// Renders MinerU's content_list_v2.json (already page-grouped: one array per
// PDF page) into per-page Markdown. This is a faithful re-render from MinerU's
// structured extraction, not a split of full.md — full.md is a flat, furniture-
// stripped concatenation that cannot be reliably realigned to page boundaries.
//
// Page furniture (page_header / page_footer / page_number / page_footnote) is
// dropped to match full.md's clean body. Unknown item types are dropped.

interface Span {
  type?: string;
  content?: string;
}

interface ContentItem {
  type?: string;
  content?: any;
}

function renderSpans(spans: Span[] | undefined): string {
  if (!Array.isArray(spans)) return "";
  return spans
    .map((s) => (s?.type === "equation_inline" ? `$${s.content ?? ""}$` : (typeof s?.content === "string" ? s.content : "")))
    .join("");
}

function renderItem(item: ContentItem): string | null {
  const c = item?.content ?? {};
  switch (item?.type) {
    case "title": {
      const level = Math.min(Math.max(Number(c.level) || 1, 1), 6);
      const text = renderSpans(c.title_content).trim();
      return text ? `${"#".repeat(level)} ${text}` : null;
    }
    case "paragraph": {
      return renderSpans(c.paragraph_content).trim() || null;
    }
    case "equation_interline": {
      const math = String(c.math_content ?? "").trim();
      return math ? `$$\n${math}\n$$` : null;
    }
    case "table": {
      const caption = renderSpans(c.table_caption).trim();
      const html = String(c.html ?? "").trim();
      if (!html && !caption) return null;
      return [caption, html].filter(Boolean).join("\n\n");
    }
    case "image": {
      // Images are not extracted from MinerU's zip yet (deferred by design), so
      // this reference points at an images/<hash>.jpg entry that is NOT written
      // to disk — the link is intentionally dangling for now.
      // FUTURE: extract the zip's images/ entries next to the page files and
      // rewrite this path to a relative on-disk location (e.g. ./images/<hash>.jpg)
      // to make the Markdown self-contained. Source field: content.image_source.path.
      const path = String(c.image_source?.path ?? "");
      const caption = renderSpans(c.image_caption).trim();
      return `![${caption}](${path})`;
    }
    case "list": {
      const ordered = c.attribute === "ordered";
      const items: any[] = Array.isArray(c.list_items) ? c.list_items : [];
      const lines = items
        .map((li, i) => {
          const text = renderSpans(li?.item_content).trim();
          if (!text) return null;
          return ordered ? `${i + 1}. ${text}` : `- ${text}`;
        })
        .filter((l): l is string => !!l);
      return lines.length ? lines.join("\n") : null;
    }
    case "algorithm": {
      const caption = renderSpans(c.algorithm_caption).trim();
      const body = renderSpans(c.algorithm_content).trim();
      if (!body) return null;
      return [caption, "```\n" + body + "\n```"].filter(Boolean).join("\n\n");
    }
    default:
      // page_header / page_footer / page_number / page_footnote and any unknown
      // type are dropped.
      return null;
  }
}

export interface RenderedPage {
  name: string;
  markdown: string;
}

export interface RenderedPages {
  pages: RenderedPage[];
  index: string;
  pageCount: number;
}

export function renderContentListPages(v2: unknown): RenderedPages {
  const groups: unknown[] = Array.isArray(v2) ? v2 : [];
  const pad = Math.max(String(groups.length).length, 2);

  const pages: RenderedPage[] = groups.map((group, idx) => {
    const items: ContentItem[] = Array.isArray(group) ? group : [];
    const blocks = items.map(renderItem).filter((b): b is string => !!b);
    return {
      name: `page-${String(idx + 1).padStart(pad, "0")}.md`,
      markdown: blocks.join("\n\n") + "\n",
    };
  });

  const index = ["# Index", "", ...pages.map((p, i) => `${i + 1}. [Page ${i + 1}](./${p.name})`), ""].join("\n");

  return { pages, index, pageCount: pages.length };
}