md_to_gost/word2md/classify.py

"""Classify DOCX paragraphs/tables into IR kinds (style first, then heuristics)."""

from __future__ import annotations

import re

from md2gost.renderable.toc import SPECIAL_TITLES

from .captions import parse_any_caption

CAPTION_STYLE_KINDS = {
    "Caption Figure": "figure",
    "Название таблицы": "table",
    "Caption Table": "table",
    "Caption Listing": "listing",
    "Caption": None,  # need text heuristic
}

HEADING_STYLE_RE = re.compile(r"^(?:Heading|Заголовок)\s*(\d+)$", re.IGNORECASE)
TOC_STYLE_RE = re.compile(r"^toc\s*\d+$", re.IGNORECASE)

SPECIAL_UPPER = {t.upper() for t in SPECIAL_TITLES}
APPENDIX_RE = re.compile(
    r"^ПРИЛОЖЕНИЕ\s+([А-ЯA-ZЁ])(?:\s+(.+))?$",
    re.IGNORECASE,
)
NUMBERED_HEADING_RE = re.compile(r"^(\d+(?:\.\d+)*)\.?\s+(.+)$")
PRACTICE_RE = re.compile(r"^Практическая\s+работа\b", re.IGNORECASE)
LIST_MARKER_RE = re.compile(r"^(?:[–—\-•▪]|(\d+)\.)\s*\t?\s*(.*)$")
BIBLIO_LINE_RE = re.compile(r"^\[(\d+(?:\.\d+)?)\]\s*[:.]?\s*(.+)$")
BODY_START_TITLES = SPECIAL_UPPER | {"СОДЕРЖАНИЕ"}


def style_name(paragraph) -> str:
    try:
        if paragraph.style is None:
            return ""
        return paragraph.style.name or ""
    except Exception:
        return ""


def is_toc_style(name: str) -> bool:
    return bool(TOC_STYLE_RE.match((name or "").strip()))


def heading_level_from_style(name: str) -> int | None:
    m = HEADING_STYLE_RE.match((name or "").strip())
    if not m:
        return None
    try:
        level = int(m.group(1))
    except ValueError:
        return None
    return level if 1 <= level <= 9 else None


def is_code_style(name: str) -> bool:
    n = (name or "").strip().lower()
    return n == "code" or n.startswith("code ")


def is_bibliography_style(name: str) -> bool:
    n = (name or "").strip().lower()
    return n in ("bibliography", "bibliography heading")


def is_caption_style(name: str) -> bool:
    return (name or "").strip() in CAPTION_STYLE_KINDS


def caption_kind_from_style(name: str) -> str | None:
    return CAPTION_STYLE_KINDS.get((name or "").strip())


def is_special_title(text: str) -> bool:
    t = (text or "").strip().upper()
    if t in SPECIAL_UPPER:
        return True
    # allow «СПИСОК ИСПОЛЬЗОВАННЫХ ИСТОЧНИКОВ» variants already in set
    if t.startswith("СПИСОК ИСПОЛЬЗ"):
        return True
    return False


def is_body_start_heading(text: str, style: str = "") -> bool:
    """True if this heading marks the start of report body (skip title pages before it)."""
    t = (text or "").strip()
    if not t:
        return False
    if heading_level_from_style(style) or style.lower().startswith("heading"):
        upper = t.upper()
        if upper.lstrip("* ").upper() in BODY_START_TITLES or is_special_title(t):
            return True
        if PRACTICE_RE.match(t):
            return True
        if NUMBERED_HEADING_RE.match(t):
            return True
        if APPENDIX_RE.match(t):
            return True
        # Any Heading 1 after front matter
        if heading_level_from_style(style) == 1:
            return True
    # Heuristic without style: special titles alone
    if is_special_title(t) or PRACTICE_RE.match(t):
        return True
    return False


def classify_heading_text(text: str, style_level: int | None) -> tuple[int, str, bool]:
    """
    Return (level, title_text, numbered).
    numbered=False → emit as # *TITLE
    """
    raw = (text or "").strip()
    level = style_level or 1

    app = APPENDIX_RE.match(raw)
    if app:
        letter = app.group(1).upper()
        name = (app.group(2) or "").strip()
        title = f"Приложение {letter}" + (f" {name}" if name else "")
        return 2, title, True

    if is_special_title(raw):
        return 1, raw.upper(), False

    m = NUMBERED_HEADING_RE.match(raw)
    if m:
        # keep digits in text for manual numbering dialect
        return level, raw, True

    if PRACTICE_RE.match(raw):
        return 1, raw, True

    return level, raw, True


def look_like_list_item(text: str) -> tuple[bool, bool, str] | None:
    """
    Detect md2gost-style list lines (–\\t / 1.\\t) or plain markers.
    Returns (ordered, marker_ok, body) or None.
    """
    t = (text or "").strip()
    if not t:
        return None
    if t[0] in ("–", "—", "-", "•", "▪") and (len(t) == 1 or t[1] in "\t "):
        body = t[1:].lstrip("\t ").strip()
        return False, True, body
    m = re.match(r"^(\d+)\.\s*\t?\s*(.+)$", t)
    if m:
        return True, True, m.group(2).strip()
    return None


def parse_biblio_line(text: str) -> tuple[str, str] | None:
    m = BIBLIO_LINE_RE.match((text or "").strip())
    if not m:
        return None
    return m.group(1), m.group(2).strip()


def paragraph_has_numpr(paragraph) -> bool:
    pPr = paragraph._element.pPr
    if pPr is None:
        return False
    return pPr.numPr is not None