Update app.py

app.py

@@ -1,885 +1,141 @@
-import gradio as gr
-import difflib
-import re
-import html
+# verify_highlight_mono.py
+# モノクロAnnotationの見た目を確認する検証用アプリ
+
 import io
+import gradio as gr
 from reportlab.lib.pagesizes import A4
-from reportlab.lib.units import mm
+from reportlab.pdfgen import canvas as rl_canvas
 from reportlab.pdfbase import pdfmetrics
 from reportlab.pdfbase.ttfonts import TTFont
-from reportlab.pdfgen import canvas as rl_canvas
-import fitz  # PyMuPDF
+import fitz
 import tempfile
-import os
-
-# フォント登録
-pdfmetrics.registerFont(TTFont("NotoSansJP", "NotoSansJP-Regular.ttf"))
-
-# ============================================================
-# 文字除去ユーティリティ
-# ============================================================
-
-LATIN_LIGATURES = set(range(0xFB00, 0xFB07))
-
-KANGXI_RADICALS = set(range(0x2F00, 0x2FD6))
-CJK_RADICALS_SUPPLEMENT = set(range(0x2E80, 0x2F00))
-
-COMPATIBILITY_IDEOGRAPHS = set(range(0xF900, 0xFB00))
-ENCLOSED_CJK             = set(range(0x1F200, 0x1F300))
-VERTICAL_BRACKETS        = set(range(0xFE10, 0xFE20))
-CJK_COMPATIBILITY        = set(range(0x3300, 0x3400))
-ENCLOSED_ALPHANUMERICS   = set(range(0x2460, 0x2500))
-ENCLOSED_CJK_LETTERS     = set(range(0x3200, 0x3300))
-
-HYPHENS = set([
-    '\u002D',  # 半角ハイフンマイナス
-    '\uFF0D',  # 全角ハイフンマイナス
-    '\u2010',  # HYPHEN
-    '\u2011',  # NON-BREAKING HYPHEN
-    '\u2012',  # FIGURE DASH
-    '\u2013',  # EN DASH
-    '\u2014',  # EM DASH
-])
-
-KANJI_NUMBERS = set('零一二三四五六七八九十')
-ARABIC_NUMBERS = set('0123456789０１２３４５６７８９')
-
-YAKUMONO = set(
-    '。、．，.,'
-    '〔〕《》〈〉【】［］（）()[]'
-    '「」『』｢｣'
-    '〝〟'
-    '！？・…―'
-    '!?'
-    '〜～'
-    '；：;:'
-    '／/＼\\'
-    '‥'
-    '＋+'
-    '＊*'
-    '＃#'
-    '＆&'
-    '％%'
-    '＠@'
-    '＄$'
-    '￥¥'
-    '＾^'
-    '｜|'
-    '`｀'
-    '′″'
-    '§'
-    '※'
-    '〒'
-    '♪♫♬♩'
-    '→←↑↓↗↘↙↖'
-    '◯○●◎□■△▲▽▼◇◆☆★'
-    '©®™'
-)
-
-BRACKET_PAIRS = [
-    ('（', '）'),
-    ('(', ')'),
-    ('[', ']'),
-    ('〔', '〕'),
-    ('《', '》'),
-    ('〈', '〉'),
-    ('【', '】'),
-    ('［', '］'),
-    ('「', '」'),
-    ('『', '』'),
-]
-OPEN_BRACKETS  = set(p[0] for p in BRACKET_PAIRS)
-CLOSE_BRACKETS = set(p[1] for p in BRACKET_PAIRS)
-BRACKET_MAP    = {p[0]: p[1] for p in BRACKET_PAIRS}
-
-def remove_brackets_with_content(text: str) -> str:
-    result = []
-    i = 0
-    n = len(text)
-    while i < n:
-        ch = text[i]
-        if ch in OPEN_BRACKETS:
-            expected_close = BRACKET_MAP[ch]
-            depth = 1
-            j = i + 1
-            while j < n and depth > 0:
-                if text[j] == ch:
-                    depth += 1
-                elif text[j] == expected_close:
-                    depth -= 1
-                elif text[j] in CLOSE_BRACKETS and text[j] != expected_close:
-                    break
-                j += 1
-            if depth == 0:
-                i = j
-            else:
-                result.append(ch)
-                i += 1
-        else:
-            result.append(ch)
-            i += 1
-    return ''.join(result)
-
-def remove_yakumono(text: str) -> str:
-    return ''.join(c for c in text if c not in YAKUMONO)
-
-# ============================================================
-# 逆マッピング
-# ============================================================
-
-def build_active_flags(original: str, settings: dict) -> list:
-    n = len(original)
-    flags = [True] * n
-
-    if settings.get('ignore_brackets_with_content'):
-        i = 0
-        while i < n:
-            ch = original[i]
-            if ch in OPEN_BRACKETS:
-                expected_close = BRACKET_MAP[ch]
-                depth = 1
-                j = i + 1
-                while j < n and depth > 0:
-                    if original[j] == ch: depth += 1
-                    elif original[j] == expected_close: depth -= 1
-                    elif original[j] in CLOSE_BRACKETS and original[j] != expected_close: break
-                    j += 1
-                if depth == 0:
-                    for k in range(i, j): flags[k] = False
-                    i = j
-                else:
-                    i += 1
-            else:
-                i += 1
-
-    if settings.get('ignore_yakumono'):
-        for i, ch in enumerate(original):
-            if ch in YAKUMONO: flags[i] = False
-
-    if settings.get('ignore_alpha'):
-        for i, ch in enumerate(original):
-            if re.match(r'[A-Za-zＡ-Ｚａ-ｚ]', ch): flags[i] = False
-
-    if settings.get('ignore_digit'):
-        for i, ch in enumerate(original):
-            if re.match(r'[0-9０-９]', ch): flags[i] = False
-
-    if settings.get('ignore_kana'):
-        for i, ch in enumerate(original):
-            if re.match(r'[ァ-ンｦ-ﾟ]', ch): flags[i] = False
-
-    if settings.get('ignore_ligature'):
-        for i, ch in enumerate(original):
-            if ord(ch) in LATIN_LIGATURES: flags[i] = False
-
-    if settings.get('ignore_kangxi'):
-        for i, ch in enumerate(original):
-            if ord(ch) in KANGXI_RADICALS or ord(ch) in CJK_RADICALS_SUPPLEMENT: flags[i] = False
-
-    if settings.get('ignore_kanjinum'):
-        for i, ch in enumerate(original):
-            if ch in KANJI_NUMBERS or ch in ARABIC_NUMBERS: flags[i] = False
-
-    if settings.get('ignore_hyphen'):
-        for i, ch in enumerate(original):
-            if ch in HYPHENS: flags[i] = False
-
-    if settings.get('ignore_kumimoji'):
-        for i, ch in enumerate(original):
-            if (ord(ch) in COMPATIBILITY_IDEOGRAPHS or ord(ch) in ENCLOSED_CJK
-                    or ord(ch) in VERTICAL_BRACKETS or ord(ch) in CJK_COMPATIBILITY
-                    or ord(ch) in ENCLOSED_ALPHANUMERICS or ord(ch) in ENCLOSED_CJK_LETTERS):
-                flags[i] = False
-
-    if settings.get('ignore_space'):
-        for i, ch in enumerate(original):
-            if ch in ' \t　': flags[i] = False
-
-    if settings.get('ignore_newline'):
-        for i, ch in enumerate(original):
-            if ch in '\r\n': flags[i] = False
-
-    return flags
-
-def build_active_text_and_map(original: str, flags: list):
-    active_chars = [(i, ch) for i, ch in enumerate(original) if flags[i]]
-    active_text = ''.join(ch for _, ch in active_chars)
-    index_map = [i for i, _ in active_chars]
-    return active_text, index_map
-
-# ============================================================
-# ポストフィルタリング
-# ============================================================
-
-EQUAL_MERGE_THRESHOLD = 0
-
-def filter_opcodes(opcodes):
-    ops = list(opcodes)
-    n = len(ops)
-    if n == 0:
-        return [], []
-
-    is_diff = []
-    for tag, i1, i2, j1, j2 in ops:
-        if tag == 'equal':
-            is_diff.append((i2 - i1) <= EQUAL_MERGE_THRESHOLD)
-        else:
-            is_diff.append(True)
-
-    # 両端のequal区間はdiff扱いにしない
-    for idx in range(n):
-        if ops[idx][0] == 'equal':
-            is_diff[idx] = False
-        else:
-            break
-    for idx in range(n-1, -1, -1):
-        if ops[idx][0] == 'equal':
-            is_diff[idx] = False
-        else:
-            break
-
-    result_src = []
-    result_dst = []
-
-    i = 0
-    while i < n:
-        if is_diff[i]:
-            j = i + 1
-            while j < n and is_diff[j]:
-                j += 1
-            for k in range(i, j):
-                tag, i1, i2, j1, j2 = ops[k]
-                if tag in ('replace', 'delete'):
-                    for pos in range(i1, i2):
-                        result_src.append(pos)
-                if tag in ('replace', 'insert'):
-                    for pos in range(j1, j2):
-                        result_dst.append(pos)
-            i = j
-        else:
-            i += 1
-
-    return result_src, result_dst
-
-# ============================================================
-# 比較コア
-# ============================================================
-
-def merge_ranges(ranges):
-    merged = []
-    for s, e in sorted(ranges):
-        if s >= e: continue
-        if merged and s <= merged[-1][1]:
-            merged[-1] = (merged[-1][0], max(merged[-1][1], e))
-        else:
-            merged.append([s, e])
-    return [(s, e) for s, e in merged]
-
-def compare_texts(src: str, dst: str, settings: dict):
-    src_flags = build_active_flags(src, settings)
-    dst_flags = build_active_flags(dst, settings)
-
-    src_active, src_map = build_active_text_and_map(src, src_flags)
-    dst_active, dst_map = build_active_text_and_map(dst, dst_flags)
-
-    color_mode = settings.get('color_mode', 'color1')
-    if color_mode == 'color1':
-        hl_style = 'background-color:#ffaaaa;color:#000000;'
-    elif color_mode == 'color2':
-        hl_style = 'background-color:#aaaaff;color:#000000;'
-    else:
-        hl_style = 'background-color:#000000;color:#ffffff;'
-
-    src_diff_ranges = []
-    dst_diff_ranges = []
-
-    if src_active != '' or dst_active != '':
-        matcher = difflib.SequenceMatcher(None, src_active, dst_active, autojunk=False)
-        opcodes = matcher.get_opcodes()
-        result_src, result_dst = filter_opcodes(opcodes)
-
-        for pos in result_src:
-            orig = src_map[pos]
-            src_diff_ranges.append((orig, orig + 1))
-
-        for pos in result_dst:
-            orig = dst_map[pos]
-            dst_diff_ranges.append((orig, orig + 1))
-
-    has_diff = bool(src_diff_ranges or dst_diff_ranges)
-
-    def build_html(text: str, highlight_ranges: list) -> str:
-        merged = merge_ranges(highlight_ranges)
-        if not merged:
-            return '<span>' + html.escape(text).replace('\n', '<br>') + '</span>'
-        parts = []
-        prev = 0
-        for s, e in merged:
-            s = max(s, prev)
-            e = min(e, len(text))
-            if s >= e:
-                continue
-            if prev < s:
-                parts.append('<span>' + html.escape(text[prev:s]).replace('\n', '<br>') + '</span>')
-            parts.append(
-                f'<span style="{hl_style}">'
-                + html.escape(text[s:e]).replace('\n', '<br>')
-                + '</span>'
-            )
-            prev = e
-        if prev < len(text):
-            parts.append('<span>' + html.escape(text[prev:]).replace('\n', '<br>') + '</span>')
-        return ''.join(parts)
-
-    src_html = build_html(src, merge_ranges(src_diff_ranges))
-    dst_html = build_html(dst, merge_ranges(dst_diff_ranges))
-
-    return src_html, dst_html, has_diff
-
-# ============================================================
-# 統計
-# ============================================================
-
-def count_stats(text: str) -> dict:
-    char_count    = len(re.sub(r'[ 　\t\r\n]', '', text))
-    space_count   = len(re.findall(r'[ 　\t]', text))
-    newline_count = len(re.findall(r'\r\n|\r|\n', text))
-    space_incl    = char_count + space_count
-    newline_incl  = char_count + newline_count
-    words         = len(re.split(r'[ 　\t\r\n]+', text.strip())) if text.strip() else 0
-    return {
-        'char':    char_count,
-        'space':   space_count,
-        'space_incl': space_incl,
-        'newline': newline_count,
-        'newline_incl': newline_incl,
-        'words':   words,
-    }
-
-# ============================================================
-# PDF生成
-# ============================================================
-
-def generate_pdf(src_orig: str, dst_orig: str, settings: dict,
-                 src_html_body: str, dst_html_body: str, has_diff: bool) -> str:
-
-    FONT       = "NotoSansJP"
-    PAGE_W, PAGE_H = A4
-    MARGIN     = 15 * mm
-    COL_GAP    = 4  * mm
-    COL_W      = (PAGE_W - MARGIN * 2 - COL_GAP) / 2
-    FONT_SIZE  = 8.5
-    LINE_HEIGHT= 13
-    COL_L      = MARGIN
-    COL_R      = MARGIN + COL_W + COL_GAP
-    DIV_X      = MARGIN + COL_W + COL_GAP / 2
-
-    color_mode = settings.get('color_mode', 'color1')
-    if color_mode == 'color1':
-        HL_COLOR_RL = (1.0, 0.6, 0.6)
-        HL_FG_RL    = (0.0, 0.0, 0.0)
-        HL_COLOR_MU = (1.0, 0.6, 0.6)
-    elif color_mode == 'color2':
-        HL_COLOR_RL = (0.6, 0.8, 1.0)
-        HL_FG_RL    = (0.0, 0.0, 0.0)
-        HL_COLOR_MU = (0.6, 0.8, 1.0)
-    else:  # monochrome
-        HL_COLOR_RL = (0.0, 0.0, 0.0)
-        HL_FG_RL    = (1.0, 1.0, 1.0)
-        HL_COLOR_MU = (0.5, 0.5, 0.5)
 
-    # ----------------------------------------------------------
-    # HTMLから差分範囲を復元
-    # ----------------------------------------------------------
-    def extract_diff_ranges(text: str, html_body: str) -> list:
-        """
-        html_bodyのspanタグからハイライト箇所の文字インデックスリストを復元する。
-        戻り値: [(start, end), ...] ※textのインデックス
-        """
-        ranges = []
-        # ハイライトspanを検索
-        pattern = re.compile(
-            r'<span style="[^"]*background-color[^"]*">(.*?)</span>',
-            re.DOTALL
-        )
-        # タグを除去してプレーンテキストを再構築しながら位置を追跡
-        plain = ''
-        pos   = 0
-        for m in re.finditer(
-            r'<span(?:\s+style="([^"]*)")?>(.*?)</span>',
-            html_body, re.DOTALL
-        ):
-            style   = m.group(1) or ''
-            content = m.group(2)
-            # <br>を改行に戻す
-            content = content.replace('<br>', '\n')
-            # HTMLエスケープを戻す
-            content = html.unescape(content)
-            is_hl   = 'background-color' in style
-            start   = len(plain)
-            plain  += content
-            end     = len(plain)
-            if is_hl:
-                ranges.append((start, end))
-        return ranges
+FONT_NAME = "NotoSansJP"
+FONT_FILE = "NotoSansJP-Regular.ttf"
+
+pdfmetrics.registerFont(TTFont(FONT_NAME, FONT_FILE))
+
+PAGE_W, PAGE_H = A4
+MARGIN     = 15
+FONT_SIZE  = 12
+LINE_HEIGHT= 18
+
+COLORS = {
+    "カラー1(赤)": {
+        "rl_bg": (1.0, 0.6, 0.6),
+        "rl_fg": (0.0, 0.0, 0.0),
+        "mu":    (1.0, 0.6, 0.6),
+    },
+    "カラー2(青)": {
+        "rl_bg": (0.6, 0.8, 1.0),
+        "rl_fg": (0.0, 0.0, 0.0),
+        "mu":    (0.6, 0.8, 1.0),
+    },
+    "モノクロ": {
+        "rl_bg": (0.0, 0.0, 0.0),
+        "rl_fg": (1.0, 1.0, 1.0),
+        "mu":    (0.5, 0.5, 0.5),
+    },
+}
+
+def build_pdf(color_mode: str) -> bytes:
+    col = COLORS[color_mode]
+    rl_bg = col["rl_bg"]
+    rl_fg = col["rl_fg"]
+    mu_color = col["mu"]
 
-    src_diff_ranges = extract_diff_ranges(src_orig, src_html_body)
-    dst_diff_ranges = extract_diff_ranges(dst_orig, dst_html_body)
-
-    # ----------------------------------------------------------
-    # テキストを折り返し済み描画単位に変換
-    # ----------------------------------------------------------
-    def to_sublines(text: str, col_w: float, font_size: float):
-        """
-        戻り値: [(char_offset, line_text), ...]
-        char_offset: テキスト先頭からの文字インデックス
-        """
-        result = []
-        lines  = text.split('\n')
-        offset = 0
-        for line in lines:
-            cur        = ''
-            cur_start  = offset
-            for ci, ch in enumerate(line):
-                test = cur + ch
-                if pdfmetrics.stringWidth(test, FONT, font_size) > col_w - 2 and cur:
-                    result.append((cur_start, cur))
-                    cur       = ch
-                    cur_start = offset + ci
-                else:
-                    cur = test
-            result.append((cur_start, cur))
-            offset += len(line) + 1  # +1 for '\n'
-        return result
-
-    # ----------------------------------------------------------
-    # ハイライト判定
-    # ----------------------------------------------------------
-    def is_highlighted(char_idx: int, diff_ranges: list) -> bool:
-        for s, e in diff_ranges:
-            if s <= char_idx < e:
-                return True
-        return False
-
-    # ----------------------------------------------------------
-    # 1行描画（背景矩形 → 文字）
-    # 戻り値: ハイライトした文字の座標リスト [(x0,y0,x1,y1), ...]
-    # ※ PyMuPDF注釈付与用
-    # ----------------------------------------------------------
-    def draw_subline(c, sub_text: str, sub_offset: int, diff_ranges: list,
-                     x_start: float, y_rl: float, font_size: float) -> list:
-        """
-        y_rl: reportlabのy座標（下原点）
-        戻り値: ハイライト文字のrect座標リスト (reportlab座標系)
-        """
-        if not sub_text:
-            return []
+    # reportlabでベースPDF生成
+    buffer = io.BytesIO()
+    c = rl_canvas.Canvas(buffer, pagesize=A4)
+    c.setFont(FONT_NAME, FONT_SIZE)
+
+    lines = [
+        ("通常テキストです。ハイライトなし。", False),
+        ("ここは差分箇所です。ハイライトあり。", True),
+        ("また通常テキストです。", False),
+        ("ここも差分箇所。ABCXYZ", True),
+    ]
 
-        hl_rects = []
+    hl_rects = []  # (x0, y0_rl, x1, y1_rl)
+    y = PAGE_H - MARGIN - 20
 
-        # パス1: 背景矩形
-        dx = x_start
-        for ci, ch in enumerate(sub_text):
-            cw = pdfmetrics.stringWidth(ch, FONT, font_size)
-            if is_highlighted(sub_offset + ci, diff_ranges):
+    for text, is_hl in lines:
+        if is_hl:
+            tw = pdfmetrics.stringWidth(text, FONT_NAME, FONT_SIZE)
+            # モノクロのみ背景矩形を描画
+            if color_mode == "モノクロ":
                 c.saveState()
-                c.setFillColorRGB(*HL_COLOR_RL)
-                c.rect(dx, y_rl - 2, cw, LINE_HEIGHT - 1, fill=1, stroke=0)
+                c.setFillColorRGB(*rl_bg)
+                c.rect(MARGIN, y - 2, tw, LINE_HEIGHT - 1, fill=1, stroke=0)
                 c.restoreState()
-                hl_rects.append((dx, y_rl - 2, dx + cw, y_rl + LINE_HEIGHT - 1))
-            dx += cw
-
-        # パス2: 文字描画
-        dx      = x_start
-        seg     = ''
-        seg_hl  = None
-        seg_x   = x_start
-
-        for ci, ch in enumerate(sub_text):
-            hl = is_highlighted(sub_offset + ci, diff_ranges)
-            cw = pdfmetrics.stringWidth(ch, FONT, font_size)
-            if seg_hl is None:
-                seg_hl, seg, seg_x = hl, ch, dx
-            elif hl == seg_hl:
-                seg += ch
-            else:
-                c.saveState()
-                c.setFont(FONT, font_size)
-                c.setFillColorRGB(*(HL_FG_RL if seg_hl else (0.0, 0.0, 0.0)))
-                c.drawString(seg_x, y_rl, seg)
-                c.restoreState()
-                seg_x += pdfmetrics.stringWidth(seg, FONT, font_size)
-                seg_hl, seg = hl, ch
-            dx += cw
-
-        if seg:
+            # ハイライト文字色設定
             c.saveState()
-            c.setFont(FONT, font_size)
-            c.setFillColorRGB(*(HL_FG_RL if seg_hl else (0.0, 0.0, 0.0)))
-            c.drawString(seg_x, y_rl, seg)
+            c.setFont(FONT_NAME, FONT_SIZE)
+            if color_mode == "モノクロ":
+                c.setFillColorRGB(*rl_fg)
+            else:
+                c.setFillColorRGB(0.0, 0.0, 0.0)
+            c.drawString(MARGIN, y, text)
             c.restoreState()
-
-        return hl_rects
-
-    # ----------------------------------------------------------
-    # reportlabでベースPDF生成
-    # ----------------------------------------------------------
-    def build_base_pdf() -> tuple:
-        """
-        戻り値: (pdf_bytes, all_hl_rects)
-        all_hl_rects: {page_no: {'src': [(x0,y0_rl,x1,y1_rl),...], 'dst': [...]}}
-        ※ y座標はreportlab座標系（下原点）
-        """
-        buffer   = io.BytesIO()
-        c        = rl_canvas.Canvas(buffer, pagesize=A4)
-        all_hl   = {}  # page_no -> {'src':[], 'dst':[]}
-        page_no  = 0
-
-        def hline(y, lw=0.5):
+            # Annotation用座標記録（カラー1/2のみ）
+            if color_mode != "モノクロ":
+                hl_rects.append((MARGIN, y - 2, MARGIN + tw, y + LINE_HEIGHT - 1))
+        else:
             c.saveState()
-            c.setStrokeColorRGB(0.5, 0.5, 0.5)
-            c.setLineWidth(lw)
-            c.line(MARGIN, y, PAGE_W - MARGIN, y)
+            c.setFont(FONT_NAME, FONT_SIZE)
+            c.setFillColorRGB(0.0, 0.0, 0.0)
+            c.drawString(MARGIN, y, text)
             c.restoreState()
+        y -= LINE_HEIGHT
 
-        def new_page():
-            nonlocal page_no
-            c.showPage()
-            page_no += 1
-
-        def ensure_page_space(y, needed=LINE_HEIGHT):
-            if y < MARGIN + needed:
-                new_page()
-                return PAGE_H - MARGIN
-            return y
-
-        y = PAGE_H - MARGIN
-
-        # タイトル
-        c.saveState()
-        c.setFont(FONT, 18)
-        c.setFillColorRGB(0, 0, 0)
-        title = "テキスト比較ツール"
-        c.drawString((PAGE_W - pdfmetrics.stringWidth(title, FONT, 18)) / 2, y, title)
-        c.restoreState()
-        y -= 24
-
-        hline(y)
-        y -= 10
-
-        # 比較設定
-        c.saveState()
-        c.setFont(FONT, 10)
-        c.setFillColorRGB(0, 0, 0)
-        c.drawString(MARGIN, y, "■ 比較設定")
-        c.restoreState()
-        y -= 14
-
-        ignore_labels = {
-            'ignore_alpha':    '英字(全半角)無視',
-            'ignore_digit':    '数字(全半角)無視',
-            'ignore_kana':     'カナ(全半角)無視',
-            'ignore_yakumono': '約物(全半角)無視',
-            'ignore_ligature': 'リガチャ(合字)無視',
-            'ignore_kangxi':   '康煕部首(CJK部首補助)無視',
-            'ignore_brackets_with_content': '括弧付き文字無視',
-            'ignore_kanjinum': '漢数字(算用数字)無視',
-            'ignore_hyphen':   'ハイフン無視',
-            'ignore_kumimoji': '組み文字無視',
-            'ignore_space':    'スペース(空白)削除',
-            'ignore_newline':  '改行削除',
-        }
-        checked = [v for k, v in ignore_labels.items() if settings.get(k)]
-        c.saveState()
-        c.setFont(FONT, 8)
-        c.setFillColorRGB(0, 0, 0)
-        c.drawString(MARGIN, y, "有効な無視設定: " + ("、".join(checked) if checked else "なし"))
-        c.restoreState()
-        y -= 12
-
-        hline(y)
-        y -= 14
-
-        # 結果ヘッダー
-        c.saveState()
-        c.setFont(FONT, 14)
-        c.setFillColorRGB(0.85, 0, 0)
-        c.drawString(MARGIN, y, "相違が見つかりました" if has_diff else "相違はありません")
-        c.restoreState()
-        y -= 20
-
-        # カラムヘッダー
-        c.saveState()
-        c.setFont(FONT, 9)
-        c.setFillColorRGB(0, 0, 0)
-        c.drawString(COL_L, y, "【比較元】")
-        c.drawString(COL_R, y, "【比較先】")
-        c.restoreState()
-        y -= 14
-
-        # 本文描画
-        sublines_src = to_sublines(src_orig, COL_W, FONT_SIZE)
-        sublines_dst = to_sublines(dst_orig, COL_W, FONT_SIZE)
-        max_lines    = max(len(sublines_src), len(sublines_dst))
-        while len(sublines_src) < max_lines:
-            sublines_src.append((len(src_orig), ''))
-        while len(sublines_dst) < max_lines:
-            sublines_dst.append((len(dst_orig), ''))
-
-        seg_y_top = y
-        all_hl[page_no] = {'src': [], 'dst': []}
-
-        for (off_s, txt_s), (off_d, txt_d) in zip(sublines_src, sublines_dst):
-            y = ensure_page_space(y)
-            if y == PAGE_H - MARGIN:
-                # ページ送り発生 → 縦区切り線を前ページに描画済み想定
-                # 新ページの縦線開始位置を更新
-                seg_y_top = y
-                all_hl[page_no] = {'src': [], 'dst': []}
-
-            rects_s = draw_subline(c, txt_s, off_s, src_diff_ranges, COL_L, y, FONT_SIZE)
-            rects_d = draw_subline(c, txt_d, off_d, dst_diff_ranges, COL_R, y, FONT_SIZE)
-            all_hl[page_no]['src'].extend(rects_s)
-            all_hl[page_no]['dst'].extend(rects_d)
-            y -= LINE_HEIGHT
-
-        # 縦区切り線
-        c.saveState()
-        c.setStrokeColorRGB(0.5, 0.5, 0.5)
-        c.setLineWidth(0.5)
-        c.line(DIV_X, seg_y_top, DIV_X, y + LINE_HEIGHT)
-        c.restoreState()
-
-        y -= 6
-
-        # 統計
-        hline(y, lw=0.3)
-        y -= 10
-
-        src_stats = count_stats(src_orig)
-        dst_stats = count_stats(dst_orig)
+    c.save()
+    pdf_bytes = buffer.getvalue()
 
-        def draw_stats(x, stats):
-            nonlocal y
-            for line in [
-                f"文字数: {stats['char']}",
-                f"空白数: {stats['space']}  空白込み文字数: {stats['space_incl']}",
-                f"改行数: {stats['newline']}  改行込み文字数: {stats['newline_incl']}",
-                f"単語数: {stats['words']}",
-            ]:
-                c.saveState()
-                c.setFont(FONT, 7.5)
-                c.setFillColorRGB(0.2, 0.2, 0.2)
-                c.drawString(x, y, line)
-                c.restoreState()
-                y -= 10
-
-        y_stats = y
-        draw_stats(COL_L, src_stats)
-        y = y_stats
-        draw_stats(COL_R, dst_stats)
-
-        c.save()
-        return buffer.getvalue(), all_hl
-
-    # ----------------------------------------------------------
-    # PyMuPDFでHighlight Annotation付与
-    # ----------------------------------------------------------
-    def add_highlight_annots(pdf_bytes: bytes, all_hl: dict) -> bytes:
+    # PyMuPDFでAnnotation付与（カラー1/2のみ）
+    if color_mode != "モノクロ" and hl_rects:
         doc = fitz.open(stream=pdf_bytes, filetype="pdf")
+        page = doc[0]
+        page_h = page.rect.height
 
-        for page_no, hl_data in all_hl.items():
-            if page_no >= len(doc):
-                continue
-            page   = doc[page_no]
-            page_h = page.rect.height  # PyMuPDF座標系の高さ（上原点）
-
-            rects_all = hl_data.get('src', []) + hl_data.get('dst', [])
-            for (x0, y0_rl, x1, y1_rl) in rects_all:
-                # reportlab座標(下原点) → PyMuPDF座標(上原点)変換
-                # reportlab: y0_rl=下端, y1_rl=上端
-                # PyMuPDF:   y_top = page_h - y1_rl, y_bottom = page_h - y0_rl
-                mu_y_top    = page_h - y1_rl
-                mu_y_bottom = page_h - y0_rl
-                rect  = fitz.Rect(x0, mu_y_top, x1, mu_y_bottom)
-                annot = page.add_highlight_annot(rect)
-                annot.set_colors(stroke=HL_COLOR_MU)
-                annot.update()
+        for (x0, y0_rl, x1, y1_rl) in hl_rects:
+            mu_y_top    = page_h - y1_rl
+            mu_y_bottom = page_h - y0_rl
+            rect  = fitz.Rect(x0, mu_y_top, x1, mu_y_bottom)
+            annot = page.add_highlight_annot(rect)
+            annot.set_colors(stroke=mu_color)
+            annot.update()
 
         result = io.BytesIO()
         doc.save(result)
         doc.close()
-        return result.getvalue()
-
-    # ----------------------------------------------------------
-    # 実行
-    # ----------------------------------------------------------
-    base_pdf_bytes, all_hl_rects = build_base_pdf()
-    final_pdf_bytes = add_highlight_annots(base_pdf_bytes, all_hl_rects)
-
-    tmp = tempfile.NamedTemporaryFile(delete=False, suffix='.pdf')
-    tmp.write(final_pdf_bytes)
-    tmp.close()
-    return tmp.name
-
-# ============================================================
-# Gradio メイン処理
-# ============================================================
+        pdf_bytes = result.getvalue()
 
-MAX_CHARS = 50000
+    return pdf_bytes
 
-def run_compare(src_raw, dst_raw,
-                ig_alpha, ig_digit, ig_kana, ig_yakumono,
-                ig_ligature, ig_kangxi, ig_bracket_content,
-                ig_kanjinum, ig_hyphen, ig_kumimoji,
-                ig_space, ig_newline,
-                color_mode):
+def run_verify(color_mode):
+    try:
+        pdf_bytes = build_pdf(color_mode)
+        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
+        tmp.write(pdf_bytes)
+        tmp.close()
+        return f"[OK] PDF生成完了 ({len(pdf_bytes)} bytes)", tmp.name
+    except Exception as e:
+        import traceback
+        return f"[ERROR] {e}\n{traceback.format_exc()}", None
 
-    src = (src_raw or '')[:MAX_CHARS]
-    dst = (dst_raw or '')[:MAX_CHARS]
-
-    settings = {
-        'ignore_alpha':    ig_alpha,
-        'ignore_digit':    ig_digit,
-        'ignore_kana':     ig_kana,
-        'ignore_yakumono': ig_yakumono,
-        'ignore_ligature': ig_ligature,
-        'ignore_kangxi':   ig_kangxi,
-        'ignore_brackets_with_content': ig_bracket_content,
-        'ignore_kanjinum': ig_kanjinum,
-        'ignore_hyphen':   ig_hyphen,
-        'ignore_kumimoji': ig_kumimoji,
-        'ignore_space':    ig_space,
-        'ignore_newline':  ig_newline,
-        'color_mode':      {'カラー1(赤)': 'color1', 'カラー2(青)': 'color2', 'モノクロ': 'monochrome'}.get(color_mode, 'color1'),
-    }
-
-    src_html_body, dst_html_body, has_diff = compare_texts(src, dst, settings)
-    src_stats = count_stats(src)
-    dst_stats = count_stats(dst)
-
-    if has_diff:
-        diff_label = '<p style="color:red;font-size:1.4em;font-weight:bold;">相違が見つかりました</p>'
-    else:
-        diff_label = '<p style="color:red;font-size:1.4em;font-weight:bold;">相違はありません</p>'
-
-    def stats_html(stats):
-        return (
-            f'<p style="font-size:0.85em;">'
-            f'文字数: {stats["char"]}　'
-            f'空白数: {stats["space"]}　空白込み文字数: {stats["space_incl"]}<br>'
-            f'改行数: {stats["newline"]}　改行込み文字数: {stats["newline_incl"]}<br>'
-            f'単語数: {stats["words"]}'
-            f'</p>'
-        )
-
-    result_html = f"""
-{diff_label}
-<div style="display:flex;gap:0;font-size:1em;">
-  <div style="flex:1;padding:8px;word-break:break-all;">
-    <div style="font-weight:bold;margin-bottom:4px;">比較元</div>
-    <div style="font-size:1em;line-height:1.8;">{src_html_body}</div>
-    {stats_html(src_stats)}
-  </div>
-  <div style="width:1px;background:#333;"></div>
-  <div style="flex:1;padding:8px;word-break:break-all;">
-    <div style="font-weight:bold;margin-bottom:4px;">比較先</div>
-    <div style="font-size:1em;line-height:1.8;">{dst_html_body}</div>
-    {stats_html(dst_stats)}
-  </div>
-</div>
-"""
-
-    pdf_path = generate_pdf(src, dst, settings, src_html_body, dst_html_body, has_diff)
-
-    return result_html, pdf_path
-
-def clear_all():
-    return (
-        '',    # src
-        '',    # dst
-        False, False, False, False,
-        False, False, False,
-        False, False, False,
-        False, False,
-        'カラー1(赤)',
-        '',    # result_html
-        None,  # pdf
-    )
-
-# ============================================================
-# UI
-# ============================================================
-
-with gr.Blocks(title="テキスト比較ツール") as demo:
-
-    gr.Markdown("# テキスト比較ツール")
-    gr.HTML("<hr>")
-    gr.Markdown(
-        "下の枠内(テキストエリア)に比較したい文章を入力し、「比較」ボタンを押します。\n\n"
-        "※最大50,000文字まで入力できます。コピーペースト時に50,000文字を超えた文字はすべて削除されます。"
-        "(参考:A4、1ページで1,000～1,500文字程度)"
-    )
-
-    with gr.Row():
-        src_input = gr.Textbox(label="比較元データ", lines=12, max_lines=30)
-        dst_input = gr.Textbox(label="比較先データ", lines=12, max_lines=30)
-
-    gr.HTML('<p style="color:red;font-size:0.85em;">※チェックを入れた項目は差分があっても検出されません</p>')
-
-    with gr.Row():
-        ig_alpha    = gr.Checkbox(label="英字(全半角)無視")
-        ig_digit    = gr.Checkbox(label="数字(全半角)無視")
-        ig_kana     = gr.Checkbox(label="カナ(全半角)無視")
-        ig_yakumono = gr.Checkbox(label="約物(全半角)無視")
-
-    with gr.Row():
-        ig_ligature        = gr.Checkbox(label="リガチャ(合字)無視")
-        ig_kangxi          = gr.Checkbox(label="康煕部首(CJK部首補助)無視")
-        ig_bracket_content = gr.Checkbox(label="括弧付き文字無視")
-        ig_kanjinum        = gr.Checkbox(label="漢数字(算用数字)無視")
-
-    with gr.Row():
-        ig_hyphen   = gr.Checkbox(label="ハイフン無視")
-        ig_kumomoji = gr.Checkbox(label="組み文字無視")
-        ig_space    = gr.Checkbox(label="スペース(空白)削除")
-        ig_newline  = gr.Checkbox(label="改行削除")
-
-    with gr.Row():
-        compare_btn = gr.Button("比較", variant="primary")
-        clear_btn   = gr.Button("テキストクリア")
-
-    gr.HTML("<hr>")
+with gr.Blocks(title="モノクロAnnotation検証") as demo:
+    gr.Markdown("## モノクロAnnotation検証")
+    gr.Markdown("カラーモードを選択して実行し、PDFの見た目を確認してください。")
 
     color_mode = gr.Radio(
         choices=["カラー1(赤)", "カラー2(青)", "モノクロ"],
-        value="カラー1(赤)",
-        label="差分箇所の色表示設定",
+        value="モノクロ",
+        label="カラーモード"
     )
+    run_btn = gr.Button("検証実行", variant="primary")
+    log_out = gr.Textbox(label="ログ", lines=5)
+    pdf_out = gr.File(label="PDFダウンロード")
 
-    result_html = gr.HTML(label="比較結果")
-    pdf_output  = gr.File(label="PDFダウンロード")
-
-    all_inputs = [
-        src_input, dst_input,
-        ig_alpha, ig_digit, ig_kana, ig_yakumono,
-        ig_ligature, ig_kangxi, ig_bracket_content,
-        ig_kanjinum, ig_hyphen, ig_kumomoji,
-        ig_space, ig_newline,
-        color_mode,
-    ]
-    all_outputs_clear = [
-        src_input, dst_input,
-        ig_alpha, ig_digit, ig_kana, ig_yakumono,
-        ig_ligature, ig_kangxi, ig_bracket_content,
-        ig_kanjinum, ig_hyphen, ig_kumomoji,
-        ig_space, ig_newline,
-        color_mode,
-        result_html, pdf_output,
-    ]
-
-    compare_btn.click(fn=run_compare, inputs=all_inputs, outputs=[result_html, pdf_output])
-    clear_btn.click(fn=clear_all, inputs=[], outputs=all_outputs_clear)
+    run_btn.click(fn=run_verify, inputs=[color_mode], outputs=[log_out, pdf_out])
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()