pdf-ocr-hotfolder/pdf_ocr_hotfolder/processor.py

"""OCR-Verarbeitung einer einzelnen PDF mit ocrmypdf + optional veraPDF."""
from __future__ import annotations

import logging
import shutil
import subprocess
from dataclasses import dataclass
from pathlib import Path

from .config import OcrConfig, OutputConfig, VeraPdfConfig

log = logging.getLogger(__name__)


def build_output_name(src_name: str, mode: str, tag: str) -> str:
    """Erzeugt den Ziel-Dateinamen für ein OCR-PDF.

    Args:
        src_name: Original-Dateiname (z.B. "scan.pdf")
        mode: "prefix" | "suffix" | "none"
        tag: Einzufügender String (verbatim, leer = kein Tag)

    Beispiele:
        prefix "OCR_":  "scan.pdf"  -> "OCR_scan.pdf"
        suffix "_OCR":  "scan.pdf"  -> "scan_OCR.pdf"
        suffix "_OCR":  "scan.tar.gz.pdf" -> "scan.tar.gz_OCR.pdf"
        none:           "scan.pdf"  -> "scan.pdf"
    """
    if mode == "none" or not tag:
        return src_name
    if mode == "prefix":
        return f"{tag}{src_name}"
    if mode == "suffix":
        # Nur die letzte Extension abspalten, sonst "foo.bar.pdf" kaputt gemacht
        p = Path(src_name)
        stem, ext = p.stem, p.suffix
        return f"{stem}{tag}{ext}"
    raise ValueError(f"Unbekannter name_mode: {mode!r}")


@dataclass
class ProcessResult:
    source: Path
    output: Path
    success: bool
    error: str = ""
    verapdf_passed: bool | None = None


def run_ocr(src: Path, dst: Path, cfg: OcrConfig) -> None:
    """Führt ocrmypdf als Library-Call aus (kein Subprozess-Overhead)."""
    import ocrmypdf  # lazy, damit Tests ohne ocrmypdf laufen

    kwargs: dict = {
        "language": cfg.languages,
        "jobs": cfg.jobs,
        "deskew": cfg.deskew,
        "clean": cfg.clean,
        "oversample": cfg.oversample,
        "progress_bar": False,
        "skip_text": cfg.skip_text,
    }
    if cfg.pdfa_level:
        kwargs["output_type"] = f"pdfa-{cfg.pdfa_level}"
    else:
        kwargs["output_type"] = "pdf"

    log.info("OCR start: %s", src.name)
    ocrmypdf.ocr(str(src), str(dst), **kwargs)
    log.info("OCR done:  %s", dst.name)


def run_verapdf(pdf: Path, cfg: VeraPdfConfig) -> bool:
    """Validiert PDF/A mit veraPDF (CLI). Gibt True zurück, wenn konform."""
    if not cfg.enabled:
        return True
    if not Path(cfg.binary).exists():
        log.warning("veraPDF binary nicht gefunden: %s", cfg.binary)
        return False
    try:
        result = subprocess.run(
            [cfg.binary, "--flavour", cfg.flavour, "--format", "text", str(pdf)],
            capture_output=True, text=True, timeout=300,
        )
        ok = result.returncode == 0 and "PASS" in result.stdout
        log.info("veraPDF %s: %s", "PASS" if ok else "FAIL", pdf.name)
        return ok
    except subprocess.TimeoutExpired:
        log.error("veraPDF Timeout: %s", pdf.name)
        return False


def process_pdf(
    src: Path,
    working_dir: Path,
    outgoing_dir: Path,
    error_dir: Path,
    ocr_cfg: OcrConfig,
    vera_cfg: VeraPdfConfig,
    output_cfg: OutputConfig,
) -> ProcessResult:
    """Verarbeitet eine einzelne PDF: move→OCR→validate→outgoing/error."""
    out_name = build_output_name(src.name, output_cfg.name_mode, output_cfg.name_tag)
    work_src = working_dir / src.name
    work_out = working_dir / f"__ocr_{out_name}"  # Temp-Name, damit er != src.name ist
    final_out = outgoing_dir / out_name

    try:
        shutil.move(str(src), str(work_src))
    except OSError as e:
        return ProcessResult(src, final_out, False, f"move to working failed: {e}")

    try:
        run_ocr(work_src, work_out, ocr_cfg)
    except Exception as e:  # noqa: BLE001 - ocrmypdf wirft viele Typen
        log.exception("OCR fehlgeschlagen für %s", src.name)
        _move_to_error(work_src, error_dir)
        return ProcessResult(src, final_out, False, f"ocr failed: {e}")

    vera_ok: bool | None = None
    if vera_cfg.enabled:
        vera_ok = run_verapdf(work_out, vera_cfg)
        if not vera_ok:
            _move_to_error(work_out, error_dir)
            work_src.unlink(missing_ok=True)
            return ProcessResult(src, final_out, False,
                                 "verapdf validation failed", verapdf_passed=False)

    outgoing_dir.mkdir(parents=True, exist_ok=True)
    shutil.move(str(work_out), str(final_out))
    _dispose_original(work_src, src.name, output_cfg)
    return ProcessResult(src, final_out, True, verapdf_passed=vera_ok)


def _dispose_original(work_src: Path, original_name: str, cfg: OutputConfig) -> None:
    """Entsorgt das Original nach erfolgreichem OCR — löschen oder archivieren."""
    if not work_src.exists():
        return
    mode = cfg.original_on_success
    if mode == "delete":
        work_src.unlink(missing_ok=True)
        return
    if mode == "archive":
        if not cfg.archive_dir:
            log.error("original_on_success=archive aber archive_dir ist leer — lösche stattdessen")
            work_src.unlink(missing_ok=True)
            return
        archive = Path(cfg.archive_dir)
        archive.mkdir(parents=True, exist_ok=True)
        dest = archive / original_name
        # Bei Namens-Kollision mit Timestamp umbenennen
        if dest.exists():
            from datetime import datetime
            ts = datetime.now().strftime("%Y%m%d-%H%M%S")
            dest = archive / f"{dest.stem}_{ts}{dest.suffix}"
        shutil.move(str(work_src), str(dest))
        log.info("Original archiviert: %s", dest)
        return
    log.warning("Unbekannter original_on_success=%r — lösche stattdessen", mode)
    work_src.unlink(missing_ok=True)


def _move_to_error(p: Path, error_dir: Path) -> None:
    error_dir.mkdir(parents=True, exist_ok=True)
    try:
        shutil.move(str(p), str(error_dir / p.name))
    except OSError:
        log.exception("Konnte %s nicht in error-Verzeichnis verschieben", p)