import { describe, expect, it } from 'vitest';
import {
  collectLinkTargets,
  isMarkdownContentPreserved,
} from '../round-trip-check';
import { createProductionMarkdown } from './production-markdown-test-helpers';

// The paste path must accept Markdown that merely normalizes and refuse only
// Markdown whose content the parser drops. A byte comparison conflates the two
// and rejected most Markdown copied from anywhere else.
function setup() {
  const markdown = createProductionMarkdown();
  // Uses the production payload collector, so the gate cannot be tested
  // against a shape the app no longer produces.
  return (source: string) =>
    isMarkdownContentPreserved(
      source,
      collectLinkTargets(markdown.parser.parse(source)),
    );
}

describe('Markdown paste content gate', () => {
  const accepted = [
    ['emphasis delimiter', '*italic text*'],
    ['strong delimiter', '__bold text__'],
    ['bullet marker', '* bullet one\n* bullet two'],
    ['plus bullet', '+ plus bullet'],
    ['paren ordered', '1) paren ordered'],
    ['setext heading', 'Setext Heading\n=============='],
    ['extra heading space', '#  extra space heading'],
    ['closed ATX heading', '# ATX closed heading #'],
    ['hard wrapped paragraph', 'para one\nhard wrapped continuation'],
    ['indented code block', '    indented code block'],
    ['uppercase task', '- [X] uppercase task'],
    ['plain paragraph', 'plain paragraph'],
    ['inline link with title', '[visible](https://example.com "the title")'],
    ['fenced code with language', '```js\nconst a = 1;\n```'],
    // The list number is never stored on the node — the serializer always
    // writes "1." — so a two-digit marker exists only in the source.
    [
      'ordered list past item nine',
      [
        '1. alpha',
        '2. bravo',
        '3. charlie',
        '4. delta',
        '5. echo',
        '6. foxtrot',
        '7. golf',
        '8. hotel',
        '9. india',
        '10. juliett',
      ].join('\n'),
    ],
    ['two-digit ordered marker', '12. twelfth item'],
    // Entities decode to punctuation, so their spelling is gone by design.
    ['named entity', 'fish & chips'],
    ['numeric entity', '© notice text'],
    // A resolved reference link keeps its content: the label is syntax, and
    // the definition reappears as the link's href.
    [
      'reference-style link',
      '[click here][docs]\n\n[docs]: https://example.com/page',
    ],
    [
      'reference-style image',
      '![diagram][img]\n\n[img]: https://example.com/d.png',
    ],
    ['collapsed reference link', '[docs][]\n\n[docs]: https://example.com/p'],
    ['shortcut reference link', '[docs]\n\n[docs]: https://example.com/p'],
  ] as const;

  for (const [name, source] of accepted) {
    it(`accepts ${name}`, () => {
      expect(setup()(source)).toBe(true);
    });
  }

  // A URL the parser percent-encodes must still be recognised as the same
  // target, or an accented or reserved character makes a lossless paste fail.
  const encodedUrls = [
    ['accented path', 'https://example.com/\u00fcber'],
    ['non-latin path', 'https://example.com/\u65e5\u672c\u8a9e'],
    ['accented query', 'https://example.com/path?q=caf\u00e9'],
    ['relative note link', './notes/\u00dcbersicht.md'],
    ['reserved characters', 'https://example.com/a|b'],
  ] as const;

  for (const [name, url] of encodedUrls) {
    it(`accepts a reference link whose URL has ${name}`, () => {
      expect(setup()(`[see][ref]\n\n[ref]: ${url}`)).toBe(true);
    });
  }

  it('refuses a dropped definition even when its URL appears in the prose', () => {
    // Matching against the document's text would let this unrelated mention
    // vouch for the definition that was actually discarded.
    expect(
      setup()(
        'alpha https://example.com/x beta\n\n[gone]: https://example.com/x',
      ),
    ).toBe(false);
  });

  it('refuses a dropped definition whose label and URL are single characters', () => {
    // The word-comparison approach skipped one-character tokens, so this
    // dropped definition slipped through and only "KEEP" was inserted.
    expect(setup()('KEEP\n\n[x]: /a')).toBe(false);
  });

  it('refuses a dropped link reference definition', () => {
    // The definition and the missing label vanish from the document, so this
    // paste really would discard what the user copied.
    expect(setup()('[visible][missing]\n\n[unused]: https://example.com')).toBe(
      false,
    );
  });

  // CommonMark consumes definitions nested inside blockquotes and list items
  // too; a gate that only scans line starts would let these vanish silently.
  const nestedDropped = [
    ['blockquote', '> quoted text\n>\n> [gone]: https://example.com/spec'],
    ['list item', '- item text\n\n- [gone]: https://example.com/spec'],
    [
      'ordered list item',
      '1. item text\n\n2. [gone]: https://example.com/spec',
    ],
    [
      'blockquote inside a list item',
      '- > quoted\n  >\n  > [gone]: https://example.com/spec',
    ],
  ] as const;

  for (const [name, source] of nestedDropped) {
    it(`refuses a dropped definition inside a ${name}`, () => {
      expect(setup()(source)).toBe(false);
    });
  }

  it('accepts a resolved definition inside a blockquote', () => {
    expect(
      setup()('> see [the spec][ref]\n>\n> [ref]: https://example.com/spec'),
    ).toBe(true);
  });
});