The Unicode standard does not formalize the notion of a character
width. Roughly based on https://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c,
https://github.com/nodejs/node/blob/master/src/node_i18n.cc,
and UAX #11 we proceed as follows.
The following code points are of width 0:
code points with general category (see stringi-search-charclass)
Me
, Mn
, and Cf
),
C0
and C1
control codes (general category Cc
)
- for compatibility with the nchar
function,
Hangul Jamo medial vowels and final consonants
(code points with enumerable property UCHAR_HANGUL_SYLLABLE_TYPE
equal to U_HST_VOWEL_JAMO
or U_HST_TRAILING_JAMO
;
note that applying the NFC normalization with stri_trans_nfc
is encouraged),
ZERO WIDTH SPACE (U+200B),
Characters with the UCHAR_EAST_ASIAN_WIDTH
enumerable property
equal to U_EA_FULLWIDTH
or U_EA_WIDE
are
of width 2.
Most emojis and characters with general category So (other symbols)
are of width 2.
SOFT HYPHEN (U+00AD) (for compatibility with nchar
)
as well as any other characters have width 1.