skill-case-get/scripts/office_extract.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""skill-case-get —— .docx / .pptx 解包抽取器（纯标准库）

本机没有 python-docx / python-pptx，Node 也没有对应库；
.docx / .pptx 本质是 zip，因此用标准库 `zipfile` + `xml.etree.ElementTree` 解包解析。

抽取内容：
  - 文字段落（按文档/幻灯片顺序，保持阅读序）
  - 表格文本（.docx 的 w:tbl，逐行拼接）
  - 内嵌媒体（word/media/*、ppt/media/*、xl/media/*）落地到 --media-out
  - 超链接（document.xml 的 w:hyperlink r:id → rels 目标；slide 的 a:hlinkClick）
  - .pptx 的幻灯片切换顺序（presentation.xml 的 sldIdLst + rels）
  - .pptx 图解/备注：p:sp 的 a:t 文本（标题、正文、备注页）
  - .docx 页眉页脚（可选，--no-headers 关闭）

用法：
  python3 office_extract.py --input <file.docx|file.pptx> [--media-out <dir>] [--out <result.json>]
  python3 office_extract.py --input a.docx --input b.pptx --out bundle.json
  python3 office_extract.py --selftest

输出：一份 JSON（stdout 或 --out），结构见 `result` 组装处。
不写日志、不落任何 PII 到 stdout 之外的位置；媒体文件按原字节落地到 media-out。
"""

import argparse
import json
import os
import re
import sys
import zipfile
import xml.etree.ElementTree as ET

NS = {
    # WordprocessingML
    "w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main",
    "r": "http://schemas.openxmlformats.org/officeDocument/2006/relationships",
    "wp": "http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing",
    "a": "http://schemas.openxmlformats.org/drawingml/2006/main",
    # PresentationML
    "p": "http://schemas.openxmlformats.org/presentationml/2006/main",
    # package relationships
    "pr": "http://schemas.openxmlformats.org/package/2006/relationships",
    "ct": "http://schemas.openxmlformats.org/package/2006/content-types",
}

DOCX_EXT = {".docx", ".docm"}
PPTX_EXT = {".pptx", ".pptm"}
MEDIA_PREFIXES = ("word/media/", "ppt/media/", "xl/media/", "media/")
IMAGE_EXTS = {".png", ".jpg", ".jpeg", ".gif", ".bmp", ".webp", ".tiff", ".emf", ".wmf"}
VIDEO_EXTS = {".mp4", ".mov", ".m4v", ".avi", ".webm", ".mkv", ".wmv"}
AUDIO_EXTS = {".mp3", ".m4a", ".wav", ".aac", ".flac", ".ogg", ".wma"}

REL_NS = "{http://schemas.openxmlformats.org/officeDocument/2006/relationships}"
PR_NS = "{http://schemas.openxmlformats.org/package/2006/relationships}"


def qname(prefix, local):
    return "{%s}%s" % (NS[prefix], local)


def safe_relpath(name):
    """把 zip 内条目名规整成安全的相对路径（防 zip slip）。

    正确处理 `..`：向上弹一层（而不是直接丢弃），这样
    `ppt/slides/../media/a.png` → `ppt/media/a.png`；
    越出根的 `..` 一律忽略，路径不可能逃出包外。
    """
    clean = name.replace("\\", "/").lstrip("/")
    parts = []
    for segment in clean.split("/"):
        if not segment or segment == ".":
            continue
        if segment == "..":
            if parts:
                parts.pop()
            continue
        parts.append(segment)
    return "/".join(parts)


def local_name(tag):
    return tag.split("}", 1)[1] if "}" in tag else tag


# ---------------------------------------------------------------------------
# relationships
# ---------------------------------------------------------------------------

def parse_rels(zf, rels_path):
    """解析 _rels 部件 → {rId: {target, type, mode}}。"""
    out = {}
    try:
        raw = zf.read(rels_path)
    except (KeyError, OSError):
        return out
    try:
        root = ET.fromstring(raw)
    except ET.ParseError:
        return out
    for rel in root:
        rid = rel.get("Id") or ""
        target = rel.get("Target") or ""
        if not rid:
            continue
        out[rid] = {
            "id": rid,
            "target": target,
            "type": (rel.get("Type") or "").rsplit("/", 1)[-1],
            "mode": rel.get("TargetMode") or "Internal",
        }
    return out


def rel_target_to_part(rels_path, target):
    """把相对 Target 规整成 zip 内条目路径。"""
    if target.startswith("/"):
        return safe_relpath(target)
    base = posix_dirname(posix_dirname(rels_path))  # _rels 的上一级
    return safe_relpath(posix_join(base, target))


def posix_dirname(p):
    return p.rsplit("/", 1)[0] if "/" in p else ""


def posix_join(a, b):
    if not a:
        return safe_relpath(b)
    return safe_relpath(a + "/" + b)


# ---------------------------------------------------------------------------
# text helpers
# ---------------------------------------------------------------------------

def collect_text_runs(element, run_tag, text_tag):
    """按阅读序收集某个节点下的全部文本片段。"""
    parts = []
    for node in element.iter():
        if node.tag == run_tag or local_name(node.tag) == local_name(text_tag):
            if node.tag == text_tag and node.text:
                parts.append(node.text)
    return "".join(parts)


def paragraph_text(p_node):
    """一个 w:p / a:p 的文本。"""
    parts = []
    for node in p_node.iter():
        ln = local_name(node.tag)
        if ln == "t" and node.text:
            parts.append(node.text)
        elif ln == "tab":
            parts.append("\t")
        elif ln in ("br", "cr"):
            parts.append("\n")
    return "".join(parts).strip()


def is_heading(p_node):
    """w:p 是否带 Heading 样式。"""
    for node in p_node.iter(qname("w", "pStyle")):
        val = node.get(qname("w", "val")) or ""
        if val.lower().startswith("heading") or val in ("Title", "标题"):
            return True
    return False


# ---------------------------------------------------------------------------
# docx
# ---------------------------------------------------------------------------

def extract_docx(zf, media_out, include_headers=True):
    result = {
        "kind": "document",
        "format": "docx",
        "paragraphs": [],
        "outline": [],
        "tables": [],
        "hyperlinks": [],
        "headings": [],
        "media": [],
        "embeddedObjects": [],
        "notes": [],
    }

    doc_part = "word/document.xml"
    if doc_part not in zf.namelist():
        for name in zf.namelist():
            if name.startswith("word/") and name.endswith("document.xml"):
                doc_part = name
                break
    if doc_part not in zf.namelist():
        result["error"] = "DOCX_MISSING_DOCUMENT_XML"
        return result

    rels = parse_rels(zf, "word/_rels/document.xml.rels")
    root = ET.fromstring(zf.read(doc_part))
    body = root.find(qname("w", "body"))
    if body is None:
        body = root

    def walk_blocks(container, scope):
        for child in container:
            ln = local_name(child.tag)
            if ln == "p":
                text = paragraph_text(child)
                if text:
                    entry = {"index": len(result["paragraphs"]), "scope": scope, "text": text}
                    if is_heading(child):
                        entry["heading"] = True
                        result["headings"].append(text)
                    result["paragraphs"].append(entry)
                    result["outline"].append({"level": 1 if entry.get("heading") else 0, "text": text})
                # 段落里的图片（drawing / pict）
                for blip in child.iter():
                    if local_name(blip.tag) == "blip":
                        rid = blip.get(REL_NS + "embed") or blip.get(REL_NS + "link")
                        if rid and rid in rels:
                            result["media"].append({
                                "rId": rid,
                                "part": rel_target_to_part("word/_rels/document.xml.rels", rels[rid]["target"]),
                                "anchor": "inline",
                                "nearbyText": text[:80],
                            })
            elif ln == "tbl":
                rows = []
                for tr in child.findall(qname("w", "tr")):
                    cells = []
                    for tc in tr.findall(qname("w", "tc")):
                        cells.append(" ".join(paragraph_text(p) for p in tc.findall(qname("w", "p"))).strip())
                    rows.append(cells)
                if rows:
                    result["tables"].append({"scope": scope, "rows": rows, "rowCount": len(rows)})
                walk_blocks(child, scope)
            elif ln in ("sdt", "sdtContent", "tc"):
                walk_blocks(child, scope)

    walk_blocks(body, "body")

    # 超链接：w:hyperlink r:id → rels
    for link in root.iter(qname("w", "hyperlink")):
        rid = link.get(REL_NS + "id") or link.get(qname("r", "id")) or ""
        target = rels.get(rid, {}).get("target", "")
        text = paragraph_text(link)
        if target or text:
            result["hyperlinks"].append({"text": text, "target": target, "scope": "body"})

    # 内嵌 OLE / 附件对象
    for obj in root.iter(qname("w", "object")):
        for node in obj.iter():
            if local_name(node.tag) == "OLEObject":
                rid = node.get(REL_NS + "id") or node.get(qname("r", "id")) or ""
                if rid in rels:
                    result["embeddedObjects"].append({"rId": rid, "part": rel_target_to_part("word/_rels/document.xml.rels", rels[rid]["target"]), "progId": node.get("ProgID") or ""})

    if include_headers:
        for name in sorted(zf.namelist()):
            if re.match(r"^word/(header|footer)\d*\.xml$", name):
                try:
                    hroot = ET.fromstring(zf.read(name))
                except ET.ParseError:
                    continue
                texts = [paragraph_text(p) for p in hroot.iter(qname("w", "p"))]
                texts = [t for t in texts if t]
                if texts:
                    result["notes"].append({"scope": name, "text": "\n".join(texts)})

    result["media"] = attach_media(zf, result["media"], media_out)
    return result


# ---------------------------------------------------------------------------
# pptx
# ---------------------------------------------------------------------------

def slide_sort_key(name):
    m = re.search(r"slide(\d+)\.xml$", name)
    return int(m.group(1)) if m else 10 ** 6


def extract_pptx(zf, media_out):
    result = {
        "kind": "presentation",
        "format": "pptx",
        "slides": [],
        "paragraphs": [],
        "outline": [],
        "tables": [],
        "hyperlinks": [],
        "headings": [],
        "media": [],
        "embeddedObjects": [],
        "notes": [],
        "slideCount": 0,
    }

    pres_rels = parse_rels(zf, "ppt/_rels/presentation.xml.rels")
    ordered_parts = []

    # 首选：presentation.xml 的 sldIdLst（真实放映顺序）
    if "ppt/presentation.xml" in zf.namelist():
        try:
            proot = ET.fromstring(zf.read("ppt/presentation.xml"))
            for sld in proot.iter(qname("p", "sldId")):
                rid = sld.get(REL_NS + "id") or sld.get(qname("r", "id")) or ""
                if rid in pres_rels:
                    ordered_parts.append(rel_target_to_part("ppt/_rels/presentation.xml.rels", pres_rels[rid]["target"]))
        except ET.ParseError:
            pass

    if not ordered_parts:
        ordered_parts = sorted(
            [n for n in zf.namelist() if re.match(r"^ppt/slides/slide\d+\.xml$", n)],
            key=slide_sort_key,
        )

    for idx, part in enumerate(ordered_parts, start=1):
        if part not in zf.namelist():
            continue
        try:
            sroot = ET.fromstring(zf.read(part))
        except ET.ParseError:
            continue
        rels = parse_rels(zf, posix_join(posix_dirname(part), "_rels/" + part.rsplit("/", 1)[-1] + ".rels"))

        title = ""
        body_lines = []
        slide_texts = []
        pic_count = 0

        for sp in sroot.iter(qname("p", "sp")):
            ph = None
            for node in sp.iter(qname("p", "ph")):
                ph = node.get("type") or "body"
                break
            texts = []
            for p_node in sp.iter(qname("a", "p")):
                t = paragraph_text(p_node)
                if t:
                    texts.append(t)
            if not texts:
                continue
            if ph == "title" or ph == "ctrTitle" or (not title and ph is None and len(texts) == 1 and len(texts[0]) < 60):
                if not title:
                    title = texts[0]
                    result["headings"].append(texts[0])
                    continue
            body_lines.extend(texts)

        # 表格
        for tbl in sroot.iter(qname("a", "tbl")):
            rows = []
            for tr in tbl.iter(qname("a", "tr")):
                cells = []
                for tc in tr.iter(qname("a", "tc")):
                    cell_text = " ".join(paragraph_text(p) for p in tc.iter(qname("a", "p"))).strip()
                    cells.append(cell_text)
                if cells:
                    rows.append(cells)
            if rows:
                result["tables"].append({"scope": part, "slide": idx, "rows": rows, "rowCount": len(rows)})

        # 图片
        for blip in sroot.iter(qname("a", "blip")):
            rid = blip.get(REL_NS + "embed") or blip.get(REL_NS + "link")
            if rid and rid in rels:
                pic_count += 1
                result["media"].append({
                    "rId": rid,
                    "part": rel_target_to_part(posix_join(posix_dirname(part), "_rels/" + part.rsplit("/", 1)[-1] + ".rels"), rels[rid]["target"]),
                    "anchor": "slide",
                    "slide": idx,
                    "nearbyText": title or (body_lines[0] if body_lines else ""),
                })

        # 超链接
        for link in sroot.iter(qname("a", "hlinkClick")):
            rid = link.get(REL_NS + "id")
            if rid and rid in rels:
                result["hyperlinks"].append({"text": title, "target": rels[rid]["target"], "scope": part, "slide": idx})

        # 备注页
        notes_part = posix_join(posix_dirname(part), "notesSlides/notesSlide%d.xml" % idx)
        notes_text = ""
        if notes_part in zf.namelist():
            try:
                nroot = ET.fromstring(zf.read(notes_part))
                notes_text = "\n".join(t for t in (paragraph_text(p) for p in nroot.iter(qname("a", "p"))) if t)
            except ET.ParseError:
                notes_text = ""
        if notes_text:
            result["notes"].append({"scope": notes_part, "slide": idx, "text": notes_text})

        slide_entry = {
            "index": idx,
            "part": part,
            "title": title,
            "texts": body_lines,
            "text": "\n".join([title] + body_lines).strip(),
            "notes": notes_text,
            "imageCount": pic_count,
        }
        result["slides"].append(slide_entry)
        for line in slide_entry["text"].split("\n"):
            if line.strip():
                result["paragraphs"].append({"index": len(result["paragraphs"]), "scope": part, "slide": idx, "text": line.strip()})
                result["outline"].append({"level": 1 if line.strip() == title else 0, "text": line.strip()})

    # 内嵌文件（ppt/embeddings/*）
    for name in zf.namelist():
        if name.startswith("ppt/embeddings/") and not name.endswith("/"):
            result["embeddedObjects"].append({"part": safe_relpath(name), "progId": ""})

    result["slideCount"] = len(result["slides"])
    result["media"] = attach_media(zf, result["media"], media_out)
    return result


# ---------------------------------------------------------------------------
# media
# ---------------------------------------------------------------------------

def classify_media(part):
    ext = os.path.splitext(part)[1].lower()
    if ext in IMAGE_EXTS or ext == "":
        return "image"
    if ext in VIDEO_EXTS:
        return "video"
    if ext in AUDIO_EXTS:
        return "audio"
    return "other"


def attach_media(zf, media_entries, media_out):
    """把媒体条目去重、补上 kind/size，并按需落地到 media_out。"""
    by_part = {}
    order = []
    for entry in media_entries:
        part = safe_relpath(entry.get("part") or "")
        if not part:
            continue
        if part not in by_part:
            by_part[part] = {**entry, "part": part, "kind": classify_media(part), "occurrences": 0}
            order.append(part)
        by_part[part]["occurrences"] += 1

    out = []
    for part in order:
        entry = by_part[part]
        entry["size"] = zf.getinfo(part).file_size if part in zf.namelist() else 0
        entry["localPath"] = ""
        if media_out and part in zf.namelist():
            safe_name = part.replace("/", "__")
            target = os.path.join(media_out, safe_name)
            os.makedirs(media_out, exist_ok=True)
            with zf.open(part) as src, open(target, "wb") as dst:
                dst.write(src.read())
            entry["localPath"] = os.path.abspath(target)
        out.append(entry)
    return out


# ---------------------------------------------------------------------------
# entry
# ---------------------------------------------------------------------------

def package_kind(path):
    ext = os.path.splitext(path)[1].lower()
    if ext in DOCX_EXT:
        return "docx"
    if ext in PPTX_EXT:
        return "pptx"
    # 扩展名不认识时，看 package 内容
    try:
        with zipfile.ZipFile(path) as zf:
            names = zf.namelist()
            if any(n.startswith("word/") for n in names):
                return "docx"
            if any(n.startswith("ppt/") for n in names):
                return "pptx"
    except (zipfile.BadZipFile, OSError):
        pass
    return ""


def extract_one(path, media_out=None, include_headers=True):
    kind = package_kind(path)
    base = {
        "input": os.path.abspath(path),
        "fileName": os.path.basename(path),
        "byteSize": os.path.getsize(path) if os.path.exists(path) else 0,
    }
    if not kind:
        return {**base, "ok": False, "error": "UNSUPPORTED_PACKAGE", "message": "不是 .docx/.pptx 包（zip 内无 word/ 或 ppt/）"}
    try:
        with zipfile.ZipFile(path) as zf:
            if kind == "docx":
                payload = extract_docx(zf, media_out, include_headers=include_headers)
            else:
                payload = extract_pptx(zf, media_out)
    except zipfile.BadZipFile:
        return {**base, "ok": False, "error": "BAD_ZIP", "message": "文件损坏或不是有效 zip 包"}
    except ET.ParseError as exc:
        return {**base, "ok": False, "error": "BAD_XML", "message": "XML 解析失败：%s" % exc}
    return {**base, "ok": True, **payload}


def selftest():
    """自检：现场构造一个最小 docx + pptx，解包并断言关键结果。"""
    import io
    import tempfile

    doc_xml = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<w:document xmlns:w="%s" xmlns:r="%s" xmlns:wp="%s" xmlns:a="%s">'
        "<w:body>"
        '<w:p><w:pPr><w:pStyle w:val="Heading1"/></w:pPr><w:r><w:t>帝国理工 计算机科学 辅导案例</w:t></w:r></w:p>'
        "<w:p><w:r><w:t>学生考前冲刺，最终</w:t></w:r><w:r><w:t>提分 18 分</w:t></w:r></w:p>"
        "<w:p><w:r><w:drawing><wp:inline><a:graphic><a:blip r:embed=\"rId5\"/></a:graphic></wp:inline></w:drawing></w:r></w:p>"
        '<w:hyperlink r:id="rId9"><w:r><w:t>查看反馈</w:t></w:r></w:hyperlink>'
        "<w:tbl><w:tr><w:tc><w:p><w:r><w:t>A</w:t></w:r></w:p></w:tc><w:tc><w:p><w:r><w:t>B</w:t></w:r></w:p></w:tc></w:tr></w:tbl>"
        "</w:body></w:document>"
    ) % (NS["w"], NS["r"], NS["wp"], NS["a"])

    doc_rels = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<Relationships xmlns="%s">'
        '<Relationship Id="rId5" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/image" Target="media/image1.png"/>'
        '<Relationship Id="rId9" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/hyperlink" Target="https://example.invalid/case" TargetMode="External"/>'
        "</Relationships>"
    ) % NS["pr"]

    slide_xml = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<p:sld xmlns:p="%s" xmlns:a="%s" xmlns:r="%s"><p:cSld><p:spTree>'
        '<p:sp><p:nvSpPr><p:nvPr><p:ph type="title"/></p:nvPr></p:nvSpPr><p:txBody><a:p><a:r><a:t>案例合集</a:t></a:r></a:p></p:txBody></p:sp>'
        "<p:sp><p:nvSpPr/><p:txBody><a:p><a:r><a:t>正文一</a:t></a:r></a:p></p:txBody></p:sp>"
        '<p:pic><p:blipFill><a:blip r:embed="rId2"/></p:blipFill></p:pic>'
        "</p:spTree></p:cSld></p:sld>"
    ) % (NS["p"], NS["a"], NS["r"])

    slide_rels = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<Relationships xmlns="%s">'
        '<Relationship Id="rId2" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/image" Target="../media/image1.png"/>'
        "</Relationships>"
    ) % NS["pr"]

    pres_xml = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<p:presentation xmlns:p="%s" xmlns:r="%s">'
        '<p:sldIdLst><p:sldId id="256" r:id="rId1"/></p:sldIdLst></p:presentation>'
    ) % (NS["p"], NS["r"])

    pres_rels = (
        '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>'
        '<Relationships xmlns="%s">'
        '<Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/slide" Target="slides/slide1.xml"/>'
        "</Relationships>"
    ) % NS["pr"]

    png_bytes = bytes.fromhex(
        "89504e470d0a1a0a0000000d49484452000000010000000108060000001f15c489"
        "0000000a49444154789c6300010000050001"
        "0d0a2db40000000049454e44ae426082"
    )

    failures = []
    with tempfile.TemporaryDirectory() as tmp:
        docx_path = os.path.join(tmp, "case.docx")
        with zipfile.ZipFile(docx_path, "w") as zf:
            zf.writestr("[Content_Types].xml", "<Types/>")
            zf.writestr("word/document.xml", doc_xml)
            zf.writestr("word/_rels/document.xml.rels", doc_rels)
            zf.writestr("word/media/image1.png", png_bytes)

        res = extract_one(docx_path, media_out=os.path.join(tmp, "media"))
        if not res.get("ok"):
            failures.append("docx 解包失败：%s" % res.get("error"))
        else:
            texts = [p["text"] for p in res["paragraphs"]]
            if "帝国理工 计算机科学 辅导案例" not in texts:
                failures.append("docx 标题段落缺失")
            if not any("提分 18 分" in t for t in texts):
                failures.append("docx 跨 run 文本拼接失败")
            if not res["headings"]:
                failures.append("docx 标题识别失败")
            if not res["tables"] or res["tables"][0]["rows"][0] != ["A", "B"]:
                failures.append("docx 表格抽取失败")
            if not res["hyperlinks"] or res["hyperlinks"][0]["target"] != "https://example.invalid/case":
                failures.append("docx 超链接抽取失败")
            media = [m for m in res["media"] if m["kind"] == "image"]
            if not media:
                failures.append("docx 内嵌图片未识别")
            elif not os.path.exists(media[0]["localPath"]):
                failures.append("docx 内嵌图片未落地")

        pptx_path = os.path.join(tmp, "deck.pptx")
        with zipfile.ZipFile(pptx_path, "w") as zf:
            zf.writestr("[Content_Types].xml", "<Types/>")
            zf.writestr("ppt/presentation.xml", pres_xml)
            zf.writestr("ppt/_rels/presentation.xml.rels", pres_rels)
            zf.writestr("ppt/slides/slide1.xml", slide_xml)
            zf.writestr("ppt/slides/_rels/slide1.xml.rels", slide_rels)
            zf.writestr("ppt/media/image1.png", png_bytes)

        res = extract_one(pptx_path, media_out=os.path.join(tmp, "media2"))
        if not res.get("ok"):
            failures.append("pptx 解包失败：%s" % res.get("error"))
        else:
            if res.get("slideCount") != 1:
                failures.append("pptx 页数不对：%s" % res.get("slideCount"))
            if not res["slides"] or res["slides"][0]["title"] != "案例合集":
                failures.append("pptx 标题占位符识别失败")
            if not any("正文一" in t for t in res["slides"][0]["texts"]):
                failures.append("pptx 正文抽取失败")
            if not [m for m in res["media"] if m["kind"] == "image"]:
                failures.append("pptx 图片未识别")
            elif res["media"][0]["part"] != "ppt/media/image1.png":
                failures.append("pptx 图片路径未正确解析 ..：%s" % res["media"][0]["part"])
            elif not os.path.exists(res["media"][0]["localPath"]):
                failures.append("pptx 内嵌图片未落地")

    if failures:
        print(json.dumps({"ok": False, "failures": failures}, ensure_ascii=False, indent=2))
        return 1
    print(json.dumps({"ok": True, "checked": ["docx", "pptx"], "message": "office 解包自检通过"}, ensure_ascii=False, indent=2))
    return 0


def main(argv=None):
    parser = argparse.ArgumentParser(description="skill-case-get .docx/.pptx 解包抽取器（标准库实现）")
    parser.add_argument("--input", action="append", default=[], help="待解包的 .docx/.pptx（可重复）")
    parser.add_argument("--media-out", default="", help="内嵌媒体落地目录")
    parser.add_argument("--out", default="", help="结果 JSON 输出路径（默认 stdout）")
    parser.add_argument("--no-headers", action="store_true", help=".docx 不解页眉页脚")
    parser.add_argument("--selftest", action="store_true", help="运行内置自检")
    args = parser.parse_args(argv)

    if args.selftest:
        return selftest()

    if not args.input:
        parser.error("至少需要一个 --input")

    results = [extract_one(p, args.media_out or None, include_headers=not args.no_headers) for p in args.input]
    payload = {
        "ok": all(r.get("ok") for r in results),
        "generator": "skill-case-get/office_extract.py",
        "count": len(results),
        "results": results,
    }
    text = json.dumps(payload, ensure_ascii=False, indent=2)
    if args.out:
        with open(args.out, "w", encoding="utf-8") as fh:
            fh.write(text + "\n")
    else:
        sys.stdout.write(text + "\n")
    return 0 if payload["ok"] else 2


if __name__ == "__main__":
    sys.exit(main())