+
+  // to support curried version of createStore
+  return typeof stateCreator === "function"
+    ? createStoreUncurried(stateCreator)
+    : createStoreUncurried;
+}) as typeof ZustandExportedTypes.createStore;
+
+// reset all stores after each test run
+afterEach(() => {
+  act(() => {
+    storeResetFns.forEach((resetFn) => {
+      resetFn();
+    });
+  });
+});

frontend/eslint.config.js

@@ -1,28 +1,33 @@
-import js from '@eslint/js'
+import js from "@eslint/js";
-import globals from 'globals'
+import pluginQuery from "@tanstack/eslint-plugin-query";
-import reactHooks from 'eslint-plugin-react-hooks'
+import reactHooks from "eslint-plugin-react-hooks";
-import reactRefresh from 'eslint-plugin-react-refresh'
+import reactRefresh from "eslint-plugin-react-refresh";
-import tseslint from 'typescript-eslint'
+import globals from "globals";
+import tseslint from "typescript-eslint";
 
 export default tseslint.config(
-  { ignores: ['dist'] },
+  { ignores: ["dist"] },
   {
-    extends: [js.configs.recommended, ...tseslint.configs.recommended],
+    extends: [
-    files: ['**/*.{ts,tsx}'],
+      js.configs.recommended,
+      ...tseslint.configs.recommended,
+      ...pluginQuery.configs["flat/recommended"],
+    ],
+    files: ["**/*.{ts,tsx}"],
     languageOptions: {
       ecmaVersion: 2020,
       globals: globals.browser,
     },
     plugins: {
-      'react-hooks': reactHooks,
+      "react-hooks": reactHooks,
-      'react-refresh': reactRefresh,
+      "react-refresh": reactRefresh,
     },
     rules: {
       ...reactHooks.configs.recommended.rules,
-      'react-refresh/only-export-components': [
+      "react-refresh/only-export-components": [
-        'warn',
+        "warn",
         { allowConstantExport: true },
       ],
     },
-  },
+  }
-)
+);

frontend/jest.config.js

@@ -0,0 +1,12 @@
+/** @type {import('ts-jest').JestConfigWithTsJest} **/
+export default {
+  preset: "ts-jest",
+  testEnvironment: "jsdom",
+  transform: {
+    "^.+\\.tsx?$": ["ts-jest", { tsconfig: "tsconfig.app.json" }],
+  },
+  moduleNameMapper: {
+    "\\.(css|less|scss|sss|styl)$": "<rootDir>/node_modules/jest-css-modules",
+  },
+  setupFilesAfterEnv: ["<rootDir>/jest.setup.ts"],
+};

frontend/jest.setup.ts

@@ -0,0 +1 @@
+import "@testing-library/jest-dom";

frontend/package.json

@@ -7,11 +7,14 @@
     "dev": "vite",
     "build": "tsc -b && vite build",
     "lint": "eslint .",
-    "preview": "vite preview"
+    "preview": "vite preview",
+    "test": "jest"
   },
   "dependencies": {
+    "@floating-ui/react": "^0.27.8",
     "@tanstack/react-query": "^5.74.4",
     "@tanstack/react-query-devtools": "^5.74.6",
+    "axios": "^1.9.0",
     "react": "^19.0.0",
     "react-dom": "^19.0.0",
     "react-router-dom": "^7.5.1",
@@ -19,14 +22,25 @@
   },
   "devDependencies": {
     "@eslint/js": "^9.22.0",
+    "@jest/globals": "^29.7.0",
     "@tanstack/eslint-plugin-query": "^5.73.3",
-    "@types/react": "^19.0.10",
+    "@testing-library/dom": "^10.4.0",
-    "@types/react-dom": "^19.0.4",
+    "@testing-library/jest-dom": "^6.6.3",
+    "@testing-library/react": "^16.3.0",
+    "@testing-library/user-event": "^14.6.1",
+    "@types/jest": "^29.5.14",
+    "@types/react": "^19.1.2",
+    "@types/react-dom": "^19.1.2",
     "@vitejs/plugin-react": "^4.3.4",
     "eslint": "^9.22.0",
     "eslint-plugin-react-hooks": "^5.2.0",
     "eslint-plugin-react-refresh": "^0.4.19",
     "globals": "^16.0.0",
+    "jest": "^29.7.0",
+    "jest-css-modules": "^2.1.0",
+    "jest-environment-jsdom": "^29.7.0",
+    "ts-jest": "^29.3.2",
+    "ts-node": "^10.9.2",
     "typescript": "~5.7.2",
     "typescript-eslint": "^8.26.1",
     "typescript-plugin-css-modules": "^5.1.0",

frontend/src/App.tsx

@@ -1,64 +1,42 @@
-import { useQueries } from "@tanstack/react-query";
-
-import { getArticleIds, getToc } from "./lib/api";
 import { Language } from "./lib/types";
 
 import ArticleSelector from "./components/ArticleSelector/ArticleSelector";
 import Panel from "./components/Panel/Panel";
 import TOC from "./components/TOC/TOC";
 
-import useNavState from "./store/navStore";
 import useUIStore from "./store/uiStore";
 
 import styles from "./App.module.css";
 import CelexSelector from "./components/CelexSelector/CelexSelector";
+import { useTOC } from "./hooks/toc";
 
 function App() {
-  const { numPanels, addPanel } = useUIStore();
+  const numPanels = useUIStore((state) => state.numPanels);
-  const { celexId, articleId } = useNavState();
+  const addPanel = useUIStore((state) => state.addPanel);
-
+  const { data: toc, isPending, error } = useTOC();
-  const results = useQueries({
-    queries: [
-      {
-        queryKey: ["articleIds", celexId],
-        queryFn: () => getArticleIds(celexId!),
-        enabled: !!celexId,
-      },
-      {
-        queryKey: ["toc", celexId],
-        queryFn: () => getToc(celexId!, Language.ENG),
-        enabled: !!celexId,
-      },
-    ],
-  });
-
-  const isPending = results.some((result) => result.isPending);
-  const error = results.find((result) => result.isError);
 
   if (isPending) {
     return <div>Loading...</div>;
   }
   if (error) {
-    return <div>Error: {error.error?.message}</div>;
+    return <div>Error: {error.message}</div>;
   }
 
   return (
     <div className={styles.App}>
       <div className={styles.controls}>
         <CelexSelector />
-        <ArticleSelector articleIds={results[0].data!} />
+        <ArticleSelector toc={toc} />
         <button onClick={addPanel}>Add Panel</button>
       </div>
       <div className={styles.panelContainer}>
-        <TOC toc={results[1].data!} />
+        <TOC toc={toc} />
         {Array.from({ length: numPanels }, (_, index) => (
           <Panel
             key={index}
-            celexId={celexId!}
             language={
               Object.values(Language)[index % Object.values(Language).length]
             }
-            articleId={articleId!}
           />
         ))}
       </div>

frontend/src/components/ArticleSelector/ArticleSelector.module.css

@@ -0,0 +1,4 @@
+.articleSelector {
+  width: 40ch;
+  text-overflow: ellipsis;
+}

frontend/src/components/ArticleSelector/ArticleSelector.test.tsx

@@ -0,0 +1,60 @@
+import { render } from "@testing-library/react";
+import type { Division } from "../../lib/types";
+import ArticleSelector from "./ArticleSelector";
+
+describe("ArticleSelector", () => {
+  const mockDivision: Division = {
+    type: "division",
+    title: "Chapter 1",
+    subtitle: "Introduction",
+    level: 0,
+    content: [
+      {
+        type: "division",
+        title: "Section 1.1",
+        subtitle: "Overview",
+        level: 2,
+        content: [
+          {
+            type: "article",
+            id: 1,
+            title: "Article 1",
+            subtitle: "Details",
+          },
+          {
+            type: "article",
+            id: 2,
+            title: "Article 2",
+            subtitle: "Summary",
+          },
+        ],
+      },
+    ],
+  };
+
+  test("renders a top-level division as an optgroup", () => {
+    const { getByRole } = render(<ArticleSelector toc={[mockDivision]} />);
+
+    const optgroup = getByRole("group", {
+      name: "Chapter 1: Introduction",
+    });
+    expect(optgroup).toBeInTheDocument();
+  });
+
+  test("renders nested divisions as options", () => {
+    const { getByText } = render(<ArticleSelector toc={[mockDivision]} />);
+
+    const option1 = getByText("Article 1: Details");
+    const option2 = getByText("Article 2: Summary");
+
+    expect(option1).toBeInTheDocument();
+    expect(option2).toBeInTheDocument();
+  });
+
+  test("flattens nested divisions correctly", () => {
+    const { container } = render(<ArticleSelector toc={[mockDivision]} />);
+
+    const options = container.querySelectorAll("option");
+    expect(options).toHaveLength(2); // Two options rendered
+  });
+});

frontend/src/components/ArticleSelector/ArticleSelector.tsx

@@ -1,33 +1,57 @@
-import useNavState from "../../store/navStore";
+import { Fragment, JSX } from "react";
+import type { Division } from "../../lib/types";
+import useNavStore from "../../store/navStore";
+import styles from "./ArticleSelector.module.css";
 
 type ArticleSelectorProps = {
-  articleIds: number[];
+  toc: Division[];
 };
 
-function ArticleSelector({ articleIds }: ArticleSelectorProps) {
+/**
-  const { articleId, setArticleId } = useNavState();
+ * Renders a division and its contents as a nested structure of optgroups and options.
+ * @param {Division} div - The division to render.
+ * @returns {JSX.Element} The rendered division, suitable for use inside a `select` tag.
+ */
+function renderDivision(div: Division): JSX.Element {
+  const contents = div.content.map((c) => {
+    if (c.type === "division") {
+      return renderDivision(c);
+    } else {
+      const title = `${c.title}: ${c.subtitle}`;
+      return (
+        <option key={title} value={c.id}>
+          {title}
+        </option>
+      );
+    }
+  });
+
+  const title = `${div.title}: ${div.subtitle}`;
+  if (div.level === 0) {
+    return (
+      // For top-level divisions, we can use optgroup
+      <optgroup key={title} label={title}>
+        {contents}
+      </optgroup>
+    );
+  } else {
+    // HTML does not support nested optgroups, so we need to flatten the structure
+    return <Fragment key={title}>{contents}</Fragment>;
+  }
+}
+
+function ArticleSelector({ toc }: ArticleSelectorProps) {
+  const articleId = useNavStore.use.articleId();
+  const setArticleId = useNavStore.use.setArticleId();
+
   return (
-    <>
+    <select
-      {articleId && articleId > 1 && (
+      value={articleId || undefined}
-        <button onClick={() => setArticleId(articleId - 1)}>prev</button>
+      className={styles.articleSelector}
-      )}
+      onChange={(e) => setArticleId(parseInt(e.target.value))}
-      <select
+    >
-        value={articleId || undefined}
+      {toc.map(renderDivision)}
-        onChange={(e) => {
+    </select>
-          const id = parseInt(e.currentTarget.value);
-          setArticleId(id);
-        }}
-      >
-        {articleIds.map((id) => (
-          <option key={id} value={id}>
-            Article {id}
-          </option>
-        ))}
-      </select>
-      {articleId && articleId < articleIds[articleIds.length - 1] && (
-        <button onClick={() => setArticleId(articleId + 1)}>next</button>
-      )}
-    </>
   );
 }
 

frontend/src/components/CelexSelector/CelexSelector.test.tsx

@@ -0,0 +1,40 @@
+import { fireEvent, render } from "@testing-library/react";
+import { examples } from "../../lib/examples";
+import useNavStore from "../../store/navStore";
+import CelexSelector from "./CelexSelector";
+
+describe("CelexSelector", () => {
+  it("renders the dropdown with options", () => {
+    const { getByLabelText, getAllByRole, getByRole } = render(
+      <CelexSelector />
+    );
+
+    expect(getByLabelText("Select example:")).toBeInTheDocument();
+    expect(getByRole("combobox")).toBeInTheDocument();
+
+    const [def, ...options] = getAllByRole("option");
+
+    // First option is the disabled placeholder option
+    expect(def).toHaveValue("");
+    expect(def).toHaveTextContent("Select an example");
+    expect(def).toBeDisabled();
+
+    expect(options).toHaveLength(examples.length);
+    for (const i in examples) {
+      expect(options[i]).toHaveValue(examples[i].id);
+      expect(options[i]).toHaveTextContent(examples[i].name);
+    }
+  });
+
+  it("calls setCelexId and setArticleId on selection change", async () => {
+    const celexId = examples[2].id;
+    const { getByRole } = render(<CelexSelector />);
+
+    fireEvent.change(getByRole("combobox"), {
+      target: { value: celexId },
+    });
+
+    expect(useNavStore.getState().celexId).toEqual(celexId);
+    expect(useNavStore.getState().articleId).toEqual(1);
+  });
+});

frontend/src/components/CelexSelector/CelexSelector.tsx

@@ -1,13 +1,10 @@
-import useNavState from "../../store/navStore";
+import { examples } from "../../lib/examples";
-
+import useNavStore from "../../store/navStore";
-const examples = [
-  { name: "GDPR", id: "32016R0679" },
-  { name: "AI Act", id: "32024R1689" },
-  { name: "Cyber Resilience Act", id: "32024R2847" },
-];
 
 function CelexSelector() {
-  const { celexId, setCelexId, setArticleId } = useNavState();
+  const celexId = useNavStore.use.celexId();
+  const setCelexId = useNavStore.use.setCelexId();
+  const setArticleId = useNavStore.use.setArticleId();
 
   return (
     <div>
@@ -20,6 +17,9 @@ function CelexSelector() {
           setArticleId(1);
         }}
       >
+        <option value="" disabled>
+          Select an example
+        </option>
         {examples.map((example) => (
           <option key={example.id} value={example.id}>
             {example.name}

frontend/src/components/LanguageSwitcher/LanguageSwitcher.test.tsx

@@ -0,0 +1,44 @@
+import { render, screen } from "@testing-library/react";
+import userEvent from "@testing-library/user-event";
+import { Language } from "../../lib/types";
+import LanguageSwitcher from "./LanguageSwitcher";
+
+describe("LanguageSwitcher", () => {
+  const mockOnChange = jest.fn();
+  const renderSwitcher = (
+    onChange = mockOnChange,
+    defaultLang = Language.ENG
+  ) => {
+    render(<LanguageSwitcher defaultLang={defaultLang} onChange={onChange} />);
+  };
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+  });
+
+  test("renders with correct default language", async () => {
+    renderSwitcher();
+
+    expect(screen.getByRole("combobox")).toHaveValue(Language.ENG);
+  });
+
+  test("calls onChange handler when a new language is selected", async () => {
+    renderSwitcher();
+
+    await userEvent.selectOptions(screen.getByRole("combobox"), Language.ESP);
+
+    expect(mockOnChange).toHaveBeenCalledWith(Language.ESP);
+  });
+
+  test("renders all language options correctly", () => {
+    renderSwitcher();
+
+    const options = screen.getAllByRole("option");
+    const languageValues = Object.values(Language);
+    expect(options).toHaveLength(languageValues.length);
+    languageValues.forEach((lang, index) => {
+      expect(options[index]).toHaveValue(lang);
+      expect(options[index]).toHaveTextContent(lang.toUpperCase());
+    });
+  });
+});

frontend/src/components/LanguageSwitcher/LanguageSwitcher.tsx

@@ -10,6 +10,7 @@ function LanguageSwitcher({
 }) {
   return (
     <select
+      data-testid="language-switcher"
       defaultValue={defaultLang}
       onChange={(ev) => onChange(ev.currentTarget.value as Language)}
       className={styles.languageSwitcher}

frontend/src/components/Panel/Panel.module.css

@@ -1,5 +1,5 @@
 .panel {
-  flex: 1 auto;
+  flex: auto;
   padding: 1rem;
   border-radius: 8px;
   border: 1px solid #ccc;

frontend/src/components/Panel/Panel.test.tsx

@@ -0,0 +1,114 @@
+import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
+import { fireEvent, render } from "@testing-library/react";
+import React from "react";
+import { getArticle } from "../../lib/api";
+import { Language } from "../../lib/types";
+import useNavStore from "../../store/navStore";
+import useUIStore from "../../store/uiStore";
+import Panel from "./Panel";
+
+jest.mock("../../store/uiStore");
+jest.mock("../../store/navStore");
+jest.mock("../../lib/api");
+jest.mock("../../constants", () =>
+  Promise.resolve({
+    API_URL: "http://localhost:8000/api", // Mock the API_URL to a local server for testing
+  })
+);
+
+const queryClient = new QueryClient({
+  defaultOptions: {
+    queries: {
+      retry: false,
+    },
+  },
+});
+const wrapper = ({ children }: { children: React.ReactNode }) => (
+  <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+);
+
+describe("Panel Component", () => {
+  const mockSetSelectedParagraphId = jest.fn();
+  const mockUseUIStore = {
+    selectedParagraphId: null,
+    setSelectedParagraphId: mockSetSelectedParagraphId,
+  };
+  const mockNavState = {
+    celexId: "123",
+    articleId: 1,
+  };
+
+  beforeEach(() => {
+    jest.mocked(useNavStore).mockReturnValue(mockNavState);
+    jest.mocked(useUIStore).mockReturnValue(mockUseUIStore);
+  });
+
+  afterEach(() => {
+    jest.clearAllMocks();
+    queryClient.clear();
+  });
+
+  test("renders loading state", () => {
+    (getArticle as jest.Mock).mockReturnValue(new Promise(() => {}));
+    const { getByText } = render(<Panel />, { wrapper });
+    expect(getByText("Loading...")).toBeInTheDocument();
+  });
+
+  test("renders error state", async () => {
+    (getArticle as jest.Mock).mockRejectedValue(new Error("Failed to fetch"));
+    const { findByText } = render(<Panel />, { wrapper });
+    expect(
+      await findByText("An error has occurred: Failed to fetch")
+    ).toBeInTheDocument();
+  });
+
+  test("renders article content", async () => {
+    const mockData = `<div class='paragraph' data-paragraph-id='1'>Test Content</div>`;
+    jest.mocked(getArticle).mockResolvedValue(mockData);
+
+    const result = render(<Panel />, { wrapper });
+    expect(await result.findByText("Test Content")).toBeInTheDocument();
+  });
+
+  test("highlights a paragraph on click", async () => {
+    const mockData = `
+            <div class='paragraph' data-paragraph-id='1'>Paragraph 1</div>
+            <div class='paragraph' data-paragraph-id='2'>Paragraph 2</div>
+        `;
+    (getArticle as jest.Mock).mockResolvedValue(mockData);
+
+    const result = render(<Panel />, { wrapper });
+
+    const paragraph1 = await result.findByText("Paragraph 1");
+    const paragraph2 = await result.findByText("Paragraph 2");
+
+    fireEvent.click(paragraph1);
+    expect(paragraph1.classList.contains("highlight")).toBe(true);
+    expect(paragraph2.classList.contains("highlight")).toBe(false);
+    expect(mockSetSelectedParagraphId).toHaveBeenCalledWith("1");
+
+    fireEvent.click(paragraph2);
+    expect(paragraph1.classList.contains("highlight")).toBe(false);
+    expect(paragraph2.classList.contains("highlight")).toBe(true);
+    expect(mockSetSelectedParagraphId).toHaveBeenCalledWith("2");
+  });
+
+  test("renders LanguageSwitcher and updates language", async () => {
+    jest
+      .mocked(getArticle)
+      .mockResolvedValue(
+        "<div class='paragraph' data-paragraph-id='1'>Test Content</div>"
+      );
+    const result = render(<Panel language={Language.FRA} />, { wrapper });
+
+    const languageSwitcher = await result.findByRole("combobox");
+    expect(languageSwitcher).toBeInTheDocument();
+
+    fireEvent.change(languageSwitcher, { target: { value: Language.ENG } });
+    expect(jest.mocked(getArticle)).toHaveBeenCalledWith(
+      "123",
+      1,
+      Language.ENG
+    );
+  });
+});

frontend/src/components/Panel/Panel.tsx

@@ -1,35 +1,70 @@
-import { useQuery } from "@tanstack/react-query";
 import { useEffect, useRef, useState } from "react";
 
-import { getArticle } from "../../lib/api";
 import { Language } from "../../lib/types";
 import useUIStore from "../../store/uiStore";
 import LanguageSwitcher from "../LanguageSwitcher/LanguageSwitcher";
 
+import { useArticle } from "../../hooks/useArticle";
+import useNavStore from "../../store/navStore";
 import "../../styles/PanelContent.css";
+import ParagraphPreview from "../ParagraphPreview/ParagraphPreview";
+import { Tooltip, TooltipContent } from "../Tooltip/Tooltip";
 import styles from "./Panel.module.css";
 
 type PanelProps = {
-  celexId: string;
   language?: Language;
-  articleId: number;
 };
 
-function Panel({ celexId, language, articleId }: PanelProps) {
+function Panel({ language }: PanelProps) {
   const { selectedParagraphId, setSelectedParagraphId } = useUIStore();
 
   const [lang, setLang] = useState(language || Language.ENG);
   const articleRef = useRef<HTMLDivElement>(null);
-  const { data, isPending, error } = useQuery({
+  const { articleId, celexId, setArticleId } = useNavStore();
-    queryKey: ["article", celexId, articleId, lang],
+  const { data, isPending, error } = useArticle(celexId, articleId, lang);
-    queryFn: () => getArticle(celexId, articleId, lang),
+
-    enabled: !!celexId && !!articleId,
+  const [hoverArticleId, setHoverArticleId] = useState<number | null>(null);
-  });
+  const [hoverParagraphId, setHoverParagraphId] = useState<number | null>(null);
+  const [isTooltipOpen, setIsTooltipOpen] = useState(false);
 
   useEffect(() => {
     const articleElement = articleRef.current;
     if (!articleElement) return;
 
+    // Replace cross-reference links with page navigation
+    const crossRefs = articleElement.querySelectorAll(
+      "a.cross-ref"
+    ) as NodeListOf<HTMLAnchorElement>;
+    crossRefs.forEach((link) => {
+      const target = link.getAttribute("data-target");
+      const targetId = link.getAttribute("data-id");
+      const paragraphId = link.getAttribute("data-paragraph-id");
+
+      if (target && targetId) {
+        if (target === "article") {
+          // Replace link for easier copying, but navigate in-place to maintain UI state
+          link.setAttribute("href", `/${celexId}/articles/${targetId}`);
+          link.onclick = () => {
+            setArticleId(parseInt(targetId));
+            return false;
+          };
+          if (paragraphId) {
+            link.onmouseover = () => {
+              setHoverArticleId(parseInt(targetId));
+              setHoverParagraphId(parseInt(paragraphId));
+              setIsTooltipOpen(true);
+            };
+
+            link.onmouseout = () => {
+              setIsTooltipOpen(false);
+            };
+          }
+        }
+      } else {
+        console.warn("No target or ID found for link:", link);
+      }
+    });
+
     const paragraphs = articleElement.querySelectorAll(".paragraph");
 
     // Highlight the selected paragraph
@@ -66,11 +101,15 @@ function Panel({ celexId, language, articleId }: PanelProps) {
     // Cleanup event listeners
     return () => {
       console.log("Cleaning up event listeners");
+      // crossRefs.forEach((link) => {
+      //   link.onmouseover = null;
+      //   link.onmouseout = null;
+      // });
       paragraphs.forEach((element) => {
         element.removeEventListener("click", handleClick(element));
       });
     };
-  }, [articleRef, data, selectedParagraphId, setSelectedParagraphId]);
+  });
 
   if (isPending) return "Loading...";
   if (error) return "An error has occurred: " + error.message;
@@ -81,10 +120,24 @@ function Panel({ celexId, language, articleId }: PanelProps) {
         defaultLang={lang}
         onChange={setLang}
       ></LanguageSwitcher>
+      <Tooltip
+        open={isTooltipOpen}
+        placement="right-start"
+        onOpenChange={setIsTooltipOpen}
+      >
+        <TooltipContent>
+          <ParagraphPreview
+            celexId={celexId!}
+            articleId={hoverArticleId!}
+            paragraphId={hoverParagraphId!}
+            lang={lang}
+          />
+        </TooltipContent>
+      </Tooltip>
       <div
         ref={articleRef}
         lang={lang.substring(0, 2)}
-        dangerouslySetInnerHTML={{ __html: data || "" }}
+        dangerouslySetInnerHTML={{ __html: data }}
       />
     </div>
   );

frontend/src/components/ParagraphPreview/ParagraphPreview.tsx

@@ -0,0 +1,34 @@
+import { useParagraph } from "../../hooks/useParagraph";
+import { Language } from "../../lib/types";
+
+type ParagraphPreviewProps = {
+  celexId: string;
+  articleId: number;
+  paragraphId: number;
+  lang: Language;
+};
+
+function ParagraphPreview({
+  celexId,
+  articleId,
+  paragraphId,
+  lang,
+}: ParagraphPreviewProps) {
+  const { data, isPending, error } = useParagraph(
+    celexId,
+    articleId,
+    paragraphId,
+    lang
+  );
+
+  if (isPending) {
+    return <p>Loading...</p>;
+  }
+  if (error) {
+    return <p>Error loading paragraph: {error.message}</p>;
+  }
+
+  return <div dangerouslySetInnerHTML={{ __html: data }} />;
+}
+
+export default ParagraphPreview;

frontend/src/components/TOC/TOC.module.css

@@ -1,16 +1,15 @@
 .toc {
   font-size: 0.8rem;
-  min-width: 25vw;
+  flex: 1 0 25vw;
-  flex: 1 auto;
 
   &.hidden {
-    flex: 0 0;
+    display: none;
-    min-width: 0;
   }
 
+  transition: flex-basis 0.1s ease-in-out;
+
   overflow-y: scroll;
   overflow-x: wrap;
-  height: 100vh;
 
   .tocDivision {
     margin-block: 0.5rem;

frontend/src/components/TOC/TOC.tsx

@@ -1,16 +1,14 @@
 import { useState } from "react";
 import { Division } from "../../lib/types";
-import useNavState from "../../store/navStore";
+import useNavStore from "../../store/navStore";
 import styles from "./TOC.module.css";
 
-type TOC = Division[];
-
 type TOCProps = {
-  toc: TOC;
+  toc: Division[];
 };
 
 function TOC({ toc }: TOCProps) {
-  const { articleId, setArticleId } = useNavState();
+  const { articleId, setArticleId } = useNavStore();
 
   function containsArticle(division: Division, articleId: number): boolean {
     return division.content.some((c) => {
@@ -57,15 +55,17 @@ function TOC({ toc }: TOCProps) {
   const [isVisible, setIsVisible] = useState(true);
 
   return (
-    <nav className={[styles.toc, isVisible ? "" : styles.hidden].join(" ")}>
+    <>
       <button
         onClick={() => setIsVisible(!isVisible)}
         className={styles.toggleButton}
       >
         {isVisible ? "<" : ">"}
       </button>
-      {toc.map((division) => renderDivision(division))}
+      <nav className={[styles.toc, isVisible ? "" : styles.hidden].join(" ")}>
-    </nav>
+        {toc.map((division) => renderDivision(division))}
+      </nav>
+    </>
   );
 }
 export default TOC;

frontend/src/components/Tooltip/Tooltip.module.css

@@ -0,0 +1,10 @@
+.Tooltip {
+  background-color: #444;
+  color: white;
+  font-size: 90%;
+  padding: 4px 8px;
+  border-radius: 4px;
+  box-sizing: border-box;
+  width: max-content;
+  max-width: 60ch;
+}

frontend/src/components/Tooltip/Tooltip.tsx

@@ -0,0 +1,163 @@
+import type { Placement } from "@floating-ui/react";
+import {
+  autoUpdate,
+  flip,
+  FloatingPortal,
+  offset,
+  shift,
+  useDismiss,
+  useFloating,
+  useFocus,
+  useHover,
+  useInteractions,
+  useMergeRefs,
+  useRole,
+} from "@floating-ui/react";
+import * as React from "react";
+
+import styles from "./Tooltip.module.css";
+
+interface TooltipOptions {
+  initialOpen?: boolean;
+  placement?: Placement;
+  open?: boolean;
+  onOpenChange?: (open: boolean) => void;
+}
+
+export function useTooltip({
+  initialOpen = false,
+  placement = "top",
+  open: controlledOpen,
+  onOpenChange: setControlledOpen,
+}: TooltipOptions = {}) {
+  const [uncontrolledOpen, setUncontrolledOpen] = React.useState(initialOpen);
+
+  const open = controlledOpen ?? uncontrolledOpen;
+  const setOpen = setControlledOpen ?? setUncontrolledOpen;
+
+  const data = useFloating({
+    placement,
+    open,
+    onOpenChange: setOpen,
+    whileElementsMounted: autoUpdate,
+    middleware: [
+      offset(5),
+      flip({
+        crossAxis: placement.includes("-"),
+        fallbackAxisSideDirection: "start",
+        padding: 5,
+      }),
+      shift({ padding: 5 }),
+    ],
+  });
+
+  const context = data.context;
+
+  const hover = useHover(context, {
+    move: false,
+    enabled: controlledOpen == null,
+  });
+  const focus = useFocus(context, {
+    enabled: controlledOpen == null,
+  });
+  const dismiss = useDismiss(context);
+  const role = useRole(context, { role: "tooltip" });
+
+  const interactions = useInteractions([hover, focus, dismiss, role]);
+
+  return React.useMemo(
+    () => ({
+      open,
+      setOpen,
+      ...interactions,
+      ...data,
+    }),
+    [open, setOpen, interactions, data]
+  );
+}
+
+type ContextType = ReturnType<typeof useTooltip> | null;
+
+const TooltipContext = React.createContext<ContextType>(null);
+
+export const useTooltipContext = () => {
+  const context = React.useContext(TooltipContext);
+
+  if (context == null) {
+    throw new Error("Tooltip components must be wrapped in <Tooltip />");
+  }
+
+  return context;
+};
+
+export function Tooltip({
+  children,
+  ...options
+}: { children: React.ReactNode } & TooltipOptions) {
+  // This can accept any props as options, e.g. `placement`,
+  // or other positioning options.
+  const tooltip = useTooltip(options);
+  return (
+    <TooltipContext.Provider value={tooltip}>
+      {children}
+    </TooltipContext.Provider>
+  );
+}
+
+export const TooltipTrigger = React.forwardRef<
+  HTMLElement,
+  React.HTMLProps<HTMLElement> & { asChild?: boolean }
+>(function TooltipTrigger({ children, asChild = false, ...props }, propRef) {
+  const context = useTooltipContext();
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  const childrenRef = (children as any).ref;
+  const ref = useMergeRefs([context.refs.setReference, propRef, childrenRef]);
+
+  // `asChild` allows the user to pass any element as the anchor
+  if (asChild && React.isValidElement(children)) {
+    return React.cloneElement(
+      children,
+      context.getReferenceProps({
+        ref,
+        ...props,
+        ...children.props,
+        "data-state": context.open ? "open" : "closed",
+      })
+    );
+  }
+
+  return (
+    <button
+      ref={ref}
+      // The user can style the trigger based on the state
+      data-state={context.open ? "open" : "closed"}
+      {...context.getReferenceProps(props)}
+    >
+      {children}
+    </button>
+  );
+});
+
+export const TooltipContent = React.forwardRef<
+  HTMLDivElement,
+  React.HTMLProps<HTMLDivElement>
+>(function TooltipContent({ style, ...props }, propRef) {
+  const context = useTooltipContext();
+  const ref = useMergeRefs([context.refs.setFloating, propRef]);
+
+  if (!context.open) return null;
+
+  return (
+    <FloatingPortal>
+      <div
+        ref={ref}
+        className={styles.Tooltip}
+        style={{
+          ...context.floatingStyles,
+          ...style,
+        }}
+        {...context.getFloatingProps(props)}
+      />
+    </FloatingPortal>
+  );
+});

frontend/src/constants.ts

@@ -0,0 +1,3 @@
+const { VITE_API_URL: API_URL } = import.meta.env;
+
+export { API_URL };

frontend/src/hooks/toc.ts

@@ -0,0 +1,14 @@
+import { useQuery } from "@tanstack/react-query";
+import { getToc } from "../lib/api";
+import { Language } from "../lib/types";
+import useNavStore from "../store/navStore";
+
+export const useTOC = () => {
+  const celexId = useNavStore((state) => state.celexId);
+  const query = useQuery({
+    queryKey: ["toc", celexId],
+    queryFn: () => getToc(celexId!, Language.ENG),
+    enabled: !!celexId,
+  });
+  return query;
+};

frontend/src/hooks/urlSync.ts

@@ -1,11 +1,15 @@
 import { useEffect } from "react";
 import { useNavigate, useParams } from "react-router-dom";
-import useNavState from "../store/navStore";
+import useNavStore from "../store/navStore";
 
 export const useUrlSync = () => {
   const navigate = useNavigate();
   const { celexId: paramCelexId, articleId: paramArticleId } = useParams();
-  const { articleId, celexId, setArticleId, setCelexId } = useNavState();
+
+  const celexId = useNavStore.use.celexId();
+  const setCelexId = useNavStore.use.setCelexId();
+  const articleId = useNavStore.use.articleId();
+  const setArticleId = useNavStore.use.setArticleId();
 
   // Effect to read from URL when URL changes
   useEffect(() => {
@@ -26,13 +30,13 @@ export const useUrlSync = () => {
 
   // Update the URL when Zustand changes
   useEffect(() => {
-    if (celexId === null) {
+    if (!celexId) {
       return;
     }
     let targetUrl = `/${celexId}`;
     if (articleId !== null) {
       targetUrl += `/articles/${articleId}`;
     }
-    navigate(targetUrl, { replace: true });
+    navigate(targetUrl);
   }, [navigate, celexId, articleId]); // Only sync URL when Zustand changes
 };

frontend/src/hooks/useArticle.ts

@@ -0,0 +1,15 @@
+import { useQuery } from "@tanstack/react-query";
+import { getArticle } from "../lib/api";
+import { Language } from "../lib/types";
+
+export const useArticle = (
+  celexId: string | null,
+  articleId: number | null,
+  lang: Language
+) => {
+  return useQuery({
+    queryKey: ["article", celexId, articleId, lang],
+    queryFn: () => getArticle(celexId!, articleId!, lang),
+    enabled: !!celexId && !!articleId,
+  });
+};

frontend/src/hooks/useParagraph.ts

@@ -0,0 +1,16 @@
+import { useQuery } from "@tanstack/react-query";
+import { getParagraph } from "../lib/api";
+import { Language } from "../lib/types";
+
+export const useParagraph = (
+  celexId: string | null,
+  articleId: number | null,
+  paragraphId: number | null,
+  lang: Language
+) => {
+  return useQuery({
+    queryKey: ["paragraph", celexId, articleId, paragraphId, lang],
+    queryFn: () => getParagraph(celexId!, articleId!, paragraphId!, lang),
+    enabled: !!celexId && !!articleId && !!paragraphId,
+  });
+};

frontend/src/lib/api.ts

@@ -1,7 +1,14 @@
-import TOC from "../components/TOC/TOC";
+import Axios from "axios";
-import { Language } from "./types";
+import { API_URL } from "../constants";
+import { Division, Language } from "./types";
 
-const API_URL = import.meta.env.VITE_API_URL;
+const axios = Axios.create({
+  baseURL: API_URL,
+  timeout: 5000,
+  headers: {
+    "Content-Type": "application/json",
+  },
+});
 
 async function getArticle(
   celexId: string,
@@ -11,22 +18,40 @@ async function getArticle(
   console.debug(
     `Fetching article ${article} for CELEX ID ${celexId} in ${language} language`
   );
-  const response = await fetch(
+  const response = await axios.get<string>(
-    `${API_URL}/${celexId}/articles/${article}/${language}`
+    `${celexId}/articles/${article}/${language}`
   );
-  return await response.text();
+  return response.data;
+}
+
+async function getParagraph(
+  celexId: string,
+  article: number,
+  paragraph: number,
+  language: string
+): Promise<string> {
+  console.debug(
+    `Fetching article ${article} paragraph ${paragraph} for CELEX ID ${celexId} in ${language} language`
+  );
+  const response = await axios.get<string>(
+    `${celexId}/articles/${article}/${paragraph}/${language}`
+  );
+  return response.data;
 }
 
 async function getArticleIds(celexId: string): Promise<number[]> {
   console.debug(`Fetching article list for CELEX ID ${celexId}`);
-  const response = await fetch(`${API_URL}/${celexId}/articles`);
+  const response = await axios.get<number[]>(`${celexId}/articles`);
-  return await response.json();
+  return response.data;
 }
 
-async function getToc(celexId: string, language: Language): Promise<TOC> {
+async function getToc(
+  celexId: string,
+  language: Language
+): Promise<Division[]> {
   console.debug(`Fetching TOC for CELEX ID ${celexId}`);
-  const response = await fetch(`${API_URL}/${celexId}/toc/${language}`);
+  const response = await axios.get<Division[]>(`${celexId}/toc/${language}`);
-  return await response.json();
+  return response.data;
 }
 
-export { getArticle, getArticleIds, getToc };
+export { getArticle, getArticleIds, getParagraph, getToc };

frontend/src/lib/examples.ts

@@ -0,0 +1,9 @@
+export const examples = [
+  { name: "GDPR", id: "32016R0679" },
+  { name: "AI Act", id: "32024R1689" },
+  { name: "Cybersecurity Act", id: "32019R0881" },
+  { name: "Cyber Resilience Act", id: "32024R2847" },
+  { name: "Medical Device Regulation", id: "32017R0745" },
+  { name: "NIS 2 Directive", id: "32022L2555" },
+  { name: "Digital Services Act", id: "32022R2065" },
+];

frontend/src/lib/types.ts

@@ -16,7 +16,7 @@ type Division = {
   title: string;
   subtitle: string;
   level: number;
-  content: Article[] | Division[];
+  content: (Article | Division)[];
 };
 
 export { Language };

frontend/src/pages/MainView.tsx

@@ -1,10 +1,10 @@
 import App from "../App";
 import { useUrlSync } from "../hooks/urlSync";
-import useNavState from "../store/navStore";
+import useNavStore from "../store/navStore";
 
 function MainView() {
   useUrlSync();
-  const celexId = useNavState((state) => state.celexId);
+  const celexId = useNavStore.use.celexId();
 
   if (!celexId) {
     return <div>Error: No CELEX ID provided</div>;

frontend/src/store/navStore.ts

@@ -1,17 +1,19 @@
-import { create } from "zustand";
+import { create, StateCreator } from "zustand";
+import { createSelectors } from "../util/zustand";
 
-interface NavState {
+export type NavStore = {
   celexId: string | null;
   articleId: number | null;
   setCelexId: (celexId: string) => void;
   setArticleId: (articleId: number | null) => void;
-}
+};
 
-const useNavState = create<NavState>((set) => ({
+const navStoreCreator: StateCreator<NavStore> = (set) => ({
-  celexId: null,
+  celexId: "",
-  articleId: null,
+  articleId: 1,
   setCelexId: (celexId) => set({ celexId }),
   setArticleId: (articleId) => set({ articleId }),
-}));
+});
 
-export default useNavState;
+const useNavStore = createSelectors(create<NavStore>()(navStoreCreator));
+export default useNavStore;

frontend/src/store/uiStore.ts

@@ -11,11 +11,11 @@ interface UIState {
 const useUIStore = create<UIState>((set) => ({
   numPanels: 1,
   selectedParagraphId: null,
+
   addPanel: () => set((state) => ({ numPanels: state.numPanels + 1 })),
   removePanel: () =>
     set((state) => ({ numPanels: Math.max(state.numPanels - 1, 1) })),
-  setSelectedParagraphId: (selectedParagraphId: string | null) =>
+  setSelectedParagraphId: (selectedParagraphId) => set({ selectedParagraphId }),
-    set({ selectedParagraphId }),
 }));
 
 export default useUIStore;

frontend/src/styles/PanelContent.css

@@ -5,6 +5,8 @@
 }
 
 article {
+  max-width: 64ch;
+
   .list-lower-alpha {
     list-style-type: lower-alpha;
   }

frontend/src/util/zustand.ts

@@ -0,0 +1,18 @@
+import { StoreApi, UseBoundStore } from "zustand";
+
+type WithSelectors<S> = S extends { getState: () => infer T }
+  ? S & { use: { [K in keyof T]: () => T[K] } }
+  : never;
+
+export const createSelectors = <S extends UseBoundStore<StoreApi<object>>>(
+  _store: S
+) => {
+  const store = _store as WithSelectors<typeof _store>;
+  store.use = {};
+  for (const k of Object.keys(store.getState())) {
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    (store.use as any)[k] = () => store((s) => s[k as keyof typeof s]);
+  }
+
+  return store;
+};

frontend/tsconfig.app.json

@@ -24,5 +24,5 @@
 
     "plugins": [{ "name": "typescript-plugin-css-modules" }]
   },
-  "include": ["src"]
+  "include": ["src", "./jest.setup.ts", "__mocks__"]
 }

main.py

@@ -19,7 +19,7 @@ app = modal.App("formex-viewer", image=app_image)
 )
 @modal.asgi_app()
 def fastapi_app():
-    import fastapi.staticfiles
+    import fastapi
 
     from formex_viewer.server import app as formex_app
 
@@ -28,7 +28,7 @@ def fastapi_app():
     @formex_app.get("/{path:path}")
     async def frontend_handler(path: str):
         fp = assets_path / path
-        if not fp.exists():
+        if not fp.exists() or not fp.is_file():
             fp = assets_path / "index.html"
         return fastapi.responses.FileResponse(fp)
 

pyproject.toml

@@ -20,3 +20,8 @@ formex-viewer = "formex_viewer:main"
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
+
+[dependency-groups]
+dev = [
+    "pytest>=8.3.5",
+]

src/formex_viewer/formex4.py

@@ -1,12 +1,16 @@
 import html
 import re
-from typing import Optional, Union
+import warnings
+from dataclasses import dataclass
+from typing import Literal, Optional, Union, cast
 
 import lxml.etree
 from lxml import etree as ET
 
+from formex_viewer.main import Language
 
-def text_content(el: lxml.etree.Element) -> str:
+
+def text_content(el: ET._Element) -> str:
     """Get the text content of an XML element, including all child elements."""
 
     def _iterate(el):
@@ -23,24 +27,130 @@ def text_content(el: lxml.etree.Element) -> str:
     return "".join(_iterate(el))
 
 
+@dataclass
+class CrossReference:
+    target: Literal["article", "annex"]
+    text: str
+    id: str
+    paragraph: int | None = None
+
+
+def extract_xrefs(el: ET._Element, language: Language) -> list[CrossReference]:
+    """Extract cross-references from an XML element.
+
+    Args:
+        el: The XML element to extract cross-references from.
+
+    Returns:
+        A dictionary with cross-reference IDs as keys and their text content as values.
+    """
+    crossrefs = []
+    text = text_content(el)
+
+    PATTERN_PARTS = {
+        Language.ENG: {
+            "article": r"(Art\.|Articles?)",
+            "annex": r"(Ann\.|Annex)",
+            "exclusion": r"(?! of(?! this))",
+        },
+        Language.DEU: {
+            "article": r"(Art\.|Artikels?)",
+            "annex": r"(Anhang)",
+            "exclusion": r"(?! von)",
+        },
+    }
+
+    if language not in PATTERN_PARTS:
+        warnings.warn(
+            f"Language '{language}' not supported for cross-reference extraction. Returning empty list."
+        )
+        return []
+
+    # Prevent zealous matching of references to other texts by using a negative lookahead
+    # Also, match only at word boundaries to prevent partial matches
+    parts = PATTERN_PARTS[language]
+    patterns = {
+        "article": rf"\b{parts["article"]}\s+(?P<art_num>\d+)(?:[(](?P<parag_num>\d+)[)])?(?:{parts["exclusion"]})",
+        "annex": rf"\b{parts["annex"]}\s+(?P<annex_num>[DILMVX]+)(?:{parts["exclusion"]})",
+    }
+    for key, pattern in patterns.items():
+        matches = re.finditer(pattern, text, flags=re.IGNORECASE)
+        for match in matches:
+            crossref_id = (
+                match.group("art_num") if key == "article" else match.group("annex_num")
+            )
+            parag_num = match.groupdict().get("parag_num")
+
+            if key not in ["article", "annex"]:
+                raise RuntimeError()
+
+            crossref_text = match.group(0)
+            crossrefs.append(
+                CrossReference(
+                    target=key,
+                    id=crossref_id,
+                    paragraph=int(parag_num) if parag_num else None,
+                    text=crossref_text,
+                )
+            )
+    return crossrefs
+
+
+def extract_article(doc: ET._Element, article_id: int) -> ET._Element | None:
+    """Extract a specific article from a Formex document.
+
+    Args:
+        doc: The XML document to extract from.
+        article_id: The article number.
+
+    Returns:
+        The extracted article element.
+    """
+
+    # Use XPath to find the specific article
+    xpath = f".//ARTICLE[@IDENTIFIER='{article_id:03d}']"
+    return doc.xpath(xpath)[0] if doc.xpath(xpath) else None
+
+
+def extract_paragraph(
+    doc: ET._Element, article_id: int, paragraph_id: int
+) -> ET._Element | None:
+    """Extract a specific paragraph from an article in a Formex document.
+
+    Args:
+        doc: The XML document to extract from.
+        article_id: The article number.
+        paragraph_id: The paragraph number.
+
+    Returns:
+        The extracted paragraph element.
+    """
+
+    # Use XPath to find the specific paragraph
+    xpath = f".//PARAG[@IDENTIFIER='{article_id:03d}.{paragraph_id:03d}']"
+    return doc.xpath(xpath)[0] if doc.xpath(xpath) else None
+
+
 class FormexArticleConverter:
     """Converts Formex XML <ARTICLE> elements to semantic HTML5."""
 
-    def __init__(self, namespace: Optional[str] = None):
+    def __init__(self, language: Language, namespace: Optional[str] = None):
         """
         Initialize the converter.
 
         Args:
+            language: Language object to determine the language for cross-reference extraction
             namespace: Optional XML namespace to use when parsing elements
         """
         self.ns = namespace
+        self.language = language
         self.ns_prefix = f"{{{namespace}}}" if namespace else ""
 
     def _get_tag(self, tag: str) -> str:
         """Get the tag name with namespace if available."""
         return f"{self.ns_prefix}{tag}"
 
-    def _get_text(self, element: ET.Element) -> str:
+    def _get_text(self, element: ET._Element) -> str:
         """Get the text content of an element, including all nested text.
 
         This uses lxml's text_content() method when available, falling back to
@@ -55,7 +165,7 @@ class FormexArticleConverter:
         except AttributeError:
             # Fall back to manual traversal if text_content() is not available
             text = element.text or ""
-            for child in element:
+            for child in element.iterchildren(tag="*"):
                 text += self._get_text(child)
                 if child.tail:
                     text += child.tail
@@ -67,7 +177,16 @@ class FormexArticleConverter:
         clean_id = re.sub(r"[^a-zA-Z0-9-]", "-", identifier)
         return f"art-{clean_id}"
 
-    def _convert_btx(self, element: ET.Element) -> str:
+    def _replace_xref(self, text: str, xref: CrossReference) -> str:
+        """Replace a cross-reference instance with semantic markup in the text."""
+        # Replace the cross-reference text with a link
+        text = text.replace(
+            xref.text,
+            f'<a href="" data-target="{xref.target}" data-id="{xref.id}" data-paragraph-id="{xref.paragraph or ''}" class="cross-ref">{xref.text}</a>',
+        )
+        return text
+
+    def _convert_btx(self, element: ET._Element) -> str:
         """
         Convert basic text elements (t_btx, t_btx.seq) to HTML.
 
@@ -79,7 +198,15 @@ class FormexArticleConverter:
 
         result = element.text or ""
 
-        for child in element:
+        is_title = element.tag in ("TI", "STI", "TI.ART", "STI.ART")
+        if not is_title and not element.getchildren():
+            # Cross-references should be treated at the deepest level
+            xrefs = extract_xrefs(element, self.language)
+            for xref in xrefs:
+                # Replace the cross-reference text with a link
+                result = self._replace_xref(result, xref)
+
+        for child in element.iterchildren(tag="*"):
             child_tag = child.tag.replace(self.ns_prefix, "")
 
             # Process common inline elements
@@ -176,11 +303,17 @@ class FormexArticleConverter:
                 result += self._convert_btx(child)
 
             if child.tail:
-                result += child.tail
+                xrefs = extract_xrefs(child, self.language)
+                tail_text = child.tail
+                for xref in xrefs:
+                    # Replace the cross-reference text with a link
+                    tail_text = self._replace_xref(tail_text, xref)
+
+                result += tail_text
 
         return result
 
-    def _convert_list(self, list_element: ET.Element) -> str:
+    def _convert_list(self, list_element: ET._Element) -> str:
         """Convert a Formex LIST element to HTML list items."""
         result = ""
         # Using lxml's xpath to get direct child ITEM elements
@@ -200,6 +333,12 @@ class FormexArticleConverter:
                     if no_p is not None and txt is not None:
                         num = self._get_text(no_p)
                         text = self._get_text(txt)
+
+                        # Handle cross-references within the text
+                        xrefs = extract_xrefs(txt, self.language)
+                        for xref in xrefs:
+                            text = self._replace_xref(text, xref)
+
                         item_content += f'<span class="item-number">{num}</span> {text}'
                 elif child_tag == "P":
                     # Regular paragraph
@@ -212,41 +351,40 @@ class FormexArticleConverter:
 
         return result
 
-    def _convert_alinea(self, alinea: ET.Element) -> str:
+    def _convert_alinea(self, alinea: ET._Element) -> str:
         """Convert an ALINEA element to HTML."""
         return f'<p class="alinea">{self._convert_btx(alinea)}</p>'
 
-    def _convert_parag(self, parag: ET.Element) -> str:
+    def _convert_parag(self, parag: ET._Element) -> str:
         """Convert a PARAG (paragraph) element to HTML."""
         identifier = parag.get("IDENTIFIER", "")
         parag_id = self._create_id(identifier) if identifier else ""
 
-        # Get the paragraph number using XPath
-        no_parag_elems = parag.xpath(f"./{self._get_tag('NO.PARAG')}")
-        parag_num = self._get_text(no_parag_elems[0]) if no_parag_elems else ""
-
-        # Process the alineas within the paragraph
         content = ""
-        for alinea in parag.xpath(f"./{self._get_tag('ALINEA')}"):
+        for child in parag.iterchildren(tag="*"):
-            content += self._convert_alinea(alinea)
+            child_tag = child.tag.replace(self.ns_prefix, "")
+            if child_tag == "ALINEA":
+                content += self._convert_alinea(child)
+            elif child_tag == "COMMENT":
+                content += f'<div class="comment">{self._convert_btx(child)}</div>'
+            elif child_tag == "QUOT.S":
+                content += f'<blockquote class="quotation">{self._convert_btx(child)}</blockquote>'
+            elif child_tag == "NO.PARAG":
+                content += (
+                    f'<span class="paragraph-number">{self._convert_btx(child)}</span>'
+                )
+            else:
+                raise RuntimeError(
+                    f"Unexpected child element '{child_tag}' in PARAG: {text_content(child)}"
+                )
 
-        # Process any comments
+        return f'<div class="paragraph" data-paragraph-id="{parag_id}">{content}</div>'
-        for comment in parag.xpath(f"./{self._get_tag('COMMENT')}"):
-            content += f'<div class="comment">{self._convert_btx(comment)}</div>'
 
-        # Process any quotations
+    def _convert_subdiv(self, subdiv: ET._Element) -> str:
-        for quot in parag.xpath(f"./{self._get_tag('QUOT.S')}"):
+        """Convert a SUBDIV (subdivision) element to HTML, preserving child order."""
-            content += (
+        # Get the title using XPath (should be the first TITLE child if present)
-                f'<blockquote class="quotation">{self._convert_btx(quot)}</blockquote>'
-            )
-
-        return f'<div class="paragraph" data-paragraph-id="{parag_id}"><span class="paragraph-number">{parag_num}</span>{content}</div>'
-
-    def _convert_subdiv(self, subdiv: ET.Element) -> str:
-        """Convert a SUBDIV (subdivision) element to HTML."""
-        # Get the title using XPath
-        title_elems = subdiv.xpath(f"./{self._get_tag('TITLE')}")
         title = ""
+        title_elems = subdiv.xpath(f"./{self._get_tag('TITLE')}")
         if title_elems:
             title_elem = title_elems[0]
             # Process TI (title) and STI (subtitle) elements
@@ -261,34 +399,30 @@ class FormexArticleConverter:
             if sti_list:
                 title += f'<h5 class="subdivision-subtitle">{" ".join(sti_list)}</h5>'
 
-        # Process content: either paragraphs, alineas, or nested subdivisions
+        # Process all children in order, skipping TITLE (already handled)
         content = ""
-
+        for child in subdiv.iterchildren(tag="*"):
-        # Process paragraphs directly under this subdivision
+            child_tag = child.tag.replace(self.ns_prefix, "")
-        for parag in subdiv.xpath(f"./{self._get_tag('PARAG')}"):
+            if child_tag == "TITLE":
-            content += self._convert_parag(parag)
+                continue  # already handled
-
+            elif child_tag == "PARAG":
-        # Process alineas directly under this subdivision
+                content += self._convert_parag(child)
-        for alinea in subdiv.xpath(f"./{self._get_tag('ALINEA')}"):
+            elif child_tag == "ALINEA":
-            content += self._convert_alinea(alinea)
+                content += self._convert_alinea(child)
-
+            elif child_tag == "COMMENT":
-        # Process comments directly under this subdivision
+                content += f'<div class="comment">{self._convert_btx(child)}</div>'
-        for comment in subdiv.xpath(f"./{self._get_tag('COMMENT')}"):
+            elif child_tag == "QUOT.S":
-            content += f'<div class="comment">{self._convert_btx(comment)}</div>'
+                content += f'<blockquote class="quotation">{self._convert_btx(child)}</blockquote>'
-
+            elif child_tag == "SUBDIV":
-        # Process quotations directly under this subdivision
+                content += self._convert_subdiv(child)
-        for quot in subdiv.xpath(f"./{self._get_tag('QUOT.S')}"):
+            else:
-            content += (
+                raise RuntimeError(
-                f'<blockquote class="quotation">{self._convert_btx(quot)}</blockquote>'
+                    f"Unexpected child element '{child_tag}' in SUBDIV: {text_content(child)}"
-            )
+                )
-
-        # Process nested subdivisions directly under this subdivision
-        for sub in subdiv.xpath(f"./{self._get_tag('SUBDIV')}"):
-            content += self._convert_subdiv(sub)
 
         return f'<section class="subdivision">{title}{content}</section>'
 
-    def convert_article(self, article: Union[str, ET.Element]) -> str:
+    def convert_article(self, article: Union[str, ET._Element]) -> str:
         """
         Convert a Formex <ARTICLE> element to HTML5.
 
@@ -302,7 +436,9 @@ class FormexArticleConverter:
         if isinstance(article, str):
             try:
                 parser = ET.XMLParser(remove_blank_text=True)
-                article = ET.fromstring(article.encode("utf-8"), parser)
+                article = cast(
+                    ET._Element, ET.fromstring(article.encode("utf-8"), parser)
+                )
             except ET.XMLSyntaxError as e:
                 return f"<p>Error parsing XML: {e}</p>"
 
@@ -325,43 +461,36 @@ class FormexArticleConverter:
         article_subtitle = self._convert_btx(sti_art) if sti_art is not None else ""
 
         # Build the header section
-        header = f'<header><h3 class="article-title">{article_title}</h3>'
+        if article_title and article_subtitle:
-        if article_subtitle:
+            header = f'<header><h3 class="article-title">{article_title}</h3>'
-            header += f'<h4 class="article-subtitle">{article_subtitle}</h4>'
+            if article_subtitle:
-        header += "</header>"
+                header += f'<h4 class="article-subtitle">{article_subtitle}</h4>'
+            header += "</header>"
+        else:
+            header = ""
 
         # Process the content based on what's present
         content = ""
 
-        # Check if we have alineas directly under the article
+        # Process all child elements (except TITLE) in tree order
-        alineas = article.xpath(f"./{self._get_tag('ALINEA')}")
+        for child in article.iterchildren(tag="*"):
-        if alineas:
+            child_tag = child.tag.replace(self.ns_prefix, "")
-            for alinea in alineas:
+            if child_tag in ["TI.ART", "STI.ART"]:
-                content += self._convert_alinea(alinea)
+                continue  # already handled
-
+            elif child_tag == "ALINEA":
-        # Check if we have paragraphs directly under the article
+                content += self._convert_alinea(child)
-        parags = article.xpath(f"./{self._get_tag('PARAG')}")
+            elif child_tag == "PARAG":
-        if parags:
+                content += self._convert_parag(child)
-            for parag in parags:
+            elif child_tag == "COMMENT":
-                content += self._convert_parag(parag)
+                content += f'<div class="comment">{self._convert_btx(child)}</div>'
-
+            elif child_tag == "QUOT.S":
-        # Check for comments directly under the article
+                content += f'<blockquote class="quotation">{self._convert_btx(child)}</blockquote>'
-        comments = article.xpath(f"./{self._get_tag('COMMENT')}")
+            elif child_tag == "SUBDIV":
-        if comments:
+                content += self._convert_subdiv(child)
-            for comment in comments:
+            else:
-                content += f'<div class="comment">{self._convert_btx(comment)}</div>'
+                raise RuntimeError(
-
+                    f"Unexpected child element '{child_tag}' in ARTICLE: {text_content(child)}"
-        # Check for quotations directly under the article
+                )
-        quots = article.xpath(f"./{self._get_tag('QUOT.S')}")
-        if quots:
-            for quot in quots:
-                content += f'<blockquote class="quotation">{self._convert_btx(quot)}</blockquote>'
-
-        # Check for subdivisions directly under the article
-        subdivs = article.xpath(f"./{self._get_tag('SUBDIV')}")
-        if subdivs:
-            for subdiv in subdivs:
-                content += self._convert_subdiv(subdiv)
 
         # Assemble the complete article
         return f'<article id="{article_id}" class="formex-article">{header}<div class="article-content">{content}</div></article>'

src/formex_viewer/server.py

@@ -2,7 +2,11 @@ import lxml.etree as ET
 from fastapi import APIRouter, FastAPI, Response
 from fastapi.middleware.cors import CORSMiddleware
 
-from formex_viewer.formex4 import FormexArticleConverter
+from formex_viewer.formex4 import (
+    FormexArticleConverter,
+    extract_article,
+    extract_paragraph,
+)
 from formex_viewer.main import (
     CellarClient,
     CellarIdentifier,
@@ -68,9 +72,16 @@ def article_ids(celex_id: str, language: Language = Language.ENG):
 
 @api_router.get("/{celex_id}/toc/{language}")
 def toc(celex_id: str, language: Language = Language.ENG):
+    def _extract_text(root: ET.Element, tag: str) -> str:
+        """
+        Extract text from the given tag in the XML element.
+        """
+        text = root.xpath(f"{tag}//text()")
+        return "".join(text) if text else ""
+
     def _handle_division(division: ET.Element, level: int):
-        title = ti_el[0] if (ti_el := division.xpath("TITLE/TI//text()")) else ""
+        title = _extract_text(division, "TITLE/TI")
-        subtitle = sti_el[0] if (sti_el := division.xpath("TITLE/STI//text()")) else ""
+        subtitle = _extract_text(division, "TITLE/STI")
 
         subdivisions = []
         for subdivision in division.xpath("DIVISION") or []:
@@ -81,10 +92,9 @@ def toc(celex_id: str, language: Language = Language.ENG):
             art_id = article.get("IDENTIFIER")
             if not art_id:
                 continue
-            art_title = ti_el[0] if (ti_el := article.xpath("TI.ART//text()")) else ""
+
-            art_subtitle = (
+            art_title = _extract_text(article, "TI.ART")
-                sti_el[0] if (sti_el := article.xpath("STI.ART//text()")) else ""
+            art_subtitle = _extract_text(article, "STI.ART")
-            )
             articles.append(
                 {
                     "id": int(art_id.lstrip("0")),
@@ -115,21 +125,51 @@ def toc(celex_id: str, language: Language = Language.ENG):
 
 
 @api_router.get("/{celex_id}/articles/{article_id}/{language}")
-def article(celex_id: str, article_id: int, language: Language = Language.ENG):
+def article(
+    celex_id: str,
+    article_id: int,
+    language: Language = Language.ENG,
+):
     """
     Fetch an article from the server.
     """
     xml = _get_fmx4_data(celex_id, language)
+    article = extract_article(xml, article_id=article_id)
 
-    article_xpath = "//ARTICLE"
+    if article is None:
-    articles = xml.xpath(article_xpath)
+        return Response(
-    for article in articles:
+            "Article not found",
-        num = article.get("IDENTIFIER").lstrip("0")
+            status_code=404,
-        if num == str(article_id):
+        )
-            return Response(
+
-                FormexArticleConverter().convert_article(article),
+    return Response(
-                media_type="text/html",
+        FormexArticleConverter(language=language).convert_article(article),
-            )
+        media_type="text/html",
+    )
+
+
+@api_router.get("/{celex_id}/articles/{article_id}/{parag_id}/{language}")
+def paragraph(
+    celex_id: str,
+    article_id: int,
+    parag_id: int,
+    language: Language = Language.ENG,
+):
+    """
+    Fetch a paragraph within an article from the server.
+    """
+    xml = _get_fmx4_data(celex_id, language)
+    parag = extract_paragraph(xml, article_id=article_id, paragraph_id=parag_id)
+    if parag is None:
+        return Response(
+            "Paragraph not found",
+            status_code=404,
+        )
+
+    return Response(
+        FormexArticleConverter(language=language)._convert_parag(parag),
+        media_type="text/html",
+    )
 
 
 app.include_router(api_router, prefix="/api")

tests/test_parser.py

@@ -0,0 +1,52 @@
+import pytest
+from lxml import etree as ET
+
+from formex_viewer.formex4 import FormexArticleConverter
+from formex_viewer.main import Language
+
+
+@pytest.fixture
+def converter():
+    return FormexArticleConverter(language=Language.ENG)
+
+
+def test_convert_tree_order(converter):
+    """Test that the order of HTML blocks in the converted article matches the order of elements in the XML tree."""
+
+    xml = """
+    <ARTICLE>
+        <SUBDIV>
+            <TITLE>
+                <TI>Subdivision Title</TI>
+                <STI>Subdivision Subtitle</STI>
+            </TITLE>
+            <PARAG IDENTIFIER="001.001">
+                <NO.PARAG>1</NO.PARAG>
+                <ALINEA>Paragraph 1 text.</ALINEA>
+            </PARAG>
+            <COMMENT>Comment text.</COMMENT>
+            <ALINEA>Alinea text.</ALINEA>
+            <QUOT.S>Quotation text.</QUOT.S>
+            <SUBDIV>
+                <TITLE>
+                    <TI>Nested Subdivision</TI>
+                </TITLE>
+                <ALINEA>Nested alinea.</ALINEA>
+            </SUBDIV>
+        </SUBDIV>
+    </ARTICLE>
+    """
+    parser = ET.XMLParser(remove_blank_text=True)
+    el = ET.fromstring(xml, parser)
+    html = converter.convert_article(el)
+
+    # Check that the order of HTML blocks matches the order of elements in the XML tree
+    idx_title = html.index("Subdivision Title")
+    idx_parag = html.index('class="paragraph"')
+    idx_comment = html.index("Comment text.")
+    idx_alinea = html.index("Alinea text.")
+    idx_quot = html.index("Quotation text.")
+    idx_nested = html.index("Nested Subdivision")
+
+    # The order in the XML: title, parag, alinea, comment, quot, nested subdiv
+    assert idx_title < idx_parag < idx_comment < idx_alinea < idx_quot < idx_nested