Re: [PATCH PR/42686] Align the help text output

["Joseph S. Myers" <joseph at codesourcery dot com>]

On Tue, 13 Apr 2010, Shujing Zhao wrote:

> > > +      else if (nbytes > 1)
> > 
> > I still see no reason you should need conditionals on the number of bytes in
> > a character, instead of always working with characters regardless of the
> > number of bytes in them.
> This condition is to make the string can be break after a multi bytes wide
> alphabetic or punctuations. The line can be break after every multi bytes wide
> alphabetic or punctuations. I think the nbytes can distinguish if it is a wide
> character before decode. The letter 'a' is recognized wide character after the
> decoding, but the line can't be broken after a 'a' at the string "after". I
> think the difference between one byte wide character and the multi bytes wide
> character is the nbytes.
> 
> Is it better changed to
> 
> +      else if ((iswalpha (wc[0]) || iswpunct (wc[0])) && nbytes > 1)
> 
> better?

I still cannot make sense of what the logical condition is that this code 
is trying to implement; that is, the logical properties of a pair of 
characters and what the conclusion is from those properties about whether 
a break is or is not permitted at a particular location in relation to 
those characters.  You have a series of conditions that might be described 
something like (and it's possible I'm not understanding your intent):

/* We can break at the end of the string if it is narrow enough.  */
/* If there is a space, we can break before the space (and not print the 
   space).  */
/* Break after '-' or '/' if the previous character was alphabetical, but 
   not in the middle of "--" (for example).  */

But what in plain English is the rule this code is trying to implement for 
a condition on breaking or not breaking the string?

Whatever the condition is, it relates to some logical properties of the 
characters in question.  The number of bytes is not a logical property; 
it's a physical property of the particular locale character set and may 
vary from character set to character set (gettext will automatically 
translate using iconv to the LC_CTYPE character set).

-- 
Joseph S. Myers
joseph@codesourcery.com