\r \r

PDF / PPTX / DOCX / XLSX 文档转换工具集\r

\r 四大办公文档格式的转换、提取、合并、拆分工具。基于 Python 生态，所有操作均可通过 execute_command 执行 Python 一行命令或短脚本完成。\r \r

依赖安装\r

\r

pip install PyMuPDF pdf2docx python-docx python-pptx openpyxl pandas Pillow pdfplumber\r
```\r
\r
部分操作（DOCX→PDF、PPTX→PDF、XLSX→PDF）需要系统安装 **LibreOffice**：\r
\r
```bash\r
# Windows (winget)\r
winget install LibreOffice.LibreOffice\r
\r
# macOS\r
brew install --cask libreoffice\r
\r
# Ubuntu/Debian\r
sudo apt install libreoffice\r
```\r
\r
PDF 转图片如需高质量渲染，可选装 poppler（`PyMuPDF` 内置渲染已足够，poppler 仅作为备选）。\r
\r
---\r
\r
## 快速参考：支持的全部转换\r
\r
| 源格式 | 目标格式 | 推荐库 | 备注 |\r
|--------|----------|--------|------|\r
| PDF | 图片 (PNG/JPG) | PyMuPDF | 逐页渲染，支持 DPI 控制 |\r
| PDF | DOCX | pdf2docx | 保留布局、表格、图片 |\r
| PDF | PPTX | PyMuPDF + python-pptx | 每页一张幻灯片 |\r
| PDF | XLSX | pdfplumber + openpyxl | 提取表格数据 |\r
| PDF | 文本 (TXT) | PyMuPDF | 提取纯文本 |\r
| DOCX | PDF | LibreOffice (CLI) | 最佳保真度 |\r
| DOCX | PPTX | python-docx + python-pptx | 段落→幻灯片 |\r
| DOCX | HTML | python-docx / mammoth | 保留基本格式 |\r
| DOCX | 纯文本 | python-docx | 提取所有段落文本 |\r
| PPTX | PDF | LibreOffice (CLI) | 最佳保真度 |\r
| PPTX | 图片 (PNG) | PyMuPDF | 每页导出为图片 |\r
| PPTX | DOCX | python-pptx + python-docx | 提取所有文本 |\r
| PPTX | 纯文本 | python-pptx | 提取幻灯片文本 |\r
| XLSX | PDF | LibreOffice (CLI) | 最佳保真度 |\r
| XLSX | CSV | pandas | 可指定 sheet |\r
| XLSX | DOCX | openpyxl + python-docx | 表格写入 Word |\r
| XLSX | JSON | pandas | 结构化数据导出 |\r
| 图片 | PDF | Pillow + reportlab | 多图合并为 PDF |\r
\r
---\r
\r
## 转换命令详解\r
\r
### 1. PDF → 图片\r
\r
```bash\r
python -c "\r
import fitz, sys, os\r
pdf_path, out_dir = sys.argv[1], sys.argv[2] if len(sys.argv)>2 else '.'\r
os.makedirs(out_dir, exist_ok=True)\r
doc = fitz.open(pdf_path)\r
dpi = int(sys.argv[3]) if len(sys.argv)>3 else 200\r
fmt = sys.argv[4] if len(sys.argv)>4 else 'png'\r
for i, page in enumerate(doc):\r
    pix = page.get_pixmap(dpi=dpi)\r
    out = os.path.join(out_dir, f'page_{i+1:04d}.{fmt}')\r
    pix.save(out)\r
    print(f'Saved: {out}')\r
doc.close()\r
print(f'Done: {len(doc)} pages -> {out_dir}')\r
" "input.pdf" "./output_images" 200 png\r
```\r
\r
**参数说明：**\r
- `arg1` — PDF 文件路径\r
- `arg2` — 输出目录（默认当前目录）\r
- `arg3` — DPI 分辨率（默认 200，推荐 150~300）\r
- `arg4` — 图片格式：`png`（默认）或 `jpg`\r
\r
### 2. PDF → DOCX\r
\r
```bash\r
python -c "\r
from pdf2docx import Converter\r
import sys\r
cv = Converter(sys.argv[1])\r
cv.convert(sys.argv[2] if len(sys.argv)>2 else 'output.docx')\r
cv.close()\r
print('Done')\r
" "input.pdf" "output.docx"\r
```\r
\r
**可选参数（通过修改脚本）：**\r
- `start=0, end=None` — 指定页码范围\r
- `multi_processing=True` — 多进程加速大文件\r
\r
### 3. PDF → PPTX（每页一张幻灯片）\r
\r
```bash\r
python -c "\r
import fitz, sys\r
from pptx import Presentation\r
from pptx.util import Inches\r
from pptx.dml.color import RGBColor\r
import io\r
\r
pdf_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.pptx'\r
doc = fitz.open(pdf_path)\r
prs = Presentation()\r
prs.slide_width = Inches(13.333)\r
prs.slide_height = Inches(7.5)\r
\r
blank_layout = prs.slide_layouts[6]  # blank\r
for i, page in enumerate(doc):\r
    pix = page.get_pixmap(dpi=200)\r
    img_data = pix.tobytes('png')\r
    slide = prs.slides.add_slide(blank_layout)\r
    slide.shapes.add_picture(io.BytesIO(img_data), Inches(0), Inches(0),\r
                             width=prs.slide_width, height=prs.slide_height)\r
    print(f'Page {i+1}/{len(doc)} added')\r
\r
prs.save(out_path)\r
doc.close()\r
print(f'Done: {out_path}')\r
" "input.pdf" "output.pptx"\r
```\r
\r
### 4. PDF → XLSX（提取表格）\r
\r
```bash\r
python -c "\r
import pdfplumber, openpyxl, sys\r
\r
pdf_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.xlsx'\r
wb = openpyxl.Workbook()\r
ws_total = wb.active\r
ws_total.title = 'All Tables'\r
row_offset = 0\r
\r
with pdfplumber.open(pdf_path) as pdf:\r
    for page_num, page in enumerate(pdf.pages):\r
        tables = page.extract_tables()\r
        for t_idx, table in enumerate(tables):\r
            if row_offset == 0 and t_idx == 0:\r
                ws = ws_total\r
            else:\r
                ws = wb.create_sheet(title=f'p{page_num+1}_t{t_idx+1}')\r
            for row in table:\r
                ws.append(row)\r
            row_offset += len(table)\r
            print(f'Page {page_num+1}, Table {t_idx+1}: {len(table)} rows')\r
\r
if ws_total.max_row == 1 and ws_total.max_column == 1:\r
    wb.remove(ws_total)\r
\r
wb.save(out_path)\r
print(f'Done: {out_path}')\r
" "input.pdf" "output.xlsx"\r
```\r
\r
### 5. PDF → 纯文本\r
\r
```bash\r
python -c "\r
import fitz, sys\r
doc = fitz.open(sys.argv[1])\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.txt'\r
with open(out_path, 'w', encoding='utf-8') as f:\r
    for i, page in enumerate(doc):\r
        text = page.get_text()\r
        f.write(f'--- Page {i+1} ---\
{text}\
\
')\r
print(f'Done: {len(doc)} pages extracted')\r
doc.close()\r
" "input.pdf" "output.txt"\r
```\r
\r
### 6. DOCX → PDF\r
\r
```bash\r
# Windows\r
python -c "\r
import subprocess, sys, os\r
docx_path = os.path.abspath(sys.argv[1])\r
out_dir = os.path.dirname(docx_path)\r
subprocess.run([\r
    r'C:\Program Files\LibreOffice\program\soffice.exe',\r
    '--headless', '--convert-to', 'pdf',\r
    '--outdir', out_dir, docx_path\r
], check=True)\r
print(f'Done')\r
" "input.docx"\r
\r
# macOS / Linux\r
soffice --headless --convert-to pdf --outdir ./ "input.docx"\r
```\r
\r
### 7. DOCX → PPTX\r
\r
```bash\r
python -c "\r
from docx import Document\r
from pptx import Presentation\r
from pptx.util import Pt, Inches\r
import sys\r
\r
docx_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.pptx'\r
doc = Document(docx_path)\r
prs = Presentation()\r
prs.slide_width = Inches(13.333)\r
prs.slide_height = Inches(7.5)\r
blank_layout = prs.slide_layouts[6]\r
\r
slide = None\r
bullet_count = 0\r
max_bullets = 8\r
\r
for para in doc.paragraphs:\r
    text = para.text.strip()\r
    if not text:\r
        continue\r
    style = para.style.name.lower()\r
    is_heading = 'heading' in style or 'title' in style\r
\r
    if is_heading or bullet_count >= max_bullets:\r
        slide = prs.slides.add_slide(blank_layout)\r
        bullet_count = 0\r
        txBox = slide.shapes.add_textbox(Inches(0.5), Inches(0.5),\r
                                          Inches(12.333), Inches(6.5))\r
        tf = txBox.text_frame\r
        tf.word_wrap = True\r
        if is_heading:\r
            p = tf.paragraphs[0]\r
            p.text = text\r
            p.font.size = Pt(32)\r
            p.font.bold = True\r
            bullet_count = 0\r
            continue\r
\r
    if slide is None:\r
        slide = prs.slides.add_slide(blank_layout)\r
        txBox = slide.shapes.add_textbox(Inches(0.5), Inches(0.5),\r
                                          Inches(12.333), Inches(6.5))\r
        tf = txBox.text_frame\r
        tf.word_wrap = True\r
\r
    p = tf.add_paragraph()\r
    p.text = text\r
    p.font.size = Pt(20)\r
    p.level = min(para.style.name.count('Heading') if 'Heading' in para.style.name else 0, 2)\r
    bullet_count += 1\r
\r
prs.save(out_path)\r
print(f'Done: {len(prs.slides)} slides -> {out_path}')\r
" "input.docx" "output.pptx"\r
```\r
\r
### 8. DOCX → HTML\r
\r
```bash\r
python -c "\r
from docx import Document\r
from docx.oxml.ns import qn\r
import sys, re\r
\r
doc = Document(sys.argv[1])\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.html'\r
\r
def paragraph_to_html(para):\r
    text = para.text\r
    style = para.style.name.lower()\r
    if 'heading 1' in style or 'title' in style:\r
        return f'\x3Ch1>{text}\x3C/h1>'\r
    elif 'heading 2' in style:\r
        return f'\x3Ch2>{text}\x3C/h2>'\r
    elif 'heading 3' in style:\r
        return f'\x3Ch3>{text}\x3C/h3>'\r
    elif 'list' in style:\r
        return f'\x3Cli>{text}\x3C/li>'\r
    else:\r
        return f'\x3Cp>{text}\x3C/p>'\r
\r
html_parts = ['\x3C!DOCTYPE html>\x3Chtml>\x3Chead>\x3Cmeta charset=\"utf-8\">\x3Cstyle>body{font-family:sans-serif;max-width:800px;margin:2em auto;padding:0 1em;}h1{color:#333;}h2{color:#555;border-bottom:1px solid #eee;}\x3C/style>\x3C/head>\x3Cbody>']\r
for para in doc.paragraphs:\r
    if para.text.strip():\r
        html_parts.append(paragraph_to_html(para))\r
html_parts.append('\x3C/body>\x3C/html>')\r
\r
with open(out_path, 'w', encoding='utf-8') as f:\r
    f.write('\
'.join(html_parts))\r
print(f'Done: {out_path}')\r
" "input.docx" "output.html"\r
```\r
\r
### 9. PPTX → PDF\r
\r
```bash\r
# Windows\r
python -c "\r
import subprocess, sys, os\r
pptx_path = os.path.abspath(sys.argv[1])\r
out_dir = os.path.dirname(pptx_path)\r
subprocess.run([\r
    r'C:\Program Files\LibreOffice\program\soffice.exe',\r
    '--headless', '--convert-to', 'pdf',\r
    '--outdir', out_dir, pptx_path\r
], check=True)\r
print('Done')\r
" "input.pptx"\r
\r
# macOS / Linux\r
soffice --headless --convert-to pdf --outdir ./ "input.pptx"\r
```\r
\r
### 10. PPTX → 图片\r
\r
```bash\r
python -c "\r
import fitz, sys, os\r
pptx_path = sys.argv[1]\r
out_dir = sys.argv[2] if len(sys.argv)>2 else '.'\r
os.makedirs(out_dir, exist_ok=True)\r
\r
# 先用 LibreOffice 转为 PDF，再用 PyMuPDF 渲染\r
import subprocess, tempfile\r
tmp = tempfile.NamedTemporaryFile(suffix='.pdf', delete=False)\r
tmp.close()\r
subprocess.run([\r
    r'C:\Program Files\LibreOffice\program\soffice.exe',\r
    '--headless', '--convert-to', 'pdf', '--outdir', os.path.dirname(tmp.name),\r
    pptx_path\r
], check=True, capture_output=True)\r
# LibreOffice 输出文件名基于输入文件名\r
pdf_tmp = os.path.join(os.path.dirname(tmp.name),\r
                        os.path.splitext(os.path.basename(pptx_path))[0] + '.pdf')\r
\r
doc = fitz.open(pdf_tmp)\r
for i, page in enumerate(doc):\r
    pix = page.get_pixmap(dpi=200)\r
    out = os.path.join(out_dir, f'slide_{i+1:04d}.png')\r
    pix.save(out)\r
    print(f'Saved: {out}')\r
doc.close()\r
os.unlink(pdf_tmp)\r
print(f'Done: {len(doc)} slides')\r
" "input.pptx" "./slides_output"\r
```\r
\r
> **注意**：macOS/Linux 下将 `soffice.exe` 路径替换为 `soffice`。\r
\r
### 11. PPTX → DOCX（提取文本）\r
\r
```bash\r
python -c "\r
from pptx import Presentation\r
from docx import Document\r
from docx.shared import Pt\r
import sys\r
\r
pptx_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.docx'\r
prs = Presentation(pptx_path)\r
doc = Document()\r
\r
for i, slide in enumerate(prs.slides):\r
    doc.add_heading(f'Slide {i+1}', level=1)\r
    for shape in slide.shapes:\r
        if shape.has_text_frame:\r
            for para in shape.text_frame.paragraphs:\r
                text = para.text.strip()\r
                if text:\r
                    p = doc.add_paragraph(text)\r
                    p.style.font.size = Pt(11)\r
        if shape.has_table:\r
            table = shape.table\r
            rows = []\r
            for row in table.rows:\r
                rows.append([cell.text for cell in row.cells])\r
            t = doc.add_table(rows=len(rows), cols=len(rows[0]) if rows else 0)\r
            for r_idx, row_data in enumerate(rows):\r
                for c_idx, cell_text in enumerate(row_data):\r
                    t.cell(r_idx, c_idx).text = cell_text\r
    doc.add_page_break()\r
\r
doc.save(out_path)\r
print(f'Done: {len(prs.slides)} slides -> {out_path}')\r
" "input.pptx" "output.docx"\r
```\r
\r
### 12. XLSX → PDF\r
\r
```bash\r
# Windows\r
python -c "\r
import subprocess, sys, os\r
xlsx_path = os.path.abspath(sys.argv[1])\r
out_dir = os.path.dirname(xlsx_path)\r
subprocess.run([\r
    r'C:\Program Files\LibreOffice\program\soffice.exe',\r
    '--headless', '--convert-to', 'pdf',\r
    '--outdir', out_dir, xlsx_path\r
], check=True)\r
print('Done')\r
" "input.xlsx"\r
\r
# macOS / Linux\r
soffice --headless --convert-to pdf --outdir ./ "input.xlsx"\r
```\r
\r
### 13. XLSX → CSV\r
\r
```bash\r
python -c "\r
import pandas as pd, sys, os\r
xlsx_path = sys.argv[1]\r
out_dir = sys.argv[2] if len(sys.argv)>2 else '.'\r
os.makedirs(out_dir, exist_ok=True)\r
xls = pd.ExcelFile(xlsx_path)\r
for sheet in xls.sheet_names:\r
    df = pd.read_excel(xls, sheet_name=sheet)\r
    out = os.path.join(out_dir, f'{sheet}.csv')\r
    df.to_csv(out, index=False, encoding='utf-8-sig')\r
    print(f'Saved: {out} ({len(df)} rows)')\r
print('Done')\r
" "input.xlsx" "./csv_output"\r
```\r
\r
### 14. XLSX → JSON\r
\r
```bash\r
python -c "\r
import pandas as pd, json, sys, os\r
xlsx_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.json'\r
xls = pd.ExcelFile(xlsx_path)\r
result = {}\r
for sheet in xls.sheet_names:\r
    df = pd.read_excel(xls, sheet_name=sheet)\r
    result[sheet] = df.to_dict(orient='records')\r
    print(f'Sheet \"{sheet}\": {len(df)} rows')\r
with open(out_path, 'w', encoding='utf-8') as f:\r
    json.dump(result, f, ensure_ascii=False, indent=2, default=str)\r
print(f'Done: {out_path}')\r
" "input.xlsx" "output.json"\r
```\r
\r
### 15. XLSX → DOCX（表格写入 Word）\r
\r
```bash\r
python -c "\r
import openpyxl, sys\r
from docx import Document\r
from docx.shared import Inches, Pt\r
\r
xlsx_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.docx'\r
wb = openpyxl.load_workbook(xlsx_path)\r
doc = Document()\r
\r
for sheet_name in wb.sheetnames:\r
    ws = wb[sheet_name]\r
    doc.add_heading(sheet_name, level=1)\r
    rows = list(ws.iter_rows(values_only=True))\r
    if not rows:\r
        continue\r
    t = doc.add_table(rows=len(rows), cols=len(rows[0]))\r
    for r_idx, row in enumerate(rows):\r
        for c_idx, val in enumerate(row):\r
            cell = t.cell(r_idx, c_idx)\r
            cell.text = str(val) if val is not None else ''\r
            if r_idx == 0:\r
                for run in cell.paragraphs[0].runs:\r
                    run.font.bold = True\r
                    run.font.size = Pt(10)\r
    doc.add_page_break()\r
\r
doc.save(out_path)\r
print(f'Done: {len(wb.sheetnames)} sheets -> {out_path}')\r
" "input.xlsx" "output.docx"\r
```\r
\r
### 16. 图片 → PDF\r
\r
```bash\r
python -c "\r
from PIL import Image\r
from reportlab.lib.pagesizes import A4\r
from reportlab.pdfgen import canvas\r
import sys, os, io\r
\r
img_dir = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'output.pdf'\r
files = sorted([f for f in os.listdir(img_dir)\r
                if f.lower().endswith(('.png','.jpg','.jpeg','.webp','.bmp'))])\r
\r
c = canvas.Canvas(out_path, pagesize=A4)\r
w, h = A4\r
\r
for f in files:\r
    img = Image.open(os.path.join(img_dir, f))\r
    iw, ih = img.size\r
    scale = min(w/iw, h/ih, 1)\r
    dw, dh = iw*scale, ih*scale\r
    x, y = (w-dw)/2, (h-dh)/2\r
    c.drawImage(os.path.join(img_dir, f), x, y, dw, dh)\r
    c.showPage()\r
    print(f'Added: {f}')\r
\r
c.save()\r
print(f'Done: {len(files)} images -> {out_path}')\r
" "./images" "output.pdf"\r
```\r
\r
---\r
\r
## PDF 高级操作\r
\r
### 合并 PDF\r
\r
```bash\r
python -c "\r
import fitz, sys, glob\r
files = sorted(glob.glob(sys.argv[1]))\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'merged.pdf'\r
merged = fitz.open()\r
for f in files:\r
    doc = fitz.open(f)\r
    merged.insert_pdf(doc)\r
    doc.close()\r
    print(f'Merged: {f}')\r
merged.save(out_path)\r
merged.close()\r
print(f'Done: {len(files)} files -> {out_path}')\r
" "./*.pdf" "merged.pdf"\r
```\r
\r
### 拆分 PDF\r
\r
```bash\r
python -c "\r
import fitz, sys, os\r
pdf_path = sys.argv[1]\r
out_dir = sys.argv[2] if len(sys.argv)>2 else './split'\r
os.makedirs(out_dir, exist_ok=True)\r
doc = fitz.open(pdf_path)\r
for i, page in enumerate(doc):\r
    new_doc = fitz.open()\r
    new_doc.insert_pdf(doc, from_page=i, to_page=i)\r
    out = os.path.join(out_dir, f'page_{i+1:04d}.pdf')\r
    new_doc.save(out)\r
    new_doc.close()\r
    print(f'Saved: {out}')\r
doc.close()\r
print(f'Done: {len(doc)} pages split')\r
" "input.pdf" "./split"\r
```\r
\r
### 按页码范围拆分\r
\r
```bash\r
python -c "\r
import fitz, sys\r
pdf_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'extracted.pdf'\r
ranges = sys.argv[3]  # e.g. '1-3,5,7-10'\r
\r
def parse_ranges(s):\r
    result = []\r
    for part in s.split(','):\r
        part = part.strip()\r
        if '-' in part:\r
            a, b = part.split('-')\r
            result.extend(range(int(a)-1, int(b)))\r
        else:\r
            result.append(int(part)-1)\r
    return result\r
\r
pages = parse_ranges(ranges)\r
doc = fitz.open(pdf_path)\r
new_doc = fitz.open()\r
new_doc.insert_pdf(doc, from_page=min(pages), to_page=max(pages))\r
# 如果不是连续页码，逐页插入更精确\r
if pages != list(range(min(pages), max(pages)+1)):\r
    new_doc = fitz.open()\r
    for p in sorted(set(pages)):\r
        new_doc.insert_pdf(doc, from_page=p, to_page=p)\r
new_doc.save(out_path)\r
new_doc.close()\r
doc.close()\r
print(f'Done: pages {ranges} -> {out_path}')\r
" "input.pdf" "extracted.pdf" "1-3,5,7-10"\r
```\r
\r
### PDF 加水印\r
\r
```bash\r
python -c "\r
import fitz, sys\r
\r
pdf_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'watermarked.pdf'\r
text = sys.argv[3] if len(sys.argv)>3 else 'CONFIDENTIAL'\r
opacity = float(sys.argv[4]) if len(sys.argv)>4 else 0.3\r
\r
doc = fitz.open(pdf_path)\r
for page in doc:\r
    # 计算页面中心\r
    rect = page.rect\r
    x, y = rect.width / 2, rect.height / 2\r
    # 旋转 -45 度\r
    rc = fitz.Rect(0, 0, rect.width, rect.height)\r
    page.insert_textbox(\r
        fitz.Rect(x - 200, y - 30, x + 200, y + 30),\r
        text, fontsize=50, color=(0.5, 0.5, 0.5),\r
        rotate=-45, opacity=opacity,\r
        fontname='helv', align=1\r
    )\r
    print(f'Watermarked page {page.number + 1}')\r
\r
doc.save(out_path)\r
doc.close()\r
print(f'Done: {out_path}')\r
" "input.pdf" "watermarked.pdf" "机密文件" 0.2\r
```\r
\r
### PDF 提取图片\r
\r
```bash\r
python -c "\r
import fitz, sys, os\r
pdf_path = sys.argv[1]\r
out_dir = sys.argv[2] if len(sys.argv)>2 else './extracted_images'\r
os.makedirs(out_dir, exist_ok=True)\r
doc = fitz.open(pdf_path)\r
count = 0\r
for page_num, page in enumerate(doc):\r
    images = page.get_images(full=True)\r
    for img_idx, img in enumerate(images):\r
        xref = img[0]\r
        base = doc.extract_image(xref)\r
        ext = base['ext']\r
        out = os.path.join(out_dir, f'p{page_num+1}_img{img_idx+1}.{ext}')\r
        with open(out, 'wb') as f:\r
            f.write(base['image'])\r
        count += 1\r
        print(f'Extracted: {out}')\r
doc.close()\r
print(f'Done: {count} images extracted')\r
" "input.pdf" "./extracted_images"\r
```\r
\r
---\r
\r
## DOCX 高级操作\r
\r
### 合并 DOCX\r
\r
```bash\r
python -c "\r
from docx import Document\r
from docx.oxml.ns import qn\r
import sys, glob, os\r
\r
files = sorted(glob.glob(sys.argv[1]))\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'merged.docx'\r
\r
if os.path.exists(out_path):\r
    merged = Document(out_path)\r
else:\r
    merged = Document()\r
\r
for i, f in enumerate(files):\r
    if i == 0 and not os.path.exists(out_path):\r
        continue\r
    doc = Document(f)\r
    for element in doc.element.body:\r
        merged.element.body.append(element)\r
    print(f'Merged: {f}')\r
\r
merged.save(out_path)\r
print(f'Done: {len(files)} files -> {out_path}')\r
" "./*.docx" "merged.docx"\r
```\r
\r
### DOCX 提取图片\r
\r
```bash\r
python -c "\r
from docx import Document\r
import sys, os, zipfile\r
\r
docx_path = sys.argv[1]\r
out_dir = sys.argv[2] if len(sys.argv)>2 else './docx_images'\r
os.makedirs(out_dir, exist_ok=True)\r
\r
with zipfile.ZipFile(docx_path, 'r') as z:\r
    for name in z.namelist():\r
        if name.startswith('word/media/'):\r
            out = os.path.join(out_dir, os.path.basename(name))\r
            with open(out, 'wb') as f:\r
                f.write(z.read(name))\r
            print(f'Extracted: {out}')\r
\r
print('Done')\r
" "input.docx" "./docx_images"\r
```\r
\r
---\r
\r
## XLSX 高级操作\r
\r
### 合并 XLSX（按 sheet 名合并）\r
\r
```bash\r
python -c "\r
import pandas as pd, sys, glob\r
\r
files = sorted(glob.glob(sys.argv[1]))\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'merged.xlsx'\r
\r
with pd.ExcelWriter(out_path, engine='openpyxl') as writer:\r
    for f in files:\r
        xls = pd.ExcelFile(f)\r
        for sheet in xls.sheet_names:\r
            df = pd.read_excel(xls, sheet_name=sheet)\r
            sheet_name = sheet if sheet not in writer.sheets else f'{sheet}_{os.path.basename(f)}'\r
            df.to_excel(writer, sheet_name=sheet_name, index=False)\r
            print(f'Added sheet \"{sheet_name}\" from {f}')\r
\r
print(f'Done: {out_path}')\r
" "./*.xlsx" "merged.xlsx"\r
```\r
\r
### XLSX 筛选导出\r
\r
```bash\r
python -c "\r
import pandas as pd, sys\r
\r
xlsx_path = sys.argv[1]\r
out_path = sys.argv[2] if len(sys.argv)>2 else 'filtered.xlsx'\r
sheet = sys.argv[3] if len(sys.argv)>3 else 0\r
filter_expr = sys.argv[4]  # e.g. '金额 > 1000'\r
\r
df = pd.read_excel(xlsx_path, sheet_name=sheet)\r
filtered = df.query(filter_expr)\r
filtered.to_excel(out_path, index=False)\r
print(f'Done: {len(df)} -> {len(filtered)} rows')\r
" "input.xlsx" "filtered.xlsx" 0 "金额 > 1000"\r
```\r
\r
---\r
\r
## 错误处理指南\r
\r
| 错误 | 原因 | 解决方案 |\r
|------|------|----------|\r
| `ModuleNotFoundError` | 缺少 Python 依赖 | 运行 `pip install \x3C包名>` |\r
| `FileNotFoundError: soffice` | 未安装 LibreOffice | 安装 LibreOffice 并确认路径 |\r
| `pdf2docx` 转换布局错乱 | 复杂排版 PDF | 改用 PDF→图片→PPTX 方案 |\r
| 中文乱码 | 编码问题 | CSV 使用 `utf-8-sig`；确保系统有中文字体 |\r
| `poppler not found` | pdf2image 需要 poppler | 改用 PyMuPDF（`fitz`），不依赖 poppler |\r
| 大文件内存不足 | 文件过大 | 使用逐页处理，避免一次性加载 |\r
\r
## LibreOffice 路径参考\r
\r
| 系统 | 默认路径 |\r
|------|----------|\r
| Windows | `C:\Program Files\LibreOffice\program\soffice.exe` |\r
| macOS | `/Applications/LibreOffice.app/Contents/MacOS/soffice` |\r
| Linux | `/usr/bin/soffice` 或 `/usr/bin/libreoffice` |\r
\r
> **提示**：如果 LibreOffice 不在默认路径，可通过 `where soffice`（Windows）或 `which soffice`（Linux/macOS）查找实际路径，替换脚本中的路径即可。\r
\r
## 脚本方式（推荐）\r
\r
除了上面的内联 `python -c` 命令，所有转换操作也提供了独立脚本，位于 `scripts/` 目录。脚本方式更易读、更易调试，适合复杂参数场景。\r
\r
**脚本根目录：** `workspace/skills/pdf-ppt-docx-xlsx/scripts/`\r
\r
### 脚本清单\r
\r
| 脚本 | 功能 | 用法示例 |\r
|------|------|----------|\r
| `pdf_to_images.py` | PDF → 图片 | `python scripts/pdf_to_images.py input.pdf ./out 200 png` |\r
| `pdf_to_docx.py` | PDF → DOCX | `python scripts/pdf_to_docx.py input.pdf output.docx --start 0 --end 5` |\r
| `pdf_to_pptx.py` | PDF → PPTX | `python scripts/pdf_to_pptx.py input.pdf output.pptx 200` |\r
| `pdf_to_xlsx.py` | PDF → XLSX (提取表格) | `python scripts/pdf_to_xlsx.py input.pdf output.xlsx` |\r
| `pdf_to_txt.py` | PDF → 纯文本 | `python scripts/pdf_to_txt.py input.pdf output.txt` |\r
| `pdf_advanced.py` | PDF 合并/拆分/水印/提取/信息 | `python scripts/pdf_advanced.py merge a.pdf b.pdf out.pdf` |\r
| `docx_to_pdf.py` | DOCX → PDF | `python scripts/docx_to_pdf.py input.docx` |\r
| `docx_to_pptx.py` | DOCX → PPTX | `python scripts/docx_to_pptx.py input.docx output.pptx` |\r
| `docx_to_html.py` | DOCX → HTML | `python scripts/docx_to_html.py input.docx output.html` |\r
| `docx_advanced.py` | DOCX 合并/提取文本/信息 | `python scripts/docx_advanced.py merge a.docx b.docx out.docx` |\r
| `pptx_to_pdf.py` | PPTX → PDF | `python scripts/pptx_to_pdf.py input.pptx` |\r
| `pptx_to_images.py` | PPTX → 图片 | `python scripts/pptx_to_images.py input.pptx ./out 200` |\r
| `pptx_to_docx.py` | PPTX → DOCX | `python scripts/pptx_to_docx.py input.pptx output.docx` |\r
| `xlsx_to_pdf.py` | XLSX → PDF | `python scripts/xlsx_to_pdf.py input.xlsx` |\r
| `xlsx_to_csv.py` | XLSX → CSV | `python scripts/xlsx_to_csv.py input.xlsx ./csv_out` |\r
| `xlsx_advanced.py` | XLSX 合并/信息/摘要 | `python scripts/xlsx_advanced.py info input.xlsx` |\r
| `images_to_pdf.py` | 图片 → PDF | `python scripts/images_to_pdf.py a.png b.jpg out.pdf` |\r
\r
### `pdf_advanced.py` 子命令\r
\r
```bash\r
# 合并\r
python scripts/pdf_advanced.py merge file1.pdf file2.pdf merged.pdf\r
\r
# 拆分（全部页）\r
python scripts/pdf_advanced.py split input.pdf ./split_output\r
\r
# 拆分（指定范围）\r
python scripts/pdf_advanced.py split input.pdf ./split_output --range 0-5,8-10\r
\r
# 提取指定页\r
python scripts/pdf_advanced.py extract input.pdf output.pdf --pages 0,2,5-8\r
\r
# 加水印\r
python scripts/pdf_advanced.py watermark input.pdf "机密" watermarked.pdf --opacity 0.2 --size 60 --color #ff0000\r
\r
# 查看信息\r
python scripts/pdf_advanced.py info input.pdf\r
```\r
\r
### `docx_advanced.py` 子命令\r
\r
```bash\r
# 合并\r
python scripts/docx_advanced.py merge a.docx b.docx merged.docx\r
\r
# 提取文本\r
python scripts/docx_advanced.py extract_text input.docx output.txt\r
\r
# 查看信息\r
python scripts/docx_advanced.py info input.docx\r
```\r
\r
### `xlsx_advanced.py` 子命令\r
\r
```bash\r
# 合并\r
python scripts/xlsx_advanced.py merge a.xlsx b.xlsx merged.xlsx\r
\r
# 查看信息\r
python scripts/xlsx_advanced.py info input.xlsx\r
\r
# 生成摘要（前5行预览）\r
python scripts/xlsx_advanced.py summary input.xlsx summary.txt\r
```\r
\r
---\r
\r
## 使用原则\r
\r
1. **优先使用 Python 库方案**（PyMuPDF、pdf2docx 等），不依赖 LibreOffice 的转换更快更可控\r
2. **Office 格式转 PDF 时使用 LibreOffice**，保真度远优于纯 Python 方案\r
3. **大文件分页处理**，避免内存溢出\r
4. **输出路径使用绝对路径或明确的相对路径**，避免歧义\r
5. **转换前检查文件是否存在**，转换后验证输出文件大小非零\r
6. **简单场景用内联命令，复杂场景用 scripts/ 脚本**——脚本支持更多参数且更易调试\r
\x3C/task_progress>