pdf

1#!/usr/bin/env python32"""Read a PDF: per-page text, tables, metadata, or form fields. JSON to stdout."""3from __future__ import annotations4 5import argparse6import csv7import json8import os9import sys10 11 12def _reconfigure_stdio() -> None:13    for stream in (sys.stdout, sys.stderr):14        try:15            stream.reconfigure(encoding="utf-8")16        except Exception:17            pass18 19 20def _need(module: str, package: str):21    try:22        return __import__(module)23    except ImportError:24        print(f"Missing dependency: install with 'python3 -m pip install {package}'", file=sys.stderr)25        raise SystemExit(2)26 27 28def read_text(path: str, password: str | None) -> dict:29    pdfplumber = _need("pdfplumber", "pdfplumber")30    pages = []31    with pdfplumber.open(path, password=password) as pdf:32        for page in pdf.pages:33            pages.append(page.extract_text() or "")34    return {"page_count": len(pages), "pages": pages}35 36 37def read_tables(path: str, password: str | None, csv_dir: str | None) -> dict:38    pdfplumber = _need("pdfplumber", "pdfplumber")39    result = []40    written = []41    with pdfplumber.open(path, password=password) as pdf:42        for pageno, page in enumerate(pdf.pages, start=1):43            for tidx, table in enumerate(page.extract_tables()):44                result.append({"page": pageno, "index": tidx, "rows": table})45                if csv_dir:46                    os.makedirs(csv_dir, exist_ok=True)47                    csv_path = os.path.join(csv_dir, f"page{pageno}_table{tidx}.csv")48                    with open(csv_path, "w", encoding="utf-8", newline="") as fh:49                        csv.writer(fh).writerows([[c if c is not None else "" for c in row] for row in table])50                    written.append(csv_path)51    out = {"table_count": len(result), "tables": result}52    if csv_dir:53        out["csv_files"] = written54    return out55 56 57def read_meta(path: str, password: str | None) -> dict:58    pypdf = _need("pypdf", "pypdf")59    reader = pypdf.PdfReader(path)60    encrypted = reader.is_encrypted61    if encrypted:62        if password is None or not reader.decrypt(password):63            return {"encrypted": True, "note": "Provide --password to read metadata of an encrypted file."}64    meta = {}65    if reader.metadata:66        for key, value in reader.metadata.items():67            meta[str(key).lstrip("/")] = str(value)68    pages = []69    for idx, page in enumerate(reader.pages, start=1):70        box = page.mediabox71        pages.append({72            "page": idx,73            "width": float(box.width),74            "height": float(box.height),75            "rotation": int(page.get("/Rotate", 0)),76        })77    # scanned-page heuristic: no extractable text but page has images78    likely_scanned = []79    try:80        pdfplumber = _need("pdfplumber", "pdfplumber")81        with pdfplumber.open(path, password=password) as pdf:82            for pageno, page in enumerate(pdf.pages, start=1):83                text = (page.extract_text() or "").strip()84                if not text and page.images:85                    likely_scanned.append(pageno)86    except SystemExit:87        raise88    except Exception as exc:  # pragma: no cover - heuristic only89        print(f"Warning: scanned-page check failed: {exc}", file=sys.stderr)90    out = {91        "encrypted": encrypted,92        "page_count": len(reader.pages),93        "metadata": meta,94        "pages": pages,95        "likely_scanned_pages": likely_scanned,96    }97    if likely_scanned:98        out["note"] = ("Image-only pages detected: no text layer to extract. "99                       "Use the references/ocr-extraction.md in this skill for OCR.")100    return out101 102 103FIELD_TYPES = {"/Tx": "text", "/Btn": "button", "/Ch": "choice", "/Sig": "signature"}104 105 106def read_fields(path: str, password: str | None) -> dict:107    pypdf = _need("pypdf", "pypdf")108    reader = pypdf.PdfReader(path)109    if reader.is_encrypted:110        if password is None or not reader.decrypt(password):111            print("File is encrypted; pass --password.", file=sys.stderr)112            raise SystemExit(3)113    fields = reader.get_fields() or {}114    out = {}115    for name, field in fields.items():116        ftype = FIELD_TYPES.get(str(field.get("/FT")), str(field.get("/FT")))117        value = field.get("/V")118        states = field.get("/_States_")119        entry = {"type": ftype, "value": None if value is None else str(value)}120        if states:121            entry["options"] = [str(s) for s in states]122        out[name] = entry123    return {"field_count": len(out), "fields": out}124 125 126def main() -> int:127    _reconfigure_stdio()128    parser = argparse.ArgumentParser(description="Extract text, tables, metadata, or form fields from a PDF.")129    parser.add_argument("pdf", help="Input PDF path")130    mode = parser.add_mutually_exclusive_group(required=True)131    mode.add_argument("--text", action="store_true", help="Per-page text as JSON")132    mode.add_argument("--tables", action="store_true", help="Tables as JSON (optionally CSV via --csv-dir)")133    mode.add_argument("--meta", action="store_true", help="Metadata, page sizes, encrypted/scanned flags")134    mode.add_argument("--fields", action="store_true", help="AcroForm fields with types and values")135    parser.add_argument("--csv-dir", help="Also write each table as a CSV file into this directory")136    parser.add_argument("--password", help="Password for encrypted PDFs")137    args = parser.parse_args()138 139    if args.text:140        result = read_text(args.pdf, args.password)141    elif args.tables:142        result = read_tables(args.pdf, args.password, args.csv_dir)143    elif args.meta:144        result = read_meta(args.pdf, args.password)145    else:146        result = read_fields(args.pdf, args.password)147    json.dump(result, sys.stdout, ensure_ascii=False, indent=2)148    print()149    return 0150 151 152if __name__ == "__main__":153    sys.exit(main())154