Full-Width and Half-Width Characters

Why Two Widths Exist

In Japanese, Chinese and Korean text every character fills a square cell. Early Japanese computers had two kinds of character sets: a single-byte set, JIS X 0201, with Latin letters, digits and a small set of katakana that took half a cell, and double-byte sets such as JIS X 0208 with kanji, kana and a second copy of the Latin letters and digits that filled a whole cell. The Korean standard KS X 1001 also has a full-width copy of the Latin alphabet.

Unicode kept both copies so that old text could be converted to Unicode and back. The full-width forms of the ASCII characters (U+FF01–FF5E) and the half-width katakana (U+FF61–FF9F) sit in the "Halfwidth and Fullwidth Forms" block, and the Unicode Character Database records which ordinary character each one stands for. That record is what a converter uses.

What Has a Full-Width and a Half-Width Form

Kanji, hiragana and Hangul syllables have only one width. There are no half-width hiragana, and the small katakana ヵ, ヶ and ヮ and the old letters ヰ and ヱ have no half-width form either.

Where the Difference Causes Trouble

• Forms. A field that checks for digits rejects ０９０, and a postcode or card number copied from a Japanese page is often full width. Converting to half width fixes it.
• Search and duplicates. A search for ABC may not find ＡＢＣ, and a customer list can hold the same name twice in two widths.
• Layout. Full-width letters in English text look spaced out; half-width katakana in Japanese text look cramped and old-fashioned.
• Systems with their own rules. Some Japanese forms want the reading of a name (フリガナ) in full-width katakana, others in hiragana, and some older systems accept only half-width katakana. Follow what the form says.

Half-Width Katakana and the Sound Marks

Half-width katakana have no letters with voicing marks. ガ is written as two characters, ｶ followed by the voiced sound mark ﾞ, and パ as ﾊ followed by the semi-voiced mark ﾟ. A good converter joins the pair into one full-width letter and splits it again on the way back. That is why the number of characters changes: ｶﾞｯｺｳ has five characters, ガッコウ four.

A sound mark that cannot join the letter before it, as in ｱﾞ, stays a separate mark (ア゛). ヴ becomes ｳﾞ, and ヷ and ヺ become ﾜﾞ and ｦﾞ.

Look-Alikes That Are Not Pairs

The backslash is a special case. JIS X 0201 put the yen sign at the code that ASCII uses for the backslash, so some Japanese fonts still draw the backslash as ¥. The character is the same; only its picture differs.

Hiragana and Katakana

Every hiragana letter from ぁ to ゖ has a katakana letter of the same name, and Unicode places them 96 code points apart, so converting between them is exact: やまだ たろう becomes ヤマダ タロウ. The long vowel mark ー is shared by both scripts and stays as it is. Katakana with no hiragana letter of their own, such as ヷ, stay katakana.

For Programmers: Normalization Does More

Unicode normalization form NFKC also turns full-width letters into half-width ones and joins ｶﾞ into ガ, but it changes many other characters at the same time: ① becomes 1, № becomes No, ㈱ becomes (株) and ㍻ becomes 平成. Use it when you want all of that, for example to compare search keys. To change only the width, convert the width pairs and leave everything else alone.

Frequently Asked Questions