On 09/01/2010 03:45 PM, Jim Meyering wrote:
It's a huge mistake, because making it an error means
changing the regex grammar (and making it unnecessarily complicated
and contrived).
It's already done in grep, and wasn't a very big change.
You're conflating the formal grammar and the code that implements
it. The latter is allowed to include this kind of "hack", the former
cannot. POSIX
It's done for grep.
I see no need to change any formal grammar.
Do you honestly imagine
http://opengroup.org/onlinepubs/007908775/xbd/re.html changed to say "If
a bracket expression consists of a colon, a series of alphabetic
characters, and another colon, the implementation's behavior is undefined?
It sounds like you're upset.
Sorry it's come to that, but I feel strongly about this, too.
I'm not upset, but I'm very much worried of making a mistake. What if
someone proposes to warn for [A-Z] and [a-z] whenever they're not
going to do what the user thinks (they're not synonyms of any of
[[:upper:]], [[:lower:]], [[:alpha:]])? Are we going to make that an
error?
"We have nothing to fear but fear itself".
But I doubt you're seriously worried about [A-Z], etc.
Not worried, but it honestly would be a sensible warning whenever
strcoll ("A", "f") <= 0 && strcoll ("f", "Z") <= 0, as is the case in
UTF-8 locales with glibc. It will basically never do the right thing,
since [A-Z] and [a-z] exclude a and Z respectively.
Paolo
