#!/usr/bin/env python3
"""
TI PDF datasheet → Markdown 转换工具（双栏 + 表格 + 书签标题）

用法:
  python3 pdf2md.py <input.pdf> [output.md]

特性:
- 双栏页面按 左栏→右栏 顺序重组正文（word 坐标 + 页面中线）
- 表格用 pdfplumber extract_tables() 转 Markdown 表格
- 标题层级取自 PDF 原始书签 (outline)，1-6 级嵌套
- 自动清理 TI 页眉页脚噪音行

依赖: pip3 install pdfplumber --break-system-packages
"""
import sys, re, os
import pdfplumber


def table_to_md(tbl):
    rows = []
    for r in tbl:
        cells = [(c or "").replace("\n", " ").strip() for c in r]
        rows.append(cells)
    rows = [r for r in rows if any(c for c in r)]
    if not rows:
        return ""
    ncols = max(len(r) for r in rows)
    norm = [r + [""] * (ncols - len(r)) for r in rows]
    md = ["| " + " | ".join(norm[0]) + " |", "|" + "---|" * ncols]
    for r in norm[1:]:
        md.append("| " + " | ".join(r) + " |")
    return "\n".join(md)


def in_bbox(w, bbox):
    x0, top, x1, bot = bbox
    return (w["x0"] >= x0 - 2 and w["x1"] <= x1 + 2
            and w["top"] >= top - 2 and w["bottom"] <= bot + 3)


def is_noise(s):
    if not s.strip():
        return True
    if "Copyright ©" in s or s.strip() == "Submit Document Feedback":
        return True
    if "Product Folder Links:" in s:
        return True
    if re.match(r"^SLUSEN\d? –", s) or re.match(r"^www\.ti\.com$", s):
        return True
    if re.match(r"^\d{1,3}$", s.strip()):
        return True
    return False


def convert(src, out):
    pdf = pdfplumber.open(src)

    # ---- outlines -> headers (page_no, title, level) ----
    try:
        outlines = list(pdf.doc.get_outlines())
    except Exception:
        outlines = []
    page_by_objid = {p.page_obj.pageid: n for n, p in enumerate(pdf.pages, 1)}
    headers = []
    for item in outlines:
        try:
            lvl, title, dest = item[0], item[1], item[2]
            if dest is None:
                continue
            pno = page_by_objid.get(dest[0].objid)
            if pno:
                headers.append((pno, title, lvl))
        except Exception:
            continue

    # ---- per page: text lines (column-aware) + tables ----
    page_data = {}
    for pno, page in enumerate(pdf.pages, 1):
        tables = page.extract_tables() or []
        tboxes = [tb.bbox for tb in (page.find_tables() or [])]
        words = page.extract_words(x_tolerance=1.5)
        mid = page.width / 2.0
        left, right = [], []
        for w in words:
            if any(in_bbox(w, tb) for tb in tboxes):
                continue
            cx = (w["x0"] + w["x1"]) / 2.0
            (left if cx < mid else right).append(w)

        def col_lines(col):
            by_top = {}
            for w in col:
                by_top.setdefault(round(w["top"], 1), []).append(w)
            res = []
            for k in sorted(by_top):
                ws = sorted(by_top[k], key=lambda w: w["x0"])
                res.append(" ".join(w["text"] for w in ws))
            return res

        lines = []
        for col in (left, right):
            lines.extend(col_lines(col))
        lines = [l for l in lines if not is_noise(l)]
        tbl_mds = [table_to_md(t) for t in tables if table_to_md(t)]
        page_data[pno] = {"lines": lines, "tables": tbl_mds}

    # ---- assemble: headers inserted at matching text lines ----
    header_lookup = {}
    for pno, title, lvl in headers:
        header_lookup.setdefault(pno, []).append((title, lvl))

    doc_lines = []
    for pno in range(1, len(pdf.pages) + 1):
        data = page_data[pno]
        page_hs = header_lookup.get(pno, [])
        emitted = set()
        for l in data["lines"]:
            ls = l.strip()
            matched = None
            for title, lvl in page_hs:
                if title in emitted:
                    continue
                if ls == title:
                    matched = (title, lvl)
                    break
            if matched:
                title, lvl = matched
                doc_lines.append("")
                doc_lines.append("#" * min(lvl + 1, 6) + " " + title)
                doc_lines.append("")
                emitted.add(title)
            else:
                doc_lines.append(l)
        for title, lvl in page_hs:
            if title not in emitted:
                doc_lines.append("")
                doc_lines.append("#" * min(lvl + 1, 6) + " " + title)
                doc_lines.append("")
        for t in data["tables"]:
            doc_lines.append("")
            doc_lines.extend(t.split("\n"))
        doc_lines.append("")

    doc = "\n".join(doc_lines)
    # final cleanup of TI page noise
    for p in [r"^www\.ti\.com\s*$", r"^Product Folder\s*$", r"^Links: .+\s*$",
              r"^SLUSEN\d? – .+\s*$", r"^Submit Document Feedback\s*$",
              r"^Copyright © .+\s*$", r"^\d{1,3}$"]:
        doc = "\n".join(l for l in doc.split("\n") if not re.match(p, l.strip()))
    doc = re.sub(r"\n{4,}", "\n\n\n", doc)

    header = (f"# {os.path.basename(src)} - Markdown\n\n"
              f"> Auto-converted from {src}\n"
              f"> 转换日期：{__import__('datetime').date.today().isoformat()}。"
              f"标题层级取自 PDF 书签；正文双栏按左→右输出；表格已转 Markdown；"
              f"下标参数名平铺（如 IQ_BAT），关键参数请对照原 PDF 核对。\n\n")
    with open(out, "w", encoding="utf-8") as f:
        f.write(header + doc)
    return len(doc), len(headers), doc.count("\n| ")


if __name__ == "__main__":
    if len(sys.argv) < 2:
        print(__doc__)
        sys.exit(1)
    src = sys.argv[1]
    out = sys.argv[2] if len(sys.argv) > 2 else re.sub(r"\.pdf$", ".md", src, flags=re.I)
    chars, nhdr, ntbl = convert(src, out)
    print(f"OK: {out} | chars={chars} | headers={nhdr} | table_rows={ntbl}")