Should character classes be mutually exclusive?

[YDX-2147483647]

Character classes in the current jlreq are not mutually exclusive.

• % (U+0025) ∈ Postfixed abbreviations (cl-13) ∩ Western characters (cl-27).
• ! (U+0021) ∈ Dividing punctuation marks (cl-04) ∩ Western characters (cl-27).
• ， (U+002C) ∈ Commas (cl-07) ∩ Grouped numerals (cl-24) ∩ Western characters (cl-27).
• …

Is this intentional? If so, e.g., in Table 1 Spacing between characters, which class should take precedence if the character hits multiple rows/columns?

Background: Unicode draft #59: East Asian Spacing suggests different treatment of 增加20%以后 (Chinese) and 進捗は20%です (Japanese). On Typst forum, we found % belongs to both cl-13 and cl-27.

[okumuralab]

Thanks for clarifying the problem. jlreq character classes are indeed not mutually exclusive. A Twitter user directed me to §3.9.2 aa, “Western characters (cl-27),” which states that “several marks are used both in a Japanese context and a Western context.” This could suggest that the ‘%’ symbol belongs to cl-13 in a Japanese context and cl-27 in a Western context. jlreq stops here, but in practice ‘％’ (U+FF05) is used in a Japanese context, while ‘%’ (U+0025) is used in a Western context.

So the simple solution is to treat ‘%’ (U+0025) as a Western character (cl-27) and insert a quarter-em space between it and most Japanese characters, and treat ‘％’ (U+FF05) as a Japanese postfixed abbreviation (cl-13).

For ‘‰’ (U+2030), which is also both cl-13 and cl-27, there is only one Unicode code point, and no such simple solution exists. YDX-2147483647 lists other examples. My intuition is that ‘‰’ is a Western character by default, ‘,’ and ‘!’ are a comma and a punctuation mark, but I hope users can set each character's class by some means.