Found while implementing #194 (PR for claude/issue-194-twp-artifacts). Not fixed — reported only, because it is a contested rendering of the #195 class, not a mechanical artifact of the #194 class.
What
I swept every TWPhrases entry that actually fires in this corpus (143 of the 603) to check whether #194 had left any sibling behind. It had not: every remaining candidate I flagged — 局部 → 區域性, 回调 → 回撥, 抛出 → 丟擲, 命名空间 → 名稱空間, 优先级 → 優先順序, 表达式 → 表示式, 外键 → 外部索引鍵 — is already a row in #195's table.
One is not.
| zh-Hans |
n (source) |
in dist/zh-Hant |
twp ships |
the alternative a reviewer would weigh |
连接器 |
17 |
34 |
聯結器 |
連接器 |
Measured on ca3983b + the #194 branch:
grep -roh '连接器' content --include='*.mdx' --include='*.ts' | grep -v zh-Hant | wc -l -> 17
grep -roh '聯結器' dist/zh-Hant | wc -l -> 34
The dist number is roughly double the source count because one of the hits is a post title — 跨系统自动化：连接器和 Webhook 如何进入受治理流程 — so it also reaches the blog index, the topic hubs, rss.xml and the <title> tag, the same chrome-amplification #192's patch round ran into.
Why it belongs in #195 and not in #194
The #194 test is "wrong in any register" — 介面卡 names different hardware, 全域性 is not a noun. 聯結器 fails no such test. It is a real Taiwanese word; it is simply the mechanical coupling sense first (a railway coupler, a connector housing), where the software/data sense is normally written 連接器 in Taiwan. That is a judgement, and #195 is explicitly the place judgements go: "No fix proposed. Nothing here is dispatchable until someone who reads Taiwanese technical writing has looked at it."
Worth noting only because of volume. At 17 source occurrences it would sit fifth in #195's table, above ten of the fifteen rows already there — and one of those occurrences is a post title, which is the most visible position in the corpus.
What to do
Nothing, until #195 gets its reviewer. If and when it does, this is one more row on the same sheet, and it costs the same one line in TWP_PHRASE_OVERRIDES plus one fixture case in CONVERSION_CASES (src/lib/zhconvert.ts) that every other row would cost.
Also recorded, and deliberately not proposed as a defect: 十进制 → 十進位制 (1 occurrence) and 顯像管 → 映象管 — the latter is a second instance of the exact 象/像 slip #194 fixes for 鏡像, still live in the TWPhrases dictionary, but no source in this corpus writes 显像管 so it emits nothing (grep -roh '映象管' dist | wc -l -> 0). It is noted in the code comment beside the 鏡像 override so the next reader does not have to rediscover it.
Related
Filed unassigned, finding. No files changed.
Found while implementing #194 (PR for
claude/issue-194-twp-artifacts). Not fixed — reported only, because it is a contested rendering of the #195 class, not a mechanical artifact of the #194 class.What
I swept every
TWPhrasesentry that actually fires in this corpus (143 of the 603) to check whether #194 had left any sibling behind. It had not: every remaining candidate I flagged — 局部 → 區域性, 回调 → 回撥, 抛出 → 丟擲, 命名空间 → 名稱空間, 优先级 → 優先順序, 表达式 → 表示式, 外键 → 外部索引鍵 — is already a row in #195's table.One is not.
dist/zh-Hanttwpships连接器Measured on
ca3983b+ the #194 branch:The
distnumber is roughly double the source count because one of the hits is a post title —跨系统自动化：连接器和 Webhook 如何进入受治理流程— so it also reaches the blog index, the topic hubs,rss.xmland the<title>tag, the same chrome-amplification #192's patch round ran into.Why it belongs in #195 and not in #194
The #194 test is "wrong in any register" — 介面卡 names different hardware, 全域性 is not a noun.
聯結器fails no such test. It is a real Taiwanese word; it is simply the mechanical coupling sense first (a railway coupler, a connector housing), where the software/data sense is normally written 連接器 in Taiwan. That is a judgement, and #195 is explicitly the place judgements go: "No fix proposed. Nothing here is dispatchable until someone who reads Taiwanese technical writing has looked at it."Worth noting only because of volume. At 17 source occurrences it would sit fifth in #195's table, above ten of the fifteen rows already there — and one of those occurrences is a post title, which is the most visible position in the corpus.
What to do
Nothing, until #195 gets its reviewer. If and when it does, this is one more row on the same sheet, and it costs the same one line in
TWP_PHRASE_OVERRIDESplus one fixture case inCONVERSION_CASES(src/lib/zhconvert.ts) that every other row would cost.Also recorded, and deliberately not proposed as a defect:
十进制→ 十進位制 (1 occurrence) and顯像管→ 映象管 — the latter is a second instance of the exact 象/像 slip #194 fixes for 鏡像, still live in theTWPhrasesdictionary, but no source in this corpus writes 显像管 so it emits nothing (grep -roh '映象管' dist | wc -l-> 0). It is noted in the code comment beside the 鏡像 override so the next reader does not have to rediscover it.Related
s2twpover-substitutions are wrong in any register — 介面卡 is a network interface card, not an adapter #194 — the mechanical artifacts, fixed.Filed unassigned,
finding. No files changed.