[stantheman0128]

Closes #905.

Problem

TSCharacters.txt had no mapping for the variant 蔕, so t2s left it unchanged:

蔕 -> 蔕
花蔕 -> 花蔕
並蔕蓮 -> 并蔕莲

Issue #905 also discussed 谿. Current HEAD already has 谿 -> 溪 谿 (default 溪 first). Phrase-level keep-谿 cases such as 豁谿 / 谿刻 / 谿卡 / 谿壑 are covered by #901 / PR #1447, not this change. Existing testcase BYVoid_OpenCC_PR_464_xi_vs_xi expects 勃谿 -> 勃溪, so this PR does not add a 勃谿 identity phrase.

Change

data/dictionary/TSCharacters.txt:

蔕 -> 蒂

Sorted with data/scripts/sort.py. No change to the existing 谿 -> 溪 谿 entry.

Test plan

• New cases BYVoid_OpenCC_Issue_905_* in test/testcases/testcases.json
• Control: ordinary 蘭谿 / 松谿 still become 兰溪 / 松溪 (溪 first)
• Local verification with OpenCC 1.4.1 wheel + branch text dictionaries (see Evidence)
• CI Bazel dictionary tests (authoritative)

Evidence

Repro before (packaged wheel dictionaries):

t2s  蔕 -> 蔕
t2s  花蔕 -> 花蔕
t2s  並蔕蓮 -> 并蔕莲
t2s  煙蔕 -> 烟蔕

After this branch (config JSONs with ocd2 refs switched to text, driving data/dictionary/*.txt on the branch):

t2s/tw2s/hk2s/tw2sp/hk2sp  蔕 -> 蒂
t2s/tw2s/hk2s/tw2sp/hk2sp  花蔕 並蔕蓮 煙蔕 -> 花蒂 并蒂莲 烟蒂
t2s/tw2s/hk2s              蘭谿 松谿 -> 兰溪 松溪

Harness summary (OpenCC 1.4.1 wheel, config ocd2 references rewritten to text):

branch dicts + new #905 cases: 8 passed, 0 failed
master vs branch on BYVoid_OpenCC_PR_464_xi_vs_xi (text harness): identical pre-existing gap (樠谿 / U+235CA), unrelated to 蔕

Please treat CI Bazel dictionary tests as authoritative.

AI assistance

This fix was developed with AI assistance (Cursor / Grok) and verified locally as shown above.

[frankslin]

罕見異體字不轉換。