/* ============================================================
   readme.ts — fetch + parse + regenerate a Build Small Space README

   Tracks and badges a participant wants to be considered for are encoded
   as namespaced entries in the README frontmatter `tags:` list:
     track:backyard   badge:tiny   ...
   Regeneration is surgical: only the `tags:` block is rewritten, so every
   other line of the participant's frontmatter (and the whole body) is kept
   byte-for-byte — which keeps the "what changed" diff clean.
   ============================================================ */

export const SPACE_OWNER = 'build-small-hackathon';

/** Raw README URL for a Space in the hackathon org. Public, no token needed. */
export function rawReadmeUrl(space: string): string {
	return `https://huggingface.co/spaces/${SPACE_OWNER}/${encodeURIComponent(space)}/raw/main/README.md`;
}

/**
 * Pull the bare space name out of whatever the user pasted — a full URL,
 * `owner/space`, or a leading-slash path all collapse to just `space`.
 */
export function normalizeSpaceName(input: string): string {
	let s = input.trim();
	// strip a pasted URL down to its path
	s = s.replace(/^https?:\/\/huggingface\.co\//i, '');
	s = s.replace(/^\/+/, '');
	s = s.replace(/^spaces\//i, '');
	s = s.replace(new RegExp(`^${SPACE_OWNER}/`, 'i'), '');
	// drop anything after the space segment (/tree/main, /blob/..., trailing slash)
	s = s.split('/')[0];
	return s.trim();
}

// The tool only manages the tags a participant self-declares: their tracks,
// the sponsor prizes they're entering, and their no-prize achievement badges.
// Anything else in `tags:` is treated as non-managed and preserved untouched.
export const MANAGED_PREFIXES = ['track:', 'sponsor:', 'achievement:'] as const;
export const isManagedTag = (t: string) => MANAGED_PREFIXES.some((p) => t.startsWith(p));
export const trackTag = (id: string) => `track:${id}`;
export const sponsorTag = (id: string) => `sponsor:${id}`;
export const achievementTag = (id: string) => `achievement:${id}`;

export interface ParsedReadme {
	hasFrontmatter: boolean;
	/** YAML text between the `---` fences (fences excluded). */
	frontmatter: string;
	/** Everything after the closing fence (or the whole doc if no frontmatter). */
	body: string;
	/** Tags parsed from the frontmatter `tags:` declaration. */
	tags: string[];
	/** Indentation of the existing `tags:` list items, so we can round-trip it. */
	tagsIndent: string;
	/** Char range of the whole `tags:` declaration within `frontmatter`, if present. */
	tagsRange: { start: number; end: number } | null;
}

const FRONTMATTER_RE = /^?---[ \t]*\r?\n([\s\S]*?)\r?\n---[ \t]*\r?\n?/;

const unquote = (s: string) => s.trim().replace(/^['"]|['"]$/g, '');

/** Locate and parse the `tags:` declaration inside a frontmatter block. */
function findTagsBlock(
	fm: string
): { tags: string[]; indent: string; start: number; end: number } | null {
	const lines = fm.split('\n');
	const offsets: number[] = [];
	let acc = 0;
	for (const l of lines) {
		offsets.push(acc);
		acc += l.length + 1; // + the '\n' that split removed
	}

	for (let i = 0; i < lines.length; i++) {
		const m = /^tags:[ \t]*(.*)$/.exec(lines[i]);
		if (!m) continue;
		const start = offsets[i];
		const inline = m[1].trim();

		// flow form on the same line: tags: [a, b, c]
		if (inline.startsWith('[')) {
			const tags = inline
				.replace(/^\[|\]$/g, '')
				.split(',')
				.map(unquote)
				.filter(Boolean);
			return { tags, indent: '  ', start, end: offsets[i] + lines[i].length };
		}

		// single scalar on the same line: tags: foo
		if (inline) {
			return { tags: [unquote(inline)], indent: '  ', start, end: offsets[i] + lines[i].length };
		}

		// block form: subsequent "- item" lines. YAML lets the sequence sit at the
		// same indent as `tags:` (no extra indent) or be indented under it — accept
		// either, and remember the indent the first item used so we round-trip it.
		const tags: string[] = [];
		let indent = '  ';
		let last = i;
		for (let j = i + 1; j < lines.length; j++) {
			const lm = /^([ \t]*)-[ \t]+(.*)$/.exec(lines[j]);
			if (!lm) break; // any non list-item line (including blank) ends the block
			if (tags.length === 0) indent = lm[1]; // first item sets the indent
			tags.push(unquote(lm[2]));
			last = j;
		}
		return { tags, indent, start, end: offsets[last] + lines[last].length };
	}
	return null;
}

export function parseReadme(raw: string): ParsedReadme {
	const m = FRONTMATTER_RE.exec(raw);
	if (!m) {
		return {
			hasFrontmatter: false,
			frontmatter: '',
			body: raw,
			tags: [],
			tagsIndent: '  ',
			tagsRange: null
		};
	}
	const frontmatter = m[1];
	const body = raw.slice(m[0].length);
	const block = findTagsBlock(frontmatter);
	return {
		hasFrontmatter: true,
		frontmatter,
		body,
		tags: block?.tags ?? [],
		tagsIndent: block?.indent ?? '  ',
		tagsRange: block ? { start: block.start, end: block.end } : null
	};
}

/** Split a README's existing tags into selected track / sponsor / achievement ids. */
export function splitManaged(tags: string[]): {
	tracks: Set<string>;
	sponsors: Set<string>;
	achievements: Set<string>;
} {
	const tracks = new Set<string>();
	const sponsors = new Set<string>();
	const achievements = new Set<string>();
	for (const t of tags) {
		if (t.startsWith('track:')) tracks.add(t.slice('track:'.length));
		else if (t.startsWith('sponsor:')) sponsors.add(t.slice('sponsor:'.length));
		else if (t.startsWith('achievement:')) achievements.add(t.slice('achievement:'.length));
	}
	return { tracks, sponsors, achievements };
}

/** The `tags:` YAML block for a list of tags, in block-list form. */
export const tagsBlock = (tags: string[], indent = '  ') =>
	tags.length ? 'tags:\n' + tags.map((t) => `${indent}- ${t}`).join('\n') : 'tags:';

/**
 * Rewrite the README so its `tags:` list holds exactly the participant's
 * non-managed tags (preserved as-is) followed by `selectedTags`
 * (already namespaced + ordered by the caller).
 */
export function regenerateReadme(raw: string, selectedTags: string[]): string {
	const parsed = parseReadme(raw);
	const preserved = parsed.tags.filter((t) => !isManagedTag(t));
	const nextTags = [...preserved, ...selectedTags];
	const block = tagsBlock(nextTags, parsed.tagsIndent);

	if (!parsed.hasFrontmatter) {
		return `---\n${block}\n---\n\n${parsed.body}`;
	}

	let fm = parsed.frontmatter;
	if (parsed.tagsRange) {
		fm = fm.slice(0, parsed.tagsRange.start) + block + fm.slice(parsed.tagsRange.end);
	} else {
		fm = fm.length ? `${fm}\n${block}` : block;
	}
	return `---\n${fm}\n---\n${parsed.body}`;
}

/** Added / removed tags between the loaded README and the regenerated one. */
export function diffTags(
	original: string[],
	next: string[]
): { added: string[]; removed: string[] } {
	const o = new Set(original);
	const n = new Set(next);
	return {
		added: next.filter((t) => !o.has(t)),
		removed: original.filter((t) => !n.has(t))
	};
}