bug#47857: fill-paragraph vs. FULLWIDTH DIGITs: bug exposed

[Eli Zaretskii]

> From: 積丹尼 Dan Jacobson <jidanni@jidanni.org>
> Cc: 47857@debbugs.gnu.org
> Date: Tue, 20 Apr 2021 07:27:05 +0800
> 
> EZ> I don't see any bug here.  This is the documented behavior: filling
> EZ> deletes any excess whitespace, so 2 or more consecutive SPC characters
> EZ> are squeezed to a single SPC.
> 
> Ah, but one needs
> 
> Today's lucky number(s)!: <-added space here
> ９ <-added space here
> １０ <-added space here
> １１
> 
> to keep the whitespace in with fill-paragraph, whereas no spaces are
> required for
> 
> Today's lucky number(s)!:
> 9
> 10
> 11
> 
> To safely become:
> 
> Today's lucky number(s)!: 9 10 11

This is a feature: full-width Latin characters, such as ０, １, and ９,
are exempted from converting newlines into spaces during filling.
They follow the rules of other CJK characters, which are processed the
same.  I think it's because of line-breaking rules: we can break the
line between any 2 CJK characters, including those above, and thus
adding spaces where there were newlines is superfluous, as it would
artificially create separate words where only one has been before.