pdf

1#!/usr/bin/env python32"""Extract text from documents using pymupdf. Lightweight (~25MB), no models.3 4Usage:5    python extract_pymupdf.py document.pdf6    python extract_pymupdf.py document.pdf --markdown7    python extract_pymupdf.py document.pdf --pages 0-48    python extract_pymupdf.py document.pdf --images output_dir/9    python extract_pymupdf.py document.pdf --tables10    python extract_pymupdf.py document.pdf --metadata11"""12import sys13import json14 15def extract_text(path, pages=None):16    import pymupdf17    doc = pymupdf.open(path)18    page_range = range(len(doc)) if pages is None else pages19    for i in page_range:20        if i < len(doc):21            print(f"\n--- Page {i+1}/{len(doc)} ---\n")22            print(doc[i].get_text())23 24def extract_markdown(path, pages=None):25    import pymupdf4llm26    md = pymupdf4llm.to_markdown(path, pages=pages)27    print(md)28 29def extract_tables(path):30    import pymupdf31    doc = pymupdf.open(path)32    for i, page in enumerate(doc):33        tables = page.find_tables()34        for j, table in enumerate(tables.tables):35            print(f"\n--- Page {i+1}, Table {j+1} ---\n")36            df = table.to_pandas()37            print(df.to_markdown(index=False))38 39def extract_images(path, output_dir):40    import pymupdf41    from pathlib import Path42    Path(output_dir).mkdir(parents=True, exist_ok=True)43    doc = pymupdf.open(path)44    count = 045    for i, page in enumerate(doc):46        for img_idx, img in enumerate(page.get_images(full=True)):47            xref = img[0]48            pix = pymupdf.Pixmap(doc, xref)49            if pix.n >= 5:50                pix = pymupdf.Pixmap(pymupdf.csRGB, pix)51            out_path = f"{output_dir}/page{i+1}_img{img_idx+1}.png"52            pix.save(out_path)53            count += 154    print(f"Extracted {count} images to {output_dir}/")55 56def show_metadata(path):57    import pymupdf58    doc = pymupdf.open(path)59    print(json.dumps({60        "pages": len(doc),61        "title": doc.metadata.get("title", ""),62        "author": doc.metadata.get("author", ""),63        "subject": doc.metadata.get("subject", ""),64        "creator": doc.metadata.get("creator", ""),65        "producer": doc.metadata.get("producer", ""),66        "format": doc.metadata.get("format", ""),67    }, indent=2))68 69if __name__ == "__main__":70    import argparse71 72    parser = argparse.ArgumentParser(73        description="Extract text/tables/images/metadata from documents using pymupdf (lightweight, no models)."74    )75    parser.add_argument("path", help="Document to read")76    parser.add_argument("--pages", help="Page selection: N or START-END (0-indexed)")77    parser.add_argument("--markdown", action="store_true", help="Markdown output via pymupdf4llm")78    parser.add_argument("--tables", action="store_true", help="Extract tables as markdown")79    parser.add_argument("--images", nargs="?", const="./images", metavar="OUTPUT_DIR",80                        help="Extract embedded images to OUTPUT_DIR (default ./images)")81    parser.add_argument("--metadata", action="store_true", help="Show document metadata as JSON")82    args = parser.parse_args()83 84    pages = None85    if args.pages:86        if "-" in args.pages:87            start, end = args.pages.split("-")88            pages = list(range(int(start), int(end) + 1))89        else:90            pages = [int(args.pages)]91 92    if args.metadata:93        show_metadata(args.path)94    elif args.tables:95        extract_tables(args.path)96    elif args.images is not None:97        extract_images(args.path, args.images)98    elif args.markdown:99        extract_markdown(args.path, pages=pages)100    else:101        extract_text(args.path, pages=pages)102