Quick, also called Sucheng or Simplified Cangjie (速成, 簡易), is a Chinese input method that types a character with only the first and last keys of its Cangjie code. We counted what that costs across the 5,391 characters of plane 1 of Taiwan's CNS 11643, the set Oracle's Solaris documentation describes as the most commonly used. Full Cangjie gives 95.4% of them a code of their own. Quick gives that to 0.9%, including 23 one-key characters it cannot shorten. Among the 5,127 whose code it does shorten, 0.5% stand alone. The median character shares its two keys with 11 others: 的, 特, 得 and 等 all land on hi. We did not find a published count of these shared codes (重碼).
Quick input throws away the middle of a Cangjie code
A Cangjie code spells a character in up to five keys, each standing for a shape or radical (字根); how one is built has its own guide. 夏 is MUHE and becomes ME. 蘋 is TYHC and becomes TC. A code of one or two keys stays as it is.
Four sources state the same rule.
- Microsoft's guide, last updated 20 June 2024: "Select the two keys that correspond to the first and the last ChangJie elements". Its summary of Quick says "stroke" for "elements".
- Apple's user guide: "The first and last keys from a Cangjie code are used as a Sucheng code."
- A computer-awareness course appendix on a Hong Kong Education Bureau server; the PDF names no issuer and its file is dated January 2002: 「以首、尾兩碼輸入」.
- The open-source rime-quick schema: 「取倉頡首尾二碼爲速成碼」.
Microsoft's steps for adding Chinese (Traditional, Hong Kong SAR) add: "By default, Traditional Chinese support is installed with the Microsoft Quick IME." Apple's examples for 蘋果公司 and Microsoft's for 夏天 all reproduce from the tables below, our independent check.
Characters in Taiwan's standard set run into each other on almost every code
We applied the first-and-last rule to every code in the CNS 11643 open data released on 5 August 2026. A group means all the characters on one code, the character itself included; a character with two codes takes the smaller group. Oracle gives plane 1 as 5,401 characters. Ten of those, built on 卑, have no Cangjie row in the release and are not counted. A release dated 28 October 2026 has since followed; rerun on it, every figure here holds except one count in the next section, 27,451 characters, which becomes 27,455.
| Plane 1, 5,391 characters | Full Cangjie | Quick |
|---|---|---|
| Codes in use | 5,421 | 607 |
| Characters alone on a code | 5,142 (95.4%) | 51 (0.9%) |
| Median group | 1 | 12 |
| Largest group | 3 (qomr) | 42 (yr) |
By code, 51 of the 607 hold one character and 222 hold ten or more. In all, 3,438 characters (63.8%) sit on a code holding more than nine. Microsoft documents number keys for choosing a candidate and a next-page key, but no page length, so that says nothing firm about a first page.
Plane 1 is a Taiwan list, unweighted by use.
Collisions mount as the character table gets bigger
One code can list a dozen characters in one lookup and nearly a hundred in another: hf holds 13 in plane 1 and 93 in the Rime table below.
| Character set | Characters | Median group | Largest group |
|---|---|---|---|
| CNS plane 1 | 5,391 | 12 | 42 |
| CNS planes 1 and 2 | 13,021 | 29 | 69 |
| Rime Cangjie 5 ideographs, X-led codes dropped | 20,910 | 46 | 109 |
| CNS ideographs, Basic Multilingual Plane | 27,451 | 59 | 162 |
The rows are a scale, and the Rime table is not a subset of the CNS row beneath it. The four-corner code poses the same question on different sets.
Our Quick input lookup reads that Rime table. Until 1 November 2026 it showed Cangjie's X selector codes as ordinary Quick codes with a false 難 radical. X is a pick key: 曰 shares A with 日, so it also takes XA. That made xi, at 114 entries, read as the largest code, and the page refused non-ideographs forward that it listed in reverse. The revision of that date flags pick codes, serves both directions from one index and shows the Cangjie code beside each Quick code. Enter 夏 and the card gives me beside MUHE with 58 characters on the code; enter 曰 and xa is labelled a pick.
Everyday characters collide on hi, mm and mg
We coded the top 200 of our Chinese Wikipedia frequency count (dump of 1 September 2026) through the Rime table without X selector codes. Of the 200, 78 share a Quick code with another character inside the 200. For Cangjie the number is 2, 业 and 共, both TC. Besides the four on hi, 三, 二 and 工 are all mm, and 至, 理 and 王 are all mg.
The list is encyclopedic text in simplified forms, while Microsoft and Apple document Quick for Traditional Chinese.
Generation changes in Cangjie mostly leave the Quick code untouched
A character can carry different Cangjie codes in the system's 3rd and 5th generations. CNS names no generation, so we compared named tables: ibus cangjie3.txt against Rime's 5th-generation table, pick codes dropped, with 5,390 plane 1 characters in both.
Of the 109 plane 1 characters whose 3rd- and 5th-generation codes have nothing in common, 92 keep the same Quick code, 84.4% on our arithmetic. 面 moves from MWYL to MWSL and stays ML. The other 17 include 乍, HS in one table and OS in the other, because the first key moved.
A shorter code brings a longer candidate list
The mean shortest Cangjie code in plane 1 is 3.98 keys, so Quick saves about two and usually adds a selection. Apple's documentation states the trade-off plainly: "A Sucheng code has fewer keys than Cangjie, but more candidates, so it may take more time to select candidates in the candidate window."
The 2002 appendix calls Quick 「較倉頡易掌握」 and also says 「因要選字，所以輸入速度較慢」. An undated Hong Kong Police article, about one organisation at one moment, says selecting from a list "reduces the speed of input". We measured no net speed and give none.
Group size counts a table, and a typist's window need not match it. A real candidate window is ordered and may be filtered, and we tested none. Microsoft's example enters MK and then 1 for 天, which is 32nd of 66 on mk in table order.
The counts come with two required credits and a long list of unknowns
The CNS data is offered under the Open Government Data License or OFL-1.1, one at a time, and we use the former. That licence and the CNS site's terms make these statements a condition of use:
數位發展部 2026 CNS11643中文標準交換碼全字庫(簡稱全字庫) 全字庫屬性資料集對照表、全字庫中文碼對照表(2026年8月5日更新) 此開放資料依政府資料開放授權條款 (Open Government Data License) 進行公眾釋出，使用者於遵守本條款各項規定之前提下，得利用之。政府資料開放授權條款：https://data.gov.tw/license
資料來源：數位發展部，CNS11643中文標準交換碼全字庫網站，https://www.cns11643.gov.tw。
The top-200 list derives from Chinese Wikipedia, CC BY-SA 4.0, with credit to its contributors. The other tables are Rime's cangjie5.base and ibus-table-chinese's cangjie3.txt, the second for counting only.
We could not establish who invented Quick, or when; which places say 速成 and which 簡易; what share of typists use it; whether Windows or macOS accepts X-led codes; or which generation Microsoft's or Apple's Quick derives from, as neither page names one.
View author profile → · Editorial policy
