pdf_read.py 5.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153
  1. #!/usr/bin/env python3
  2. """Read a PDF: per-page text, tables, metadata, or form fields. JSON to stdout."""
  3. from __future__ import annotations
  4. import argparse
  5. import csv
  6. import json
  7. import os
  8. import sys
  9. def _reconfigure_stdio() -> None:
  10. for stream in (sys.stdout, sys.stderr):
  11. try:
  12. stream.reconfigure(encoding="utf-8")
  13. except Exception:
  14. pass
  15. def _need(module: str, package: str):
  16. try:
  17. return __import__(module)
  18. except ImportError:
  19. print(f"Missing dependency: '{package}' is not importable here; rerun with the documents interpreter $PY from the pdf skill's Prerequisites setup block", file=sys.stderr)
  20. raise SystemExit(2)
  21. def read_text(path: str, password: str | None) -> dict:
  22. pdfplumber = _need("pdfplumber", "pdfplumber")
  23. pages = []
  24. with pdfplumber.open(path, password=password) as pdf:
  25. for page in pdf.pages:
  26. pages.append(page.extract_text() or "")
  27. return {"page_count": len(pages), "pages": pages}
  28. def read_tables(path: str, password: str | None, csv_dir: str | None) -> dict:
  29. pdfplumber = _need("pdfplumber", "pdfplumber")
  30. result = []
  31. written = []
  32. with pdfplumber.open(path, password=password) as pdf:
  33. for pageno, page in enumerate(pdf.pages, start=1):
  34. for tidx, table in enumerate(page.extract_tables()):
  35. result.append({"page": pageno, "index": tidx, "rows": table})
  36. if csv_dir:
  37. os.makedirs(csv_dir, exist_ok=True)
  38. csv_path = os.path.join(csv_dir, f"page{pageno}_table{tidx}.csv")
  39. with open(csv_path, "w", encoding="utf-8", newline="") as fh:
  40. csv.writer(fh).writerows([[c if c is not None else "" for c in row] for row in table])
  41. written.append(csv_path)
  42. out = {"table_count": len(result), "tables": result}
  43. if csv_dir:
  44. out["csv_files"] = written
  45. return out
  46. def read_meta(path: str, password: str | None) -> dict:
  47. pypdf = _need("pypdf", "pypdf")
  48. reader = pypdf.PdfReader(path)
  49. encrypted = reader.is_encrypted
  50. if encrypted:
  51. if password is None or not reader.decrypt(password):
  52. return {"encrypted": True, "note": "Provide --password to read metadata of an encrypted file."}
  53. meta = {}
  54. if reader.metadata:
  55. for key, value in reader.metadata.items():
  56. meta[str(key).lstrip("/")] = str(value)
  57. pages = []
  58. for idx, page in enumerate(reader.pages, start=1):
  59. box = page.mediabox
  60. pages.append({
  61. "page": idx,
  62. "width": float(box.width),
  63. "height": float(box.height),
  64. "rotation": int(page.get("/Rotate", 0)),
  65. })
  66. # scanned-page heuristic: no extractable text but page has images
  67. likely_scanned = []
  68. try:
  69. pdfplumber = _need("pdfplumber", "pdfplumber")
  70. with pdfplumber.open(path, password=password) as pdf:
  71. for pageno, page in enumerate(pdf.pages, start=1):
  72. text = (page.extract_text() or "").strip()
  73. if not text and page.images:
  74. likely_scanned.append(pageno)
  75. except SystemExit:
  76. raise
  77. except Exception as exc: # pragma: no cover - heuristic only
  78. print(f"Warning: scanned-page check failed: {exc}", file=sys.stderr)
  79. out = {
  80. "encrypted": encrypted,
  81. "page_count": len(reader.pages),
  82. "metadata": meta,
  83. "pages": pages,
  84. "likely_scanned_pages": likely_scanned,
  85. }
  86. if likely_scanned:
  87. out["note"] = ("Image-only pages detected: no text layer to extract. "
  88. "Use the references/ocr-extraction.md in this skill for OCR.")
  89. return out
  90. FIELD_TYPES = {"/Tx": "text", "/Btn": "button", "/Ch": "choice", "/Sig": "signature"}
  91. def read_fields(path: str, password: str | None) -> dict:
  92. pypdf = _need("pypdf", "pypdf")
  93. reader = pypdf.PdfReader(path)
  94. if reader.is_encrypted:
  95. if password is None or not reader.decrypt(password):
  96. print("File is encrypted; pass --password.", file=sys.stderr)
  97. raise SystemExit(3)
  98. fields = reader.get_fields() or {}
  99. out = {}
  100. for name, field in fields.items():
  101. ftype = FIELD_TYPES.get(str(field.get("/FT")), str(field.get("/FT")))
  102. value = field.get("/V")
  103. states = field.get("/_States_")
  104. entry = {"type": ftype, "value": None if value is None else str(value)}
  105. if states:
  106. entry["options"] = [str(s) for s in states]
  107. out[name] = entry
  108. return {"field_count": len(out), "fields": out}
  109. def main() -> int:
  110. _reconfigure_stdio()
  111. parser = argparse.ArgumentParser(description="Extract text, tables, metadata, or form fields from a PDF.")
  112. parser.add_argument("pdf", help="Input PDF path")
  113. mode = parser.add_mutually_exclusive_group(required=True)
  114. mode.add_argument("--text", action="store_true", help="Per-page text as JSON")
  115. mode.add_argument("--tables", action="store_true", help="Tables as JSON (optionally CSV via --csv-dir)")
  116. mode.add_argument("--meta", action="store_true", help="Metadata, page sizes, encrypted/scanned flags")
  117. mode.add_argument("--fields", action="store_true", help="AcroForm fields with types and values")
  118. parser.add_argument("--csv-dir", help="Also write each table as a CSV file into this directory")
  119. parser.add_argument("--password", help="Password for encrypted PDFs")
  120. args = parser.parse_args()
  121. if args.text:
  122. result = read_text(args.pdf, args.password)
  123. elif args.tables:
  124. result = read_tables(args.pdf, args.password, args.csv_dir)
  125. elif args.meta:
  126. result = read_meta(args.pdf, args.password)
  127. else:
  128. result = read_fields(args.pdf, args.password)
  129. json.dump(result, sys.stdout, ensure_ascii=False, indent=2)
  130. print()
  131. return 0
  132. if __name__ == "__main__":
  133. sys.exit(main())