Reorder 廣韻字頭 and correct some 字 & 釋義

- All entries are reordered according to 澤存堂本
  - This solves a long-standing issue with both 韻典 and 廣韻字音表's
    data: Both data tables combine 字頭 from 廣韻全字表 and 釋義 from
    宋本廣韻データ. However 廣韻全字表 is based on 巾箱本 while
    宋本廣韻データ is based on 澤存堂本, which creates mismatches.

- Entries missing in poem's data are added back
  - This includes characters only representable with IDS, and seveval
    additions from 廣韻校本

- More errors in 字頭 & 釋義 are corrected
  - These were discovered when the new 字序表 was being made, and are
    still WIP.

README.md

@@ -3,12 +3,30 @@
 A database of the Qieyun phonological system.
 
 - 韻書
-    - 王一：`王一.csv` (not completed)
-    - 王三：`王三.csv` (小韻內部待校)
-    - 廣韻（澤存堂本）：`廣韻.csv`
+  - 王一：`王一.csv` (not completed)
+  - 王三：`王三.csv` (小韻內部待校)
+  - 廣韻 (澤存堂本, with corrections from 廣韻校本)：`廣韻.csv`
 - 韻圖
-    - 韻鏡（嘉吉本）：`韻鏡（嘉吉本）.csv` (not completed)
-    - 韻鏡（古逸叢書本）：`韻鏡（古逸叢書本）.csv`
+  - 韻鏡（嘉吉本）：`韻鏡（嘉吉本）.csv` (not completed)
+  - 韻鏡（古逸叢書本）：`韻鏡（古逸叢書本）.csv`
 - 反切音韻地位
-    - 王三：`王三反切音韻地位表.csv` (rev. Ayaka & unt)
-    - 廣韻：`廣韻反切音韻地位表.csv` (beta)
+  - 王三：`王三反切音韻地位表.csv` (rev. Ayaka & unt)
+  - 廣韻：`廣韻反切音韻地位表.csv` (beta)
+
+## About fields in 韻書/廣韻.csv
+
+- 小韻號: May contain -a/-b/-c if a 小韻 has multiple 音韻地位s
+- 小韻字號: May contain -a1, -a2 etc for entries not present in 澤存堂本 but added back according to 廣韻校本
+- 反切: May contain annotations:
+  - 脫字: `[徒]候` (小韻 #3067 豆)
+  - 訛字: `士<七>演` (小韻 #1625 淺)
+  - 改用其他來源的音韻地位: `姊宜⦉規⦊` (小韻 #133 厜)
+  - 替換成近似等價字，反切結果改變: `符咸(䒦)` (小韻 #1155 凡)
+  - 替換成音近字，反切結果改變: `式之(脂)` (小韻 #157 尸)
+  - 替換成等價字，反切結果不變: `甫⦅府⦆妄` (小韻 #2918 放)
+  - 替換成同音字，反切結果不變: `呼東⦅紅⦆` (小韻 #32 烘)
+  - 複合使用: `以沼⦅小⦆<水>` (小韻 #1692a 鷕)
+- 字頭當刪: if nonempty, indicates this entry in 澤存堂本 is errorneous and should be removed according to 廣韻校本
+- 釋義參照: 
+  - `上` if 釋義 refers to the entry above ("同上", "俗", "古文" etc.)
+  - `下` if it shares 釋義 with the entry below ("並上同", "並古文" etc.)

build.py

@@ -104,7 +104,7 @@ class 廣韻Row:
     音韻地位: str
     反切: str
     字頭: str
-    # 字頭當刪: str  # TODO
+    字頭當刪: str
     釋義: str
     釋義參照: str
 
@@ -166,6 +166,7 @@ def main():
                 釋義參照 = ''
 
         # 修正
+        字頭當刪 = ''
         if (patch := patches.get(字序_key)) is not None:
             assert patch.原字頭 == 字頭, (
                 f'patching 小韻 #{原書小韻號}/{小韻字號} 字 "{patch.原字頭}", but the actual 字 is "{字頭}"'
@@ -174,8 +175,12 @@ def main():
             if patch.校正字頭 and patch.校正字頭 != '～':
                 corrected = patch.校正字頭
                 if corrected.startswith('['):
-                    # TODO 暫忽略當刪字
-                    corrected = corrected[-2] if corrected[-2] != '-' else corrected[1]
+                    if corrected[-2] == '-':
+                        字頭當刪 = patch.當刪說明 or '當刪'
+                        corrected = corrected[1]
+                    else:
+                        assert not patch.當刪說明
+                        corrected = corrected[-2]
                 字頭 = corrected
 
             if patch.校正釋義 or patch.原釋義:
@@ -189,7 +194,15 @@ def main():
                     f'patching 釋義參照 on 小韻 #{原書小韻號}/{小韻字號} 字 "{patch.原字頭}", but the actual 釋義參照 is "{釋義參照}"'
                 )
                 釋義參照 = patch.校正釋義參照
-        # TODO 當刪字
+        elif 字序_data[字序_key].sbgy_字.endswith('/-]'):
+            字頭當刪 = '當刪'
+
+        字_check = 字序_data[字序_key].字
+        if 字_check.startswith('['):
+            字_check = 字_check[-2] if 字_check[-2] != '-' else 字_check[1]
+        assert 字頭 == 字_check, (
+            f'字頭 mismatch between 字序表 and patched data: "{字_check}" != "{字頭}"'
+        )
 
         # 小韻號
         if 原書小韻號 in has_細分:
@@ -223,7 +236,7 @@ def main():
             釋義 = 釋義.replace(poem_反切 + '切', 反切原貌 + '切')
 
         廣韻_data[字序_key] = 廣韻Row(
-            小韻號, 小韻字號, 韻目原貌, 音韻地位, 反切, 字頭, 釋義, 釋義參照
+            小韻號, 小韻字號, 韻目原貌, 音韻地位, 反切, 字頭, 字頭當刪, 釋義, 釋義參照
         )
 
     for 小韻號, cov in 小韻細分_coverage.items():