#!/usr/bin/env python3
"""Convert the rendered delivery package from PDF to Word.

The work products are authored in Markdown and rendered to PDF by
render-sdlc.mjs; the .docx beside each .pdf is a layout reconstruction of that
PDF, so the Word file keeps the page the reviewer signed off rather than being
laid out a second time from the source.

Requires pdf2docx:  pip install -r requirements.txt

Usage:
    convert-to-docx.py DELIVERY_DIRECTORY [PDF…]
"""
import sys
import os
from pathlib import Path

from pdf2docx import Converter


def convert(pdf: Path) -> Path:
    docx = pdf.with_suffix('.docx')
    converter = Converter(str(pdf))
    try:
        converter.convert(str(docx))
    finally:
        converter.close()
    return docx


def main(argv):
    if not argv:
        raise SystemExit(__doc__)

    root = Path(argv[0])
    pdfs = [Path(p) for p in argv[1:]] or sorted(root.rglob('*.pdf'))
    if not pdfs:
        raise SystemExit(f'No PDFs found under {root}')

    failures = []
    for index, pdf in enumerate(pdfs, start=1):
        try:
            docx = convert(pdf)
            print(f'[{index}/{len(pdfs)}] {docx.relative_to(root)}')
        except Exception as error:                     # noqa: BLE001 - reported, not swallowed
            failures.append((pdf, error))
            print(f'[{index}/{len(pdfs)}] FAILED {pdf.relative_to(root)}: {error}', file=sys.stderr)

    print(f'{len(pdfs) - len(failures)}/{len(pdfs)} documents converted')
    if failures:
        raise SystemExit(1)


if __name__ == '__main__':
    # pdf2docx is chatty; the per-document progress above is enough.
    os.environ.setdefault('PDF2DOCX_QUIET', '1')
    main(sys.argv[1:])