Add 反切原貌 (#8)

The 反切 field in 廣韻.csv now contains the original fanqie with annotations for the correction.

Fanqies in the 釋義 and 釋義補充 fields are also reverted from poem's version to the original text (because annotations in 釋義 are not supported yet).

For the meaning of the annotations, see: #8 (comment)

build.py

@@ -1,12 +1,8 @@
 import csv
+from dataclasses import dataclass
+import re
 
 
-# 「通俗地位」
-音韻地位_patches = {
-    '892': ('幫二庚平', '幫二耕平'),
-    '1016': ('明一侯平', '明三C尤平'),
-    '3059': ('明一侯去', '明三C尤去'),
-}
 # 補全缺失釋義補充
 釋義補充_patch_from = {
     ('949', '蔆'): None,
@@ -17,18 +13,6 @@
 }
 
 
-def process_音韻地位(row: list[str]) -> str:
-    母, 呼, 等類, 韻, 聲 = row[10:15]
-    if not 母:
-        return ''
-    if (pos := 韻.find('→')) != -1:
-        韻 = 韻[pos + 1 :]
-    # NOTE 原資料莊組真殷韻依原貌。由於資料中已列「韻目原貌」，故地位不需再分
-    if 韻 in ('真', '殷') and 呼 == '開' and 母 in tuple('莊初崇生俟'):
-        韻 = '臻'
-    return 母 + 呼 + 等類 + 韻 + 聲
-
-
 def fix_pua(s: str) -> str:
     fixed = s.replace('\uee42', '𧞬').replace('\uece0', '勳')
     for ch in fixed:
@@ -38,18 +22,27 @@ def fix_pua(s: str) -> str:
     return fixed
 
 
+@dataclass
+class 小韻Row:
+    小韻號: str
+    首字: str
+    反切: str
+    音韻地位: str
+
+
 def main():
-    小韻_data: dict[str, list[str]] = {}
-    with open('src/rime-table-0b69606.tsv') as fin:
-        next(fin)
+    小韻_data: dict[str, 小韻Row] = {}
+    with open('src/小韻表.tsv') as fin:
+        header = next(fin)
+        assert header.rstrip('\n').split('\t') == [
+            '小韻號',
+            '首字',
+            '反切',
+            '音韻地位',
+        ], repr(header)
         for line in fin:
             row = line.rstrip('\n').split('\t')
-            小韻號 = row[0]
-            小韻_data[小韻號] = row
-
-    音韻地位_data: dict[str, str] = {
-        key: process_音韻地位(row) for key, row in 小韻_data.items()
-    }
+            小韻_data[row[0]] = 小韻Row(*row)
 
     has_細分: dict[str, list[str]] = {}
     小韻細分_data: dict[str, list[str]] = {}
@@ -60,71 +53,90 @@ def main():
             assert 小韻號[-1].isalpha()
             反切 = row[1]
             assert (
-                小韻_data[小韻號][2] == 反切
+                小韻_data[小韻號].反切 == 反切
             ), f'反切 mismatch in 小韻 #{小韻號}, 小韻_data: {小韻_data[小韻號][2]}, 小韻細分_data: {反切}'
             has_細分.setdefault(小韻號[:-1], []).append(小韻號[-1])
             小韻細分_data[小韻號] = row
 
+    釋義反切_patch: tuple[str, str] | None = None
     小韻細分_coverage: dict[str, set[str]] = {}
     廣韻_data: list[tuple[tuple[int, float], list[str]]] = []
     with open('src/廣韻(20170209).csv') as fin:
         for row in csv.DictReader(fin):
-            # Formerly used fields:
-            # - 廣韻反切原貌(覈校前)
-            # - 廣韻反切(覈校後)
-            # - 廣韻字頭原貌(覈校前)
-            # - 廣韻頁序
+            # Formerly used fields (field number is 1-based, same as awk & MS Excel):
+            # '廣韻反切原貌(覈校前)',  # 20
+            # '廣韻字頭原貌(覈校前)',  # 24
+            # '廣韻頁序',  # 57
             (
                 增刪說明,
+                poem_反切,
                 字頭,
                 釋義,
                 釋義補充,
                 韻目原貌,
-                小韻號原貌,
+                原書小韻號,
                 小韻內字序,
             ) = (
                 row[key]
                 for key in (
-                    '字頭-補',
-                    '廣韻字頭(覈校後)',
-                    '廣韻釋義',
-                    '釋義補充',
-                    '廣韻韻部原貌(調整前)',
-                    '小韻序',
-                    '小韻內字序',
+                    '字頭-補',  # 19
+                    '廣韻反切(覈校後)',  # 21
+                    '廣韻字頭(覈校後)',  # 25
+                    '廣韻釋義',  # 26
+                    '釋義補充',  # 27
+                    '廣韻韻部原貌(調整前)',  # 40
+                    '小韻序',  # 59
+                    '小韻內字序',  # 60
                 )
             )
 
             if 增刪說明 == '應刪':
                 continue
 
-            order_key = (int(小韻號原貌), float(小韻內字序))
+            order_key = (int(原書小韻號), float(小韻內字序))
 
             # 小韻號
-            if 小韻號原貌 in has_細分:
-                for 細分 in has_細分[小韻號原貌]:
-                    小韻號 = 小韻號原貌 + 細分
+            if 原書小韻號 in has_細分:
+                for 細分 in has_細分[原書小韻號]:
+                    小韻號 = 原書小韻號 + 細分
                     if 字頭 in 小韻細分_data[小韻號][2]:
                         小韻細分_coverage.setdefault(小韻號, set()).add(字頭)
                         break
                 else:
                     raise ValueError(
-                        f'cannot determine 小韻細分 for {字頭} (小韻 #{小韻號原貌})'
+                        f'cannot determine 小韻細分 for {字頭} (小韻 #{原書小韻號})'
                     )
             else:
-                小韻號 = 小韻號原貌
+                小韻號 = 原書小韻號
 
-            音韻地位 = 音韻地位_data[小韻號]
-            patch = 音韻地位_patches.get(小韻號)
-            if patch is not None:
-                assert (
-                    音韻地位 == patch[0]
-                ), f'invalid patch: expect {patch[0]} -> {patch[1]}, got {音韻地位}'
-                音韻地位 = patch[1]
-            反切 = 小韻_data[小韻號][2]
-            if 反切 == '無':
+            音韻地位 = 小韻_data[小韻號].音韻地位
+
+            反切 = 小韻_data[小韻號].反切
+            if 反切 == '-':
                 反切 = ''
 
+            if 小韻內字序 == '1':
+                if 反切:
+                    反切原貌 = re.sub(r'\[.\]|<.>|⦉.⦊|\(.\)|⦅.⦆', '', 反切)
+                    if 反切原貌 == poem_反切:
+                        釋義反切_patch = None
+                    else:
+                        assert (
+                            釋義.count(poem_反切 + '切') == 1
+                        ), f'釋義 not containing {反切}切 exactly once: {釋義}'
+                        釋義反切_patch = (
+                            釋義,
+                            釋義.replace(poem_反切 + '切', 反切原貌 + '切'),
+                        )
+                        釋義 = 釋義反切_patch[1]
+                else:
+                    釋義反切_patch = None
+            elif 釋義反切_patch is not None:
+                if 釋義補充 == 釋義反切_patch[0]:
+                    釋義補充 = 釋義反切_patch[1]
+                else:
+                    釋義反切_patch = None
+
             釋義_key = (小韻號, 字頭)
             if 釋義_key in 釋義補充_patch_from:
                 assert (

check.py

@@ -9,6 +9,12 @@
 PATTERN_描述 = re.compile(
     f'([{所有母}])([開合])?([{所有等}])([ABC])?([{所有韻}])([{所有聲}])'
 )
+PATTERN_反切 = re.compile(
+    r"""(?x)(
+        \[.\] |  # 脫字
+        . ( <.> | ⦉.⦊ | \(.\) | ⦅.⦆ )*  # 原貌及校正
+    ){2}"""
+)
 
 
 def contains_ascii(s: str):
@@ -35,15 +41,16 @@ def contains_ascii(s: str):
                 釋義,
                 釋義補充,
             ) = line.rstrip('\n').split(',')
-            if 音韻地位描述 != '':
-                assert (
-                    PATTERN_描述.fullmatch(音韻地位描述) is not None
-                ), f'invalid 音韻地位: {音韻地位描述}'
-            assert len(反切) in (
-                2,
-                0,
-            ), 'The length of 反切 should be 2, otherwise it should be an empty string'
+
+            assert (
+                PATTERN_描述.fullmatch(音韻地位描述) is not None
+            ), f'invalid 音韻地位: {音韻地位描述}'
+
+            if 反切:
+                assert PATTERN_反切.fullmatch(反切) is not None, f'invalid 反切: {反切}'
             assert len(字頭) == 1, 'The length of 字頭 should be 1'
+
+            assert 釋義 + 釋義補充, '釋義 and 釋義補充 should not be both empty'
             assert not contains_ascii(
                 釋義
             ), '釋義 should not contain any ASCII characters'