path: root/src/loaders/pleroma.ts

import type { Loader } from "astro/loaders";
import { marked } from "marked";
import TurndownService from "turndown";

interface Logger {
	info: (message: string) => void;
	warn: (message: string) => void;
}

interface PleromaFeedConfig {
	instanceUrl: string;
	username: string;
	maxPosts?: number;
	accountId?: string; // Optional: if provided, skips account lookup
	allowedTags?: string[]; // Optional: if provided, only posts with these tags are included
	mergeThreads?: boolean; // Optional: if true, merges thread posts into single entry (default: true)
}

interface PleromaAccount {
	id: string;
	username: string;
	acct: string;
	display_name: string;
	url: string;
}

interface PleromaMediaAttachment {
	id: string;
	type: "image" | "video" | "gifv" | "audio" | "unknown";
	url: string;
	preview_url: string;
	description?: string;
}

interface PleromaStatus {
	id: string;
	created_at: string;
	content: string;
	url: string;
	reblog: PleromaStatus | null;
	in_reply_to_id: string | null;
	sensitive: boolean;
	media_attachments: PleromaMediaAttachment[];
	visibility: string;
	account: PleromaAccount;
}

/**
 * Detect if a post is a thread starter by checking for thread markers
 * Matches patterns like: 🧵, 👇, ⬇️, 1/n, (1/n), [1/n], Thread:, etc.
 */
function isThreadStarter(content: string): boolean {
	// Check for thread emojis
	const threadEmojis = ["🧵", "👇", "⬇️", "📝", "📖", "⤵️", "🔽"];
	if (threadEmojis.some((emoji) => content.includes(emoji))) {
		return true;
	}

	// Check for numbered thread patterns:
	// - 1/n, 1/*, 1/2, 1/10 (plain)
	// - (1/n), (1/*), (1/2) (parentheses)
	// - [1/n], [1/*], [1/2] (brackets)
	const numberedPatterns = [
		/\b1\/([n*]|\d+)\b/i, // 1/n, 1/*, 1/2
		/\(1\/([n*]|\d+)\)/i, // (1/n), (1/*)
		/\[1\/([n*]|\d+)\]/i, // [1/n], [1/*]
	];

	if (numberedPatterns.some((pattern) => pattern.test(content))) {
		return true;
	}

	// Check for text markers (case insensitive)
	const textMarkers = [
		/\bthread:/i, // Thread:
		/\[thread\]/i, // [Thread]
		/^thread about/i, // Thread about... (start of text)
		/^a thread about/i, // A thread about...
	];

	if (textMarkers.some((pattern) => pattern.test(content))) {
		return true;
	}

	return false;
}

/**
 * Parse the Link header to extract the max_id for the next page
 * Link header format: <url?max_id=123>; rel="next", <url?since_id=456>; rel="prev"
 */
function parseNextPageMaxId(linkHeader: string | null): string | null {
	if (!linkHeader) {
		return null;
	}

	// Split by comma to get individual links
	const links = linkHeader.split(",");

	for (const link of links) {
		// Check if this is the "next" rel link
		if (link.includes('rel="next"')) {
			// Extract URL from angle brackets
			const urlMatch = link.match(/<([^>]+)>/);
			if (urlMatch?.[1]) {
				// Parse the URL to extract max_id parameter
				try {
					const url = new URL(urlMatch[1]);
					const maxId = url.searchParams.get("max_id");
					return maxId;
				} catch {}
			}
		}
	}

	return null;
}

/**
 * Fetch the context (ancestors and descendants) for a given status
 * Returns only the descendants array for thread building
 */
async function fetchStatusContext(
	instanceUrl: string,
	statusId: string,
	logger: Logger,
): Promise<PleromaStatus[]> {
	try {
		const contextUrl = `${instanceUrl}/api/v1/statuses/${statusId}/context`;
		logger.info(`Fetching context for status: ${statusId}`);

		const controller = new AbortController();
		const timeoutId = setTimeout(() => controller.abort(), 10000);

		const response = await fetch(contextUrl, {
			headers: {
				"User-Agent": "Astro Blog (pleroma-loader)",
			},
			signal: controller.signal,
		});

		clearTimeout(timeoutId);

		if (!response.ok) {
			logger.warn(`Failed to fetch context: HTTP ${response.status}`);
			return [];
		}

		const context: { ancestors: PleromaStatus[]; descendants: PleromaStatus[] } =
			await response.json();
		logger.info(`Fetched ${context.descendants.length} descendants for status ${statusId}`);
		return context.descendants;
	} catch (error) {
		logger.warn(`Failed to fetch status context: ${error}`);
		return [];
	}
}

/**
 * Build a direct author-to-author reply chain from the thread starter
 * Stops when encountering a reply from another user or a missing link
 */
function buildAuthorChain(starter: PleromaStatus, descendants: PleromaStatus[]): PleromaStatus[] {
	const chain: PleromaStatus[] = [starter];
	const authorAccountId = starter.account.id;
	let currentId = starter.id;

	// Keep following the chain as long as we find direct author replies
	while (true) {
		// Find the next post in the chain: it must be by the same author and reply to the current post
		const nextPost = descendants.find(
			(status) => status.in_reply_to_id === currentId && status.account.id === authorAccountId,
		);

		if (!nextPost) {
			// No more direct author replies found, chain ends here
			break;
		}

		chain.push(nextPost);
		currentId = nextPost.id;
	}

	return chain;
}

/**
 * Strip thread markers from content (1/n, 2/n, 3/4, etc.)
 */
function stripThreadMarkers(content: string): string {
	return content
		.replace(/\s*\d+\/[n*\d]+\s*/gi, " ")
		.replace(/🧵/g, "")
		.trim();
}

/**
 * Extract trailing hashtags from content
 * Handles both hashtag-only lines and hashtags at the end of text lines
 * Returns the main content without trailing tags and the extracted tags
 */
function extractTrailingHashtags(content: string): {
	mainContent: string;
	tags: string[];
} {
	const tags: string[] = [];
	let modifiedContent = content;

	// Regex patterns
	const hashtagOnlyLine = /^\s*((\[#\w+\]\([^)]+\)|#\w+)\s*)+\s*$/;
	const trailingHashtags = /((?:\s*(?:\[#\w+\]\([^)]+\)|#\w+))+)\s*$/;
	const hashtagExtract = /\[#(\w+)\]\([^)]+\)|#(\w+)/g;

	// First, handle hashtag-only lines at the end
	const lines = modifiedContent.split("\n");
	while (lines.length > 0) {
		const lastLine = lines[lines.length - 1]?.trim() || "";
		if (!lastLine) {
			lines.pop(); // Remove empty trailing lines
			continue;
		}
		if (hashtagOnlyLine.test(lastLine)) {
			// Extract tag names from this line
			let match: RegExpExecArray | null = hashtagExtract.exec(lastLine);
			while (match !== null) {
				const tag = match[1] || match[2];
				if (tag) {
					tags.push(tag.toLowerCase());
				}
				match = hashtagExtract.exec(lastLine);
			}
			hashtagExtract.lastIndex = 0;
			lines.pop();
		} else {
			break;
		}
	}
	modifiedContent = lines.join("\n");

	// Second, handle trailing hashtags at the end of the last line (even if there's other text)
	const trailingMatch = modifiedContent.match(trailingHashtags);
	if (trailingMatch?.[1]) {
		const trailingText = trailingMatch[1];
		// Extract tag names from trailing hashtags
		let match: RegExpExecArray | null = hashtagExtract.exec(trailingText);
		while (match !== null) {
			const tag = match[1] || match[2];
			if (tag) {
				tags.push(tag.toLowerCase());
			}
			match = hashtagExtract.exec(trailingText);
		}
		hashtagExtract.lastIndex = 0;
		// Remove trailing hashtags from content
		modifiedContent = modifiedContent.replace(trailingHashtags, "").trim();
	}

	return {
		mainContent: modifiedContent,
		tags: [...new Set(tags)], // Deduplicate within this segment
	};
}

/**
 * Merge thread posts into a single content structure with image grids per segment
 */
function mergeThreadContent(chain: PleromaStatus[]): {
	content: string;
	attachments: Array<{ url: string; type: string }>;
} {
	const segments: string[] = [];
	const allAttachments: Array<{ url: string; type: string }> = [];
	const allTags = new Set<string>(); // Collect all tags from all segments

	for (const post of chain) {
		// Clean and strip thread markers from content
		const cleanedContent = cleanContent(post.content || "");
		const contentWithoutMarkers = stripThreadMarkers(cleanedContent);

		// Extract trailing hashtags from content
		const { mainContent, tags } = extractTrailingHashtags(contentWithoutMarkers);
		tags.forEach((tag) => allTags.add(tag));

		// Build segment with text (without trailing hashtags)
		let segment = mainContent;

		// Add image attachments as HTML grid after the text
		const imageAttachments = post.media_attachments.filter(
			(attachment) => attachment.type === "image",
		);

		if (imageAttachments.length > 0) {
			// Build HTML grid for images
			const imageGrid = `
<div class="mt-4 mb-4 grid grid-cols-1 gap-4 sm:grid-cols-2">
${imageAttachments
	.map((attachment) => {
		const description = attachment.description || "Image";
		allAttachments.push({
			url: attachment.url,
			type: `image/${attachment.url.split(".").pop() || "jpeg"}`,
		});
		return `<a href="${attachment.url}" target="_blank" rel="noopener noreferrer" class="block overflow-hidden rounded-lg border border-gray-200 transition-colors hover:border-gray-300 dark:border-gray-700 dark:hover:border-gray-600">
<img src="${attachment.url}" alt="${description}" class="h-48 w-full object-cover" loading="lazy" />
</a>`;
	})
	.join("\n")}
</div>`;

			segment = `${segment}\n\n${imageGrid}`;
		}

		segments.push(segment);
	}

	// Join segments with horizontal rule separator
	let content = segments.join("\n\n---\n\n");

	// Append consolidated tags at the end as markdown links
	if (allTags.size > 0) {
		// Get the instance URL from the first post to construct tag URLs
		const instanceUrl = chain[0]?.account.url.split("/@")[0] || "https://social.craftknight.com";
		const tagLine = [...allTags].map((t) => `[#${t}](${instanceUrl}/tag/${t})`).join(" ");
		content = `${content}\n\n${tagLine}`;
	}

	return { content, attachments: [] }; // Return empty attachments to avoid duplicate grid at end
}

async function getAccountId(
	instanceUrl: string,
	username: string,
	logger: Logger,
): Promise<string | null> {
	try {
		const searchUrl = `${instanceUrl}/api/v1/accounts/search?q=${encodeURIComponent(username)}&limit=1`;
		logger.info(`Looking up account ID for username: ${username}`);

		const controller = new AbortController();
		const timeoutId = setTimeout(() => controller.abort(), 10000);

		const response = await fetch(searchUrl, {
			headers: {
				"User-Agent": "Astro Blog (pleroma-loader)",
			},
			signal: controller.signal,
		});

		clearTimeout(timeoutId);

		if (!response.ok) {
			logger.warn(`Failed to search for account: HTTP ${response.status}`);
			return null;
		}

		const accounts: PleromaAccount[] = await response.json();

		if (accounts.length === 0 || !accounts[0]) {
			logger.warn(`No account found for username: ${username}`);
			return null;
		}

		const account = accounts[0];
		logger.info(`Found account ID: ${account.id} for @${account.acct}`);
		return account.id;
	} catch (error) {
		logger.warn(`Failed to lookup account ID: ${error}`);
		return null;
	}
}

async function fetchAccountStatuses(
	instanceUrl: string,
	accountId: string,
	maxPosts: number,
	logger: Logger,
): Promise<PleromaStatus[]> {
	const allStatuses: PleromaStatus[] = [];
	let maxId: string | null = null;
	let pageCount = 0;
	const pageLimit = 40; // Mastodon/Pleroma API max per page
	const fetchAll = maxPosts === -1;

	// Fetch pages until we have enough posts or no more pages available
	while (fetchAll || allStatuses.length < maxPosts) {
		pageCount++;
		let response: Response | undefined;
		let lastError: unknown;

		// Build URL with pagination parameters
		// If fetching all, always use pageLimit; otherwise calculate remaining
		const requestLimit = fetchAll ? pageLimit : Math.min(pageLimit, maxPosts - allStatuses.length);
		const params = new URLSearchParams({
			limit: String(requestLimit),
			exclude_replies: "true",
			exclude_reblogs: "true",
		});

		if (maxId) {
			params.set("max_id", maxId);
		}

		const statusesUrl = `${instanceUrl}/api/v1/accounts/${accountId}/statuses?${params.toString()}`;

		// Add retry logic for network issues
		for (let attempt = 1; attempt <= 3; attempt++) {
			try {
				const modeMsg = fetchAll ? " [fetching all posts]" : ` [target: ${maxPosts}]`;
				logger.info(
					`Attempt ${attempt} to fetch statuses page ${pageCount}${maxId ? ` (max_id: ${maxId})` : ""}${modeMsg}...`,
				);

				// Create timeout controller
				const controller = new AbortController();
				const timeoutId = setTimeout(() => controller.abort(), 10000);

				response = await fetch(statusesUrl, {
					headers: {
						"User-Agent": "Astro Blog (pleroma-loader)",
					},
					signal: controller.signal,
				});

				clearTimeout(timeoutId);

				if (response.ok) {
					break; // Success, exit retry loop
				}
				throw new Error(`HTTP ${response.status}: ${response.statusText}`);
			} catch (error) {
				lastError = error;
				logger.warn(`Attempt ${attempt} failed: ${error}`);

				if (attempt < 3) {
					logger.info("Retrying in 2 seconds...");
					await new Promise((resolve) => setTimeout(resolve, 2000));
				}
			}
		}

		if (!response || !response.ok) {
			throw new Error(`Failed to fetch statuses after 3 attempts. Last error: ${lastError}`);
		}

		const statuses: PleromaStatus[] = await response.json();
		logger.info(`Fetched ${statuses.length} statuses from page ${pageCount}`);

		// If no statuses returned, we've reached the end
		if (statuses.length === 0) {
			logger.info("No more statuses available");
			break;
		}

		// Add statuses to our accumulated list
		allStatuses.push(...statuses);

		// Parse Link header to get next page max_id
		const linkHeader = response.headers.get("link");
		const nextMaxId = parseNextPageMaxId(linkHeader);

		if (!nextMaxId) {
			logger.info("No more pages available (no next link in header)");
			break;
		}

		// If the max_id hasn't changed, we're stuck in a loop - break
		if (nextMaxId === maxId) {
			logger.warn("Pagination returned same max_id, stopping to prevent infinite loop");
			break;
		}

		maxId = nextMaxId;
	}

	const summaryMsg = fetchAll
		? `Total fetched: ${allStatuses.length} statuses (all available) across ${pageCount} page(s)`
		: `Total fetched: ${allStatuses.length} statuses (target: ${maxPosts}) across ${pageCount} page(s)`;
	logger.info(summaryMsg);
	return allStatuses;
}

function isFilteredStatus(status: PleromaStatus): boolean {
	// Filter out boosts/reblogs (already handled by API parameter, but double-check)
	if (status.reblog) {
		return true;
	}

	// Filter out replies (already handled by API parameter, but double-check)
	if (status.in_reply_to_id) {
		return true;
	}

	// Filter out NSFW/sensitive content
	if (status.sensitive) {
		return true;
	}

	return false;
}

function extractHashtags(htmlContent: string): string[] {
	// Extract hashtags from HTML spans and plain text
	const hashtagPattern = /#(\w+)/gi;
	const matches = htmlContent.match(hashtagPattern);
	return matches ? [...new Set(matches.map((tag) => tag.toLowerCase()))] : [];
}

function hasAllowedTag(status: PleromaStatus, allowedTags: string[]): boolean {
	if (!allowedTags || allowedTags.length === 0) {
		return true; // No filtering if no tags specified
	}

	const content = status.content || "";
	const hashtags = extractHashtags(content);
	const normalizedAllowedTags = allowedTags.map((tag) => tag.toLowerCase().replace(/^#/, ""));
	const normalizedHashtags = hashtags.map((tag) => tag.toLowerCase().replace(/^#/, ""));

	return normalizedHashtags.some((tag) => normalizedAllowedTags.includes(tag));
}

function cleanContent(htmlContent: string): string {
	const turndownService = new TurndownService({
		headingStyle: "atx",
		codeBlockStyle: "fenced",
	});

	// Remove or replace common Pleroma/Mastodon elements
	const cleanedContent = htmlContent
		.replace(/<span class="[^"]*mention[^"]*"[^>]*>/gi, "") // Remove mention spans but keep content
		.replace(/<\/span>/gi, "")
		.replace(/<span class="[^"]*hashtag[^"]*"[^>]*>/gi, "") // Remove hashtag spans but keep content
		.replace(/<span class="[^"]*ellipsis[^"]*"[^>]*>.*?<\/span>/gi, "") // Remove ellipsis
		.replace(/<span class="[^"]*invisible[^"]*"[^>]*>.*?<\/span>/gi, ""); // Remove invisible text

	// Convert to markdown
	const markdown = turndownService.turndown(cleanedContent);

	// Clean up extra whitespace
	return markdown.trim().replace(/\n\s*\n\s*\n/g, "\n\n");
}

function markdownToHtml(markdown: string): string {
	// Configure marked options for safe rendering
	marked.setOptions({
		breaks: true, // Convert line breaks to <br>
		gfm: true, // GitHub flavored markdown
	});

	// Convert markdown to HTML
	const html = marked.parse(markdown);

	// Return as string (marked.parse can return string or Promise<string>)
	return typeof html === "string" ? html : "";
}

function extractTitle(content: string): string {
	// Extract first line or first sentence as title
	const firstLine = content.split("\n")[0];
	if (!firstLine) return "Micro post";

	const firstSentence = firstLine.split(/[.!?]/)[0];
	if (!firstSentence) return "Micro post";

	// Limit title length and clean it up
	const title = (firstSentence.length > 60 ? `${firstSentence.substring(0, 57)}...` : firstSentence)
		.replace(/[#*_`]/g, "") // Remove markdown formatting
		.trim();

	return title || "Micro post";
}

export function pleromaLoader(config: PleromaFeedConfig): Loader {
	return {
		name: "pleroma-loader",
		load: async ({ store, logger }) => {
			try {
				const { instanceUrl, username, maxPosts = 20, accountId: configAccountId } = config;

				logger.info(`Fetching Pleroma posts via API for user: ${username}`);

				// Get account ID (use provided one or lookup by username)
				let accountId: string | undefined = configAccountId;
				if (!accountId) {
					const lookedUpAccountId = await getAccountId(instanceUrl, username, logger);
					if (!lookedUpAccountId) {
						logger.warn("Failed to get account ID. Continuing without Pleroma posts...");
						store.clear();
						return;
					}
					accountId = lookedUpAccountId;
				}

				// Fetch statuses from API
				const statuses = await fetchAccountStatuses(instanceUrl, accountId, maxPosts, logger);
				logger.info(`Fetched ${statuses.length} statuses from API`);

				// Filter statuses
				const validStatuses = statuses.filter((status) => {
					if (isFilteredStatus(status)) return false;
					if (config.allowedTags && !hasAllowedTag(status, config.allowedTags)) return false;
					return true;
				});
				logger.info(`After filtering: ${validStatuses.length} valid posts`);

				// Clear existing entries
				store.clear();

				// Process each status
				for (const status of validStatuses) {
					try {
						const content = status.content || "";
						let cleanedContent: string;
						let attachments: Array<{ url: string; type: string }>;
						let postId: string;
						let sourceUrl: string;

						// Check if this is a thread starter and thread merging is enabled
						if (config.mergeThreads !== false && isThreadStarter(content)) {
							logger.info(`Detected thread starter: ${status.id}`);

							// Fetch context and build the author chain
							const descendants = await fetchStatusContext(instanceUrl, status.id, logger);
							const chain = buildAuthorChain(status, descendants);

							logger.info(`Built chain with ${chain.length} post(s) for thread ${status.id}`);

						// Merge thread content
						const merged = mergeThreadContent(chain);
						cleanedContent = merged.content;
						attachments = merged.attachments;
							postId = status.id;
							sourceUrl = status.url;
						} else {
							// Process as single post
							cleanedContent = cleanContent(content);
							postId = status.id;
							sourceUrl = status.url;

							// Extract image attachments only
							attachments = status.media_attachments
								.filter((attachment) => attachment.type === "image")
								.map((attachment) => ({
									url: attachment.url,
									type: `image/${attachment.url.split(".").pop() || "jpeg"}`,
									alt: attachment.description || undefined,
								}));
						}

						const title = extractTitle(cleanedContent);

						// Create note entry
						store.set({
							id: `pleroma-${postId}`,
							data: {
								title,
								description:
									cleanedContent.substring(0, 160) + (cleanedContent.length > 160 ? "..." : ""),
								publishDate: new Date(status.created_at),
								sourceUrl,
								attachments,
							},
							body: cleanedContent,
							rendered: {
								html: markdownToHtml(cleanedContent),
							},
						});

						logger.info(`Processed post: ${title.substring(0, 50)}...`);
					} catch (error) {
						logger.warn(`Failed to process status ${status.id}: ${error}`);
					}
				}

				logger.info(`Successfully loaded ${validStatuses.length} Pleroma posts`);
			} catch (error) {
				logger.warn(`Pleroma loader failed: ${error}`);
				logger.info("Continuing build without Pleroma posts...");
				// Don't throw error to prevent build failure
				store.clear();
			}
		},
	};
}