marki-docx/src/markidocx/differ.py

"""Structural drift detection for markidocx (FR-700)."""

from __future__ import annotations

import re
from dataclasses import dataclass, field

HEADING_RE = re.compile(r"^(#{1,6})\s+(.+)$", re.MULTILINE)
LIST_ITEM_RE = re.compile(r"^(\s*[-*+]|\s*\d+\.)\s+(.+)$", re.MULTILINE)
TABLE_ROW_RE = re.compile(r"^\|.+\|$", re.MULTILINE)
FOOTNOTE_RE = re.compile(r"\[\^[^\]]+\]")
LINK_RE = re.compile(r"\[([^\]]+)\]\(([^)]+)\)")


@dataclass
class DriftReport:
    has_drift: bool
    preserved: list[str] = field(default_factory=list)
    degraded: list[str] = field(default_factory=list)
    broken: list[str] = field(default_factory=list)
    unsupported: list[str] = field(default_factory=list)


def compare(original: str, reimported: str) -> DriftReport:
    """Compare *original* Markdown against *reimported* Markdown.

    Classifies each structural element as:
    - preserved: identical in both
    - degraded: present but modified
    - broken: present in original, missing in reimported
    - unsupported: construct not supported by the round-trip

    Returns a DriftReport.
    """
    preserved: list[str] = []
    degraded: list[str] = []
    broken: list[str] = []
    unsupported: list[str] = []

    # --- Headings (FR-501) ---
    orig_headings = _extract_headings(original)
    reim_headings = _extract_headings(reimported)
    _compare_sets("heading", orig_headings, reim_headings, preserved, degraded, broken)

    # --- Lists (FR-502) ---
    orig_lists = _extract_list_items(original)
    reim_lists = _extract_list_items(reimported)
    _compare_sets("list_item", orig_lists, reim_lists, preserved, degraded, broken)

    # --- Tables (FR-503) ---
    orig_tables = _count_tables(original)
    reim_tables = _count_tables(reimported)
    if orig_tables == reim_tables:
        if orig_tables > 0:
            preserved.append(f"tables:{orig_tables}")
    elif reim_tables < orig_tables:
        broken.append(f"tables:missing {orig_tables - reim_tables} of {orig_tables}")
    else:
        degraded.append(f"tables:count changed {orig_tables}→{reim_tables}")

    # --- Footnotes (FR-504) ---
    orig_fn = set(FOOTNOTE_RE.findall(original))
    reim_fn = set(FOOTNOTE_RE.findall(reimported))
    for fn in orig_fn:
        if fn in reim_fn:
            preserved.append(f"footnote:{fn}")
        else:
            broken.append(f"footnote:{fn}")

    # --- Links (FR-506) ---
    orig_links = {m.group(0) for m in LINK_RE.finditer(original)}
    reim_links = {m.group(0) for m in LINK_RE.finditer(reimported)}
    for link in orig_links:
        if link in reim_links:
            preserved.append(f"link:{link[:40]}")
        else:
            degraded.append(f"link:lost {link[:40]}")

    has_drift = bool(degraded or broken)
    return DriftReport(
        has_drift=has_drift,
        preserved=preserved,
        degraded=degraded,
        broken=broken,
        unsupported=unsupported,
    )


def _extract_headings(text: str) -> list[str]:
    return [f"{'#' * len(m.group(1))} {m.group(2).strip()}" for m in HEADING_RE.finditer(text)]


def _extract_list_items(text: str) -> list[str]:
    return [m.group(2).strip() for m in LIST_ITEM_RE.finditer(text)]


def _count_tables(text: str) -> int:
    rows = TABLE_ROW_RE.findall(text)
    if not rows:
        return 0
    # Count separator rows as table boundaries
    sep_re = re.compile(r"^\|[-| :]+\|$")
    count = sum(1 for r in rows if sep_re.match(r))
    return count


def _compare_sets(
    kind: str,
    orig: list[str],
    reim: list[str],
    preserved: list[str],
    degraded: list[str],
    broken: list[str],
) -> None:
    orig_counts: dict[str, int] = {}
    for item in orig:
        orig_counts[item] = orig_counts.get(item, 0) + 1

    reim_counts: dict[str, int] = {}
    for item in reim:
        reim_counts[item] = reim_counts.get(item, 0) + 1

    for item, count in orig_counts.items():
        reim_count = reim_counts.get(item, 0)
        if reim_count >= count:
            preserved.append(f"{kind}:{item[:60]}")
        elif reim_count > 0:
            degraded.append(f"{kind}:partial '{item[:60]}' ({reim_count}/{count})")
        else:
            broken.append(f"{kind}:missing '{item[:60]}'")