"""
Words-First Grid Builder (Bottom-Up).

Builds a cell grid from Tesseract word_boxes directly, without requiring
pre-detected columns or rows.  Algorithm:

  1. Cluster words into columns by X-gap analysis
  2. Cluster words into rows by Y-proximity
  3. Build cells at (column, row) intersections

Returns the same (cells, columns_meta) format as build_cell_grid_v2().

Lizenz: Apache 2.0 (kommerziell nutzbar)
DATENSCHUTZ: Alle Verarbeitung erfolgt lokal.
"""

import logging
import re
import statistics
from typing import Any, Dict, List, Tuple

from cv_ocr_engines import (
    _group_words_into_lines,
    _words_to_reading_order_text,
)

logger = logging.getLogger(__name__)


# ---------------------------------------------------------------------------
# 1. Column clustering
# ---------------------------------------------------------------------------

def _cluster_columns(
    words: List[Dict],
    img_w: int,
    min_gap_pct: float = 3.0,
) -> List[Dict[str, Any]]:
    """Cluster words into columns by finding large horizontal gaps.

    Returns a list of column dicts:
        [{'index': 0, 'type': 'column_1', 'x_min': ..., 'x_max': ...}, ...]
    sorted left-to-right.
    """
    if not words:
        return []

    # Sort by X center
    sorted_w = sorted(words, key=lambda w: w['left'] + w['width'] / 2)

    # Collect word heights to compute adaptive threshold
    heights = [w['height'] for w in sorted_w if w.get('height', 0) > 0]
    median_h = statistics.median(heights) if heights else 30

    # Adaptive gap threshold: 3× median word height, but at least min_gap_pct of image width
    min_gap_px = max(median_h * 3, img_w * min_gap_pct / 100) if img_w > 0 else median_h * 3

    # Find X-gap boundaries between consecutive words (sorted by X-center)
    # For each word, compute right edge; for next word, compute left edge
    boundaries: List[float] = []  # X positions where columns split
    for i in range(len(sorted_w) - 1):
        right_edge = sorted_w[i]['left'] + sorted_w[i]['width']
        left_edge = sorted_w[i + 1]['left']
        gap = left_edge - right_edge
        if gap > min_gap_px:
            # Split point is midway through the gap
            boundaries.append((right_edge + left_edge) / 2)

    # Build column ranges from boundaries
    # Column ranges: (-inf, boundary[0]), (boundary[0], boundary[1]), ..., (boundary[-1], +inf)
    col_edges = [0.0] + boundaries + [float(img_w)]
    columns = []
    for ci in range(len(col_edges) - 1):
        columns.append({
            'index': ci,
            'type': f'column_{ci + 1}' if len(col_edges) > 2 else 'column_text',
            'x_min': col_edges[ci],
            'x_max': col_edges[ci + 1],
        })

    return columns


# ---------------------------------------------------------------------------
# 2. Row clustering
# ---------------------------------------------------------------------------

def _cluster_rows(
    words: List[Dict],
) -> List[Dict[str, Any]]:
    """Cluster words into visual rows by Y-proximity.

    Uses half the median word height as Y-tolerance.

    Returns a list of row dicts:
        [{'index': 0, 'y_min': ..., 'y_max': ..., 'y_center': ...}, ...]
    sorted top-to-bottom.
    """
    if not words:
        return []

    heights = [w['height'] for w in words if w.get('height', 0) > 0]
    median_h = statistics.median(heights) if heights else 20
    y_tol = max(median_h * 0.5, 5)

    lines = _group_words_into_lines(words, y_tolerance_px=int(y_tol))

    rows = []
    for ri, line_words in enumerate(lines):
        y_min = min(w['top'] for w in line_words)
        y_max = max(w['top'] + w['height'] for w in line_words)
        rows.append({
            'index': ri,
            'y_min': y_min,
            'y_max': y_max,
            'y_center': (y_min + y_max) / 2,
        })

    return rows


# ---------------------------------------------------------------------------
# 3. Build cells
# ---------------------------------------------------------------------------

def _assign_word_to_column(word: Dict, columns: List[Dict]) -> int:
    """Return column index for a word based on its X-center."""
    x_center = word['left'] + word['width'] / 2
    for col in columns:
        if col['x_min'] <= x_center < col['x_max']:
            return col['index']
    # Fallback: nearest column
    return min(columns, key=lambda c: abs((c['x_min'] + c['x_max']) / 2 - x_center))['index']


def _assign_word_to_row(word: Dict, rows: List[Dict]) -> int:
    """Return row index for a word based on its Y-center."""
    y_center = word['top'] + word['height'] / 2
    # Find the row whose y_range contains this word's center
    for row in rows:
        if row['y_min'] <= y_center <= row['y_max']:
            return row['index']
    # Fallback: nearest row by Y-center
    return min(rows, key=lambda r: abs(r['y_center'] - y_center))['index']


def _build_cells(
    words: List[Dict],
    columns: List[Dict],
    rows: List[Dict],
    img_w: int,
    img_h: int,
) -> List[Dict[str, Any]]:
    """Build cell dicts from word assignments to (column, row) pairs."""
    if not columns or not rows:
        return []

    # Bucket words into (col_idx, row_idx)
    buckets: Dict[Tuple[int, int], List[Dict]] = {}
    for w in words:
        ci = _assign_word_to_column(w, columns)
        ri = _assign_word_to_row(w, rows)
        buckets.setdefault((ci, ri), []).append(w)

    cells = []
    for (ci, ri), cell_words in sorted(buckets.items(), key=lambda kv: (kv[0][1], kv[0][0])):
        col = columns[ci]
        row = rows[ri]

        # Compute tight bbox from actual word positions
        x_min = min(w['left'] for w in cell_words)
        y_min = min(w['top'] for w in cell_words)
        x_max = max(w['left'] + w['width'] for w in cell_words)
        y_max = max(w['top'] + w['height'] for w in cell_words)
        bw = x_max - x_min
        bh = y_max - y_min

        # Text from words in reading order
        text = _words_to_reading_order_text(cell_words, y_tolerance_px=max(10, int(bh * 0.4)))

        # Average confidence
        confs = [w.get('conf', 0) for w in cell_words if w.get('conf', 0) > 0]
        avg_conf = sum(confs) / len(confs) if confs else 0.0

        # Word boxes with absolute pixel coordinates (consistent with cv_cell_grid.py).
        # PaddleOCR returns phrase-level boxes (e.g. "competition [kompa'tifn]"),
        # but the overlay slide mechanism expects one box per word. Split multi-word
        # boxes into individual word positions proportional to character length.
        # Also split at "[" boundaries (IPA patterns like "badge[bxd3]").
        word_boxes = []
        for w in sorted(cell_words, key=lambda ww: (ww['top'], ww['left'])):
            raw_text = w.get('text', '').strip()
            # Split by whitespace, at "[" boundaries (IPA), and after leading "!"
            # e.g. "badge[bxd3]" → ["badge", "[bxd3]"]
            # e.g. "profit['proft]" → ["profit", "['proft]"]
            # e.g. "!Betonung" → ["!", "Betonung"]
            tokens = re.split(r'\s+|(?=\[)|(?<=!)(?=[A-Za-z\u00c0-\u024f])', raw_text)
            tokens = [t for t in tokens if t]  # remove empty strings
            if len(tokens) <= 1:
                # Single word — keep as-is
                word_boxes.append({
                    'text': raw_text,
                    'left': w['left'],
                    'top': w['top'],
                    'width': w['width'],
                    'height': w['height'],
                    'conf': w.get('conf', 0),
                })
            else:
                # Multi-word phrase — split proportionally by character count
                total_chars = sum(len(t) for t in tokens)
                if total_chars == 0:
                    continue
                # Small gap between words (2% of box width per gap)
                n_gaps = len(tokens) - 1
                gap_px = w['width'] * 0.02
                usable_w = w['width'] - gap_px * n_gaps
                cursor = w['left']
                for t in tokens:
                    token_w = max(1, usable_w * len(t) / total_chars)
                    word_boxes.append({
                        'text': t,
                        'left': round(cursor),
                        'top': w['top'],
                        'width': round(token_w),
                        'height': w['height'],
                        'conf': w.get('conf', 0),
                    })
                    cursor += token_w + gap_px

        cells.append({
            'cell_id': f"R{ri:02d}_C{ci}",
            'row_index': ri,
            'col_index': ci,
            'col_type': col['type'],
            'text': text,
            'confidence': round(avg_conf, 1),
            'bbox_px': {'x': x_min, 'y': y_min, 'w': bw, 'h': bh},
            'bbox_pct': {
                'x': round(x_min / img_w * 100, 2) if img_w else 0,
                'y': round(y_min / img_h * 100, 2) if img_h else 0,
                'w': round(bw / img_w * 100, 2) if img_w else 0,
                'h': round(bh / img_h * 100, 2) if img_h else 0,
            },
            'word_boxes': word_boxes,
            'ocr_engine': 'words_first',
            'is_bold': False,
        })

    return cells


# ---------------------------------------------------------------------------
# 4. Public API
# ---------------------------------------------------------------------------

def build_grid_from_words(
    word_dicts: List[Dict],
    img_w: int,
    img_h: int,
    min_confidence: int = 30,
) -> Tuple[List[Dict[str, Any]], List[Dict[str, Any]]]:
    """Build a cell grid bottom-up from Tesseract word boxes.

    Args:
        word_dicts: Flat list of word dicts with keys:
            text, left, top, width, height, conf
            (absolute pixel coordinates).
        img_w: Image width in pixels.
        img_h: Image height in pixels.
        min_confidence: Minimum OCR confidence to keep a word.

    Returns:
        (cells, columns_meta) — same format as build_cell_grid_v2().
        cells: list of cell dicts with cell_id, bbox_px, bbox_pct, etc.
        columns_meta: list of {'index', 'type', 'x', 'width'} dicts.
    """
    if not word_dicts:
        logger.info("build_grid_from_words: no words — returning empty grid")
        return [], []

    # Filter by confidence
    words = [
        w for w in word_dicts
        if w.get('conf', 0) >= min_confidence and w.get('text', '').strip()
    ]
    if not words:
        logger.info("build_grid_from_words: all words filtered (conf < %d)", min_confidence)
        return [], []

    logger.info("build_grid_from_words: %d words (after confidence filter from %d)", len(words), len(word_dicts))

    # Step 1: cluster columns
    columns = _cluster_columns(words, img_w)
    logger.info("build_grid_from_words: %d column(s) detected", len(columns))

    # Step 2: cluster rows
    rows = _cluster_rows(words)
    logger.info("build_grid_from_words: %d row(s) detected", len(rows))

    # Step 3: build cells
    cells = _build_cells(words, columns, rows, img_w, img_h)
    logger.info("build_grid_from_words: %d cells built", len(cells))

    # Build columns_meta in same format as build_cell_grid_v2
    columns_meta = []
    for col in columns:
        x = int(col['x_min'])
        w = int(col['x_max'] - col['x_min'])
        columns_meta.append({
            'index': col['index'],
            'type': col['type'],
            'x': x,
            'width': w,
        })

    return cells, columns_meta