md_to_gost/md2gost/biblio_processor.py

"""Convert [n]: source lines under bibliography heading into Bibliography paragraphs."""

from __future__ import annotations

import re

from .bibliography import BIBLIO_LINE_RE, BiblioEntry, extract_year
from .bibliography_renderable import Bibliography
from .renderable import Renderable
from .renderable.heading import Heading
from .renderable.paragraph import Paragraph


BIBLIO_HEADING = re.compile(
    r"СПИСОК\s+ИСПОЛЬЗОВАНН?ЫХ\s+ИСТОЧНИКОВ",
    re.IGNORECASE,
)
APPENDIX_HEADING = re.compile(r"^ПРИЛОЖЕН", re.IGNORECASE)

# Marko merges consecutive [n]: lines into one paragraph (soft breaks ignored),
# so we must find every entry inside the paragraph text.
BIBLIO_FIND_RE = re.compile(
    r"\[(\d+(?:\.\d+)?)\]:\s*(.*?)(?=\s*\[\d+(?:\.\d+)?\]:|\s*$)",
    re.DOTALL,
)


def entries_from_text(text: str) -> list[BiblioEntry]:
    """Pull all [n]: … entries from a (possibly concatenated) paragraph."""
    entries: list[BiblioEntry] = []
    if not text or not text.strip():
        return entries
    for m in BIBLIO_FIND_RE.finditer(text.strip()):
        key = m.group(1)
        body = re.sub(r"\s+", " ", m.group(2)).strip()
        if not body:
            continue
        entries.append(BiblioEntry(key=key, text=body, year=extract_year(body)))
    return entries


# After citation linking, `[5]: …` often becomes `[]: …` in paragraph.text
# (the digits live in a hyperlink). Still a biblio source line to drop.
BIBLIO_RESIDUE_RE = re.compile(r"^\[\d*(?:\.\d+)?\]:\s*\S")


def is_biblio_source_paragraph(text: str) -> bool:
    """True if paragraph is a [n]: / []: bibliography source line (or empty)."""
    stripped = (text or "").strip()
    if not stripped:
        return True
    if entries_from_text(stripped):
        return True
    return bool(BIBLIO_RESIDUE_RE.match(stripped))


def extract_bibliography_from_markdown(md: str) -> list[BiblioEntry] | list[tuple[str, list[BiblioEntry]]]:
    """
    Parse bibliography from raw markdown (reliable even if Marko merges lines).
    Returns either a flat list of entries, or a list of (section_title, entries) for VKR.
    """
    m = re.search(
        r"^#\s*\*?\s*СПИСОК\s+ИСПОЛЬЗОВАНН?ЫХ\s+ИСТОЧНИКОВ\s*$",
        md,
        re.M | re.I,
    )
    if not m:
        return []
    start = m.end()
    rest = md[start:]
    next_h = re.search(r"^#\s+", rest, re.M)
    block = rest[: next_h.start()] if next_h else rest

    sections: list[tuple[str, list[BiblioEntry]]] = []
    current_title: str | None = None
    current: list[BiblioEntry] = []
    flat: list[BiblioEntry] = []

    for line in block.splitlines():
        hm = re.match(r"^#{2,6}\s+(\*?)(.+)$", line)
        if hm:
            title = hm.group(2).strip()
            if current_title is not None:
                sections.append((current_title, current))
            elif current:
                flat.extend(current)
            current_title = title
            current = []
            continue
        entries = entries_from_text(line)
        if entries:
            if current_title is not None:
                current.extend(entries)
            else:
                flat.extend(entries)
            continue
        # concatenated line with several [n]:
        if "[" in line and "]:" in line:
            more = entries_from_text(line)
            if current_title is not None:
                current.extend(more)
            else:
                flat.extend(more)

    if current_title is not None:
        sections.append((current_title, current))
    elif current:
        flat.extend(current)

    if sections:
        return sections
    return flat


def fold_bibliography(renderables: list[Renderable], parent,
                      raw_markdown: str | None = None) -> list[Renderable]:
    """Replace bibliography source paragraphs with a Bibliography renderable."""
    # Prefer raw markdown extraction (handles Marko soft-break merge)
    raw_entries = None
    if raw_markdown:
        raw_entries = extract_bibliography_from_markdown(raw_markdown)

    result: list[Renderable] = []
    i = 0
    while i < len(renderables):
        r = renderables[i]
        result.append(r)
        if isinstance(r, Heading) and BIBLIO_HEADING.search(r.text or ""):
            i += 1
            # Skip / drop following paragraphs that are biblio lines (already extracted)
            consumed = 0
            while i + consumed < len(renderables):
                item = renderables[i + consumed]
                if isinstance(item, Heading):
                    title = (item.text or "").strip()
                    if APPENDIX_HEADING.match(title):
                        break
                    if item.level == 1:
                        break
                    # subsection inside biblio — skip heading, content comes from raw
                    consumed += 1
                    continue
                if isinstance(item, Paragraph):
                    text = item._docx_paragraph.text or ""
                    if is_biblio_source_paragraph(text):
                        consumed += 1
                        continue
                    break
                break

            if raw_entries:
                if raw_entries and isinstance(raw_entries[0], tuple):
                    result.append(Bibliography(parent, [], sections=raw_entries))  # type: ignore
                else:
                    result.append(Bibliography(parent, raw_entries))  # type: ignore
            else:
                # Fallback: parse from renderable paragraphs
                entries: list[BiblioEntry] = []
                for j in range(consumed):
                    item = renderables[i + j]
                    if isinstance(item, Paragraph):
                        entries.extend(entries_from_text(item._docx_paragraph.text or ""))
                if entries:
                    result.append(Bibliography(parent, entries))

            i += consumed
            continue
        i += 1
    return result