md_to_gost/md2gost/office/detect.py

"""Detect installed Word / LibreOffice / OpenOffice for post-save table splits."""

from __future__ import annotations

import os
import shutil
import sys
from dataclasses import dataclass
from pathlib import Path

from .uno_cli import soffice_cli_path

OFFICE_ENGINE_MODES = ("auto", "word", "libre", "openoffice")
DEFAULT_OFFICE_ENGINE = "auto"


@dataclass(frozen=True)
class OfficeInstall:
    engine: str  # "word" | "libre" | "openoffice"
    label: str
    soffice_path: str | None = None
    python_path: str | None = None


# First successful ``auto`` resolution is reused for the process lifetime so a
# second convert does not re-probe Word and fall through to LibreOffice.
_cached_auto_resolve: tuple[str, OfficeInstall] | None = None


def _program_dirs(kind: str) -> list[Path]:
    roots: list[Path] = []
    for env in ("ProgramFiles", "ProgramFiles(x86)"):
        base = os.environ.get(env)
        if base:
            roots.append(Path(base))
    if sys.platform != "win32":
        roots.extend(
            [
                Path("/usr/lib/libreoffice"),
                Path("/usr/lib64/libreoffice"),
                Path("/opt/libreoffice"),
                Path("/opt/openoffice4"),
                Path("/Applications/LibreOffice.app/Contents"),
            ]
        )
    candidates: list[Path] = []
    if kind == "libre":
        for root in roots:
            candidates.extend(
                [
                    root / "LibreOffice" / "program",
                    root / "libreoffice" / "program",
                    root / "program",
                    root / "MacOS",
                ]
            )
    else:
        for root in roots:
            candidates.extend(
                [
                    root / "OpenOffice 4" / "program",
                    root / "openoffice4" / "program",
                ]
            )
    return candidates


def _soffice_names() -> tuple[str, ...]:
    if sys.platform == "win32":
        # soffice.com = CLI/headless; soffice.exe opens the Start Center GUI.
        return ("soffice.com", "soffice.exe")
    return ("soffice", "soffice.bin")


def _python_names() -> tuple[str, ...]:
    if sys.platform == "win32":
        return ("python.exe",)
    return ("python", "python3")


def _find_soffice_install(kind: str) -> OfficeInstall | None:
    label = "LibreOffice" if kind == "libre" else "OpenOffice"
    seen: set[str] = set()

    for directory in _program_dirs(kind):
        for soffice_name in _soffice_names():
            soffice = directory / soffice_name
            key = str(soffice.resolve()) if soffice.exists() else ""
            if not key or key in seen:
                continue
            seen.add(key)
            python_path = None
            for py_name in _python_names():
                candidate = directory / py_name
                if candidate.is_file():
                    python_path = str(candidate)
                    break
            if python_path is None:
                continue
            return OfficeInstall(
                engine=kind,
                label=label,
                soffice_path=soffice_cli_path(str(soffice)),
                python_path=python_path,
            )

    which = shutil.which("soffice") or shutil.which("soffice.bin")
    if which:
        program_dir = Path(which).resolve().parent
        python_path = None
        for py_name in _python_names():
            candidate = program_dir / py_name
            if candidate.is_file():
                python_path = str(candidate)
                break
        if python_path:
            lower = str(program_dir).lower()
            detected = "openoffice" if "openoffice" in lower else "libre"
            if kind != detected:
                return None
            return OfficeInstall(
                engine=kind,
                label=label,
                soffice_path=soffice_cli_path(str(Path(which).resolve())),
                python_path=python_path,
            )
    return None


def detect_office_install(engine: str) -> OfficeInstall | None:
    """Return install info for a concrete engine id."""
    if engine == "word":
        return _detect_word()
    if engine == "libre":
        return _find_soffice_install("libre")
    if engine == "openoffice":
        return _find_soffice_install("openoffice")
    return None


def _word_com_available() -> bool:
    """True when pywin32 is importable and Word.Application is registered."""
    if sys.platform != "win32":
        return False
    try:
        import win32com.client  # type: ignore  # noqa: F401
    except ImportError:
        return False
    try:
        import winreg

        with winreg.OpenKey(winreg.HKEY_CLASSES_ROOT, r"Word.Application"):
            pass
        return True
    except OSError:
        return False


def _detect_word() -> OfficeInstall | None:
    if not _word_com_available():
        return None
    return OfficeInstall(engine="word", label="Microsoft Word")


def _install_still_available(install: OfficeInstall) -> bool:
    if install.engine == "word":
        return _word_com_available()
    if install.soffice_path and install.python_path:
        return Path(install.soffice_path).is_file() and Path(install.python_path).is_file()
    return False


def clear_office_engine_cache() -> None:
    """Drop cached ``auto`` resolution (tests / long-running hosts)."""
    global _cached_auto_resolve
    _cached_auto_resolve = None


def resolve_office_engine(mode: str) -> tuple[str | None, OfficeInstall | None]:
    """
    Resolve ``mode`` to a concrete engine id and optional install record.

    ``auto`` tries Word, then LibreOffice, then OpenOffice. The first successful
    ``auto`` pick is cached for subsequent calls in the same process.
    """
    global _cached_auto_resolve

    if mode not in OFFICE_ENGINE_MODES:
        mode = DEFAULT_OFFICE_ENGINE

    if mode == "auto":
        if _cached_auto_resolve is not None:
            cached_engine, cached_install = _cached_auto_resolve
            if _install_still_available(cached_install):
                return _cached_auto_resolve
            _cached_auto_resolve = None

        for candidate in ("word", "libre", "openoffice"):
            install = detect_office_install(candidate)
            if install is not None:
                _cached_auto_resolve = (candidate, install)
                return candidate, install
        _cached_auto_resolve = None
        return None, None

    install = detect_office_install(mode)
    if install is None:
        return None, None
    return mode, install