import * as path from "path";
import { ImageMarker } from "../cache/WebpageCache";
import { ImageDownloader } from "../image_search/utils/ImageDownloader";
import { Logger } from "../utils/Logger";

export const MAX_MAIN_CONTENT_IMAGES = 8;

export interface ImageExtractionResult {
  markers: ImageMarker[];
  foundCount: number;
  filteredCount: number;
}

export interface ImageProcessOutcome {
  updatedMarkdown: string;
  downloadedCount: number;
  downloadedFilenames: string[];
}

const REJECT_KEYWORDS = [
  "logo",
  "icon",
  "favicon",
  "avatar",
  "emoji",
  "button",
  "badge",
  "banner",
  "social",
  "share",
  "ad",
  "advertisement",
  "pixel",
  "spacer",
  "tracker",
  "widget",
  "header",
  "footer",
  "nav",
  "profile",
  "thumb-small",
  "spinner",
  "loading",
];

/**
 * Derives a stable, clean human-readable slug from alt text or URL basename.
 * Handles filename collision deduplication using an active set.
 */
export function generateStableFilename(
  srcUrl: string,
  altOrCaption: string | undefined,
  existingFilenames: Set<string>
): string {
  let ext = ".webp";
  try {
    const parsedUrl = new URL(srcUrl);
    const urlExt = path.extname(parsedUrl.pathname).toLowerCase();
    if ([".jpg", ".jpeg", ".png", ".gif", ".webp", ".avif", ".bmp"].includes(urlExt)) {
      ext = urlExt === ".jpeg" ? ".jpg" : urlExt;
    }
  } catch {
    // Ignore URL parse error
  }

  let slug = "";
  if (altOrCaption && altOrCaption.trim()) {
    slug = altOrCaption
      .trim()
      .toLowerCase()
      .replace(/[^a-z0-9]/g, "_")
      .replace(/_+/g, "_");
    if (slug.startsWith("_")) slug = slug.substring(1);
    if (slug.endsWith("_")) slug = slug.substring(0, slug.length - 1);
    if (slug.length > 30) slug = slug.substring(0, 30);
  }

  if (!slug) {
    try {
      const parsedUrl = new URL(srcUrl);
      const base = path.basename(parsedUrl.pathname, path.extname(parsedUrl.pathname));
      slug = base.toLowerCase().replace(/[^a-z0-9]/g, "_").replace(/_+/g, "_");
    } catch {
      // Ignore
    }
  }

  if (!slug || slug.length < 3 || ["image", "photo", "picture", "file", "img"].includes(slug)) {
    slug = "content_image";
  }

  let candidate = `${slug}${ext}`;
  let counter = 2;
  while (existingFilenames.has(candidate)) {
    candidate = `${slug}_${counter}${ext}`;
    counter++;
  }

  existingFilenames.add(candidate);
  return candidate;
}

/**
 * Scans DOM document for main content images strictly within <article>, <main>, or <figure>.
 * Filters out decorative icons, logos, trackers, and ads.
 * Assigns insertionIndex based on preceding content block elements.
 */
export function extractMainContentImageMarkers(
  doc: Document,
  baseUrl: string
): ImageExtractionResult {
  const DOCUMENT_POSITION_PRECEDING =
    doc.defaultView?.Node.DOCUMENT_POSITION_PRECEDING ?? 2;

  const mainContainers = Array.from(doc.querySelectorAll("article, main, .post-content, .article-content, #content"));
  const container = mainContainers[0] || doc.body;

  if (!container) {
    return { markers: [], foundCount: 0, filteredCount: 0 };
  }

  const allImgElements = Array.from(container.querySelectorAll("img"));
  const foundCount = allImgElements.length;

  // Gather all structural content blocks in container for insertionIndex calculation
  const contentBlocks = Array.from(
    container.querySelectorAll("h1, h2, h3, h4, h5, h6, p, ul, ol, table, blockquote, pre")
  );

  const markers: ImageMarker[] = [];
  let filteredCount = 0;

  for (const img of allImgElements) {
    const src = img.getAttribute("src") || img.getAttribute("data-src");
    if (!src || src.startsWith("data:") || src.startsWith("blob:")) {
      filteredCount++;
      continue;
    }

    let absoluteUrl: string;
    try {
      absoluteUrl = new URL(src, baseUrl).href;
    } catch {
      filteredCount++;
      continue;
    }

    // Check URL keywords for rejection
    const lowerUrl = absoluteUrl.toLowerCase();
    if (REJECT_KEYWORDS.some((kw) => lowerUrl.includes(kw))) {
      filteredCount++;
      continue;
    }

    // Check class names / IDs for rejection
    const classAndId = `${img.className} ${img.id}`.toLowerCase();
    if (REJECT_KEYWORDS.some((kw) => classAndId.includes(kw))) {
      filteredCount++;
      continue;
    }

    // Dimension check (ignore tiny icons < 50px)
    const widthAttr = parseInt(img.getAttribute("width") || "0", 10);
    const heightAttr = parseInt(img.getAttribute("height") || "0", 10);
    if ((widthAttr > 0 && widthAttr < 50) || (heightAttr > 0 && heightAttr < 50)) {
      filteredCount++;
      continue;
    }

    const alt = img.getAttribute("alt")?.trim();
    const title = img.getAttribute("title")?.trim();

    // Check figure caption
    const figure = img.closest("figure");
    const captionNode = figure?.querySelector("figcaption");
    const caption = captionNode?.textContent?.trim();

    const cleanAlt = alt && !["image", "photo", "picture", "icon"].includes(alt.toLowerCase()) ? alt : undefined;
    const cleanCaption = caption || title || undefined;

    // Reject if no alt, caption, or title description available
    if (!cleanAlt && !cleanCaption) {
      filteredCount++;
      continue;
    }

    // Calculate insertionIndex: count content blocks preceding this image in DOM
    let insertionIndex = 0;
    for (const block of contentBlocks) {
      if (img.compareDocumentPosition(block) & DOCUMENT_POSITION_PRECEDING) {
        insertionIndex++;
      } else {
        break;
      }
    }

    markers.push({
      src: absoluteUrl,
      alt: cleanAlt,
      caption: cleanCaption,
      insertionIndex,
    });

    if (markers.length >= MAX_MAIN_CONTENT_IMAGES) {
      break;
    }
  }

  return {
    markers,
    foundCount,
    filteredCount: filteredCount + Math.max(0, allImgElements.length - filteredCount - markers.length),
  };
}


/**
 * Downloads main content images and inserts inline semantic image blocks into clean Markdown body.
 * Formats image blocks using the required user format:
 *
 * <img file="college_sorted.webp">
 * Description of image...
 */
export async function downloadAndInsertImages(
  markdown: string,
  markers: ImageMarker[],
  workingDir: string,
  downloader?: ImageDownloader,
  maxImages: number = MAX_MAIN_CONTENT_IMAGES
): Promise<ImageProcessOutcome> {
  if (!markers || markers.length === 0 || !markdown) {
    return {
      updatedMarkdown: markdown,
      downloadedCount: 0,
      downloadedFilenames: [],
    };
  }

  const activeDownloader = downloader ?? new ImageDownloader();
  const existingFilenames = new Set<string>();
  const downloadedFilenames: string[] = [];
  const selectedMarkers = markers.slice(0, maxImages);

  // Split markdown into structural content blocks (by double newlines)
  const blocks = markdown.split(/\n\s*\n/);
  const blockImageMap = new Map<number, string[]>();

  for (let i = 0; i < selectedMarkers.length; i++) {
    const marker = selectedMarkers[i];
    const targetFilename = generateStableFilename(marker.src, marker.alt || marker.caption, existingFilenames);

    try {
      const downloadRes = await activeDownloader.downloadSingleImage(
        marker.src,
        path.basename(targetFilename, path.extname(targetFilename)),
        workingDir
      );

      if (downloadRes) {
        downloadedFilenames.push(downloadRes.filename);
        const rawDesc = marker.alt || marker.caption || "Main content image";
        const safeDesc = rawDesc
          .replace(/\\/g, "\\\\")
          .replace(/\[/g, "\\[")
          .replace(/\]/g, "\\]")
          .replace(/\s+/g, " ")
          .trim();

        const imageBlock = `![${safeDesc}](${downloadRes.filename})`;

        // Map image block to target insertionIndex
        const targetIndex = Math.min(marker.insertionIndex, blocks.length);
        const list = blockImageMap.get(targetIndex) || [];
        list.push(imageBlock);
        blockImageMap.set(targetIndex, list);
      }
    } catch (err) {
      Logger.debug("[WebpageReader]", `Failed to download image ${marker.src}: ${err}`);
    }
  }

  if (downloadedFilenames.length === 0) {
    return {
      updatedMarkdown: markdown,
      downloadedCount: 0,
      downloadedFilenames: [],
    };
  }

  // Re-assemble markdown blocks with inline image blocks inserted at exact positions
  const resultBlocks: string[] = [];
  for (let idx = 0; idx <= blocks.length; idx++) {
    if (blockImageMap.has(idx)) {
      resultBlocks.push(...blockImageMap.get(idx)!);
    }
    if (idx < blocks.length) {
      resultBlocks.push(blocks[idx]);
    }
  }

  return {
    updatedMarkdown: resultBlocks.join("\n\n"),
    downloadedCount: downloadedFilenames.length,
    downloadedFilenames,
  };
}

import * as path from "path";
import { ImageMarker } from "../cache/WebpageCache";
import { ImageDownloader } from "../image_search/utils/ImageDownloader";
import { Logger } from "../utils/Logger";

export const MAX_MAIN_CONTENT_IMAGES = 8;

export interface ImageExtractionResult {
  markers: ImageMarker[];
  foundCount: number;
  filteredCount: number;
}

export interface ImageProcessOutcome {
  updatedMarkdown: string;
  downloadedCount: number;
  downloadedFilenames: string[];
}

const REJECT_KEYWORDS = [
  "logo",
  "icon",
  "favicon",
  "avatar",
  "emoji",
  "button",
  "badge",
  "banner",
  "social",
  "share",
  "ad",
  "advertisement",
  "pixel",
  "spacer",
  "tracker",
  "widget",
  "header",
  "footer",
  "nav",
  "profile",
  "thumb-small",
  "spinner",
  "loading",
];

/**
 * Derives a stable, clean human-readable slug from alt text or URL basename.
 * Handles filename collision deduplication using an active set.
 */
export function generateStableFilename(
  srcUrl: string,
  altOrCaption: string | undefined,
  existingFilenames: Set<string>
): string {
  let ext = ".webp";
  try {
    const parsedUrl = new URL(srcUrl);
    const urlExt = path.extname(parsedUrl.pathname).toLowerCase();
    if ([".jpg", ".jpeg", ".png", ".gif", ".webp", ".avif", ".bmp"].includes(urlExt)) {
      ext = urlExt === ".jpeg" ? ".jpg" : urlExt;
    }
  } catch {
    // Ignore URL parse error
  }

  let slug = "";
  if (altOrCaption && altOrCaption.trim()) {
    slug = altOrCaption
      .trim()
      .toLowerCase()
      .replace(/[^a-z0-9]/g, "_")
      .replace(/_+/g, "_");
    if (slug.startsWith("_")) slug = slug.substring(1);
    if (slug.endsWith("_")) slug = slug.substring(0, slug.length - 1);
    if (slug.length > 30) slug = slug.substring(0, 30);
  }

  if (!slug) {
    try {
      const parsedUrl = new URL(srcUrl);
      const base = path.basename(parsedUrl.pathname, path.extname(parsedUrl.pathname));
      slug = base.toLowerCase().replace(/[^a-z0-9]/g, "_").replace(/_+/g, "_");
    } catch {
      // Ignore
    }
  }

  if (!slug || slug.length < 3 || ["image", "photo", "picture", "file", "img"].includes(slug)) {
    slug = "content_image";
  }

  let candidate = `${slug}${ext}`;
  let counter = 2;
  while (existingFilenames.has(candidate)) {
    candidate = `${slug}_${counter}${ext}`;
    counter++;
  }

  existingFilenames.add(candidate);
  return candidate;
}

/**
 * Scans DOM document for main content images strictly within <article>, <main>, or <figure>.
 * Filters out decorative icons, logos, trackers, and ads.
 * Assigns insertionIndex based on preceding content block elements.
 */
export function extractMainContentImageMarkers(
  doc: Document,
  baseUrl: string
): ImageExtractionResult {
  const DOCUMENT_POSITION_PRECEDING =
    doc.defaultView?.Node.DOCUMENT_POSITION_PRECEDING ?? 2;

  const mainContainers = Array.from(doc.querySelectorAll("article, main, .post-content, .article-content, #content"));
  const container = mainContainers[0] || doc.body;

  if (!container) {
    return { markers: [], foundCount: 0, filteredCount: 0 };
  }

  const allImgElements = Array.from(container.querySelectorAll("img"));
  const foundCount = allImgElements.length;

  // Gather all structural content blocks in container for insertionIndex calculation
  const contentBlocks = Array.from(
    container.querySelectorAll("h1, h2, h3, h4, h5, h6, p, ul, ol, table, blockquote, pre")
  );

  const markers: ImageMarker[] = [];
  let filteredCount = 0;

  for (const img of allImgElements) {
    const src = img.getAttribute("src") || img.getAttribute("data-src");
    if (!src || src.startsWith("data:") || src.startsWith("blob:")) {
      filteredCount++;
      continue;
    }

    let absoluteUrl: string;
    try {
      absoluteUrl = new URL(src, baseUrl).href;
    } catch {
      filteredCount++;
      continue;
    }

    // Check URL keywords for rejection
    const lowerUrl = absoluteUrl.toLowerCase();
    if (REJECT_KEYWORDS.some((kw) => lowerUrl.includes(kw))) {
      filteredCount++;
      continue;
    }

    // Check class names / IDs for rejection
    const classAndId = `${img.className} ${img.id}`.toLowerCase();
    if (REJECT_KEYWORDS.some((kw) => classAndId.includes(kw))) {
      filteredCount++;
      continue;
    }

    // Dimension check (ignore tiny icons < 50px)
    const widthAttr = parseInt(img.getAttribute("width") || "0", 10);
    const heightAttr = parseInt(img.getAttribute("height") || "0", 10);
    if ((widthAttr > 0 && widthAttr < 50) || (heightAttr > 0 && heightAttr < 50)) {
      filteredCount++;
      continue;
    }

    const alt = img.getAttribute("alt")?.trim();
    const title = img.getAttribute("title")?.trim();

    // Check figure caption
    const figure = img.closest("figure");
    const captionNode = figure?.querySelector("figcaption");
    const caption = captionNode?.textContent?.trim();

    const cleanAlt = alt && !["image", "photo", "picture", "icon"].includes(alt.toLowerCase()) ? alt : undefined;
    const cleanCaption = caption || title || undefined;

    // Reject if no alt, caption, or title description available
    if (!cleanAlt && !cleanCaption) {
      filteredCount++;
      continue;
    }

    // Calculate insertionIndex: count content blocks preceding this image in DOM
    let insertionIndex = 0;
    for (const block of contentBlocks) {
      if (img.compareDocumentPosition(block) & DOCUMENT_POSITION_PRECEDING) {
        insertionIndex++;
      } else {
        break;
      }
    }

    markers.push({
      src: absoluteUrl,
      alt: cleanAlt,
      caption: cleanCaption,
      insertionIndex,
    });

    if (markers.length >= MAX_MAIN_CONTENT_IMAGES) {
      break;
    }
  }

  return {
    markers,
    foundCount,
    filteredCount: filteredCount + Math.max(0, allImgElements.length - filteredCount - markers.length),
  };
}


/**
 * Downloads main content images and inserts inline semantic image blocks into clean Markdown body.
 * Formats image blocks using the required user format:
 *
 * <img file="college_sorted.webp">
 * Description of image...
 */
export async function downloadAndInsertImages(
  markdown: string,
  markers: ImageMarker[],
  workingDir: string,
  downloader?: ImageDownloader,
  maxImages: number = MAX_MAIN_CONTENT_IMAGES
): Promise<ImageProcessOutcome> {
  if (!markers || markers.length === 0 || !markdown) {
    return {
      updatedMarkdown: markdown,
      downloadedCount: 0,
      downloadedFilenames: [],
    };
  }

  const activeDownloader = downloader ?? new ImageDownloader();
  const existingFilenames = new Set<string>();
  const downloadedFilenames: string[] = [];
  const selectedMarkers = markers.slice(0, maxImages);

  // Split markdown into structural content blocks (by double newlines)
  const blocks = markdown.split(/\n\s*\n/);
  const blockImageMap = new Map<number, string[]>();

  for (let i = 0; i < selectedMarkers.length; i++) {
    const marker = selectedMarkers[i];
    const targetFilename = generateStableFilename(marker.src, marker.alt || marker.caption, existingFilenames);

    try {
      const downloadRes = await activeDownloader.downloadSingleImage(
        marker.src,
        path.basename(targetFilename, path.extname(targetFilename)),
        workingDir
      );

      if (downloadRes) {
        downloadedFilenames.push(downloadRes.filename);
        const rawDesc = marker.alt || marker.caption || "Main content image";
        const safeDesc = rawDesc
          .replace(/\\/g, "\\\\")
          .replace(/\[/g, "\\[")
          .replace(/\]/g, "\\]")
          .replace(/\s+/g, " ")
          .trim();

        const imageBlock = `![${safeDesc}](${downloadRes.filename})`;

        // Map image block to target insertionIndex
        const targetIndex = Math.min(marker.insertionIndex, blocks.length);
        const list = blockImageMap.get(targetIndex) || [];
        list.push(imageBlock);
        blockImageMap.set(targetIndex, list);
      }
    } catch (err) {
      Logger.debug("[WebpageReader]", `Failed to download image ${marker.src}: ${err}`);
    }
  }

  if (downloadedFilenames.length === 0) {
    return {
      updatedMarkdown: markdown,
      downloadedCount: 0,
      downloadedFilenames: [],
    };
  }

  // Re-assemble markdown blocks with inline image blocks inserted at exact positions
  const resultBlocks: string[] = [];
  for (let idx = 0; idx <= blocks.length; idx++) {
    if (blockImageMap.has(idx)) {
      resultBlocks.push(...blockImageMap.get(idx)!);
    }
    if (idx < blocks.length) {
      resultBlocks.push(blocks[idx]);
    }
  }

  return {
    updatedMarkdown: resultBlocks.join("\n\n"),
    downloadedCount: downloadedFilenames.length,
    downloadedFilenames,
  };
}