Re: [^\]] in basic regexes

[Wacek Kusnierczyk]

Andreas Schwab wrote:
> Wacek Kusnierczyk <address@hidden> writes:
>
>   
>> if, as you seem to imply, the pattern is 'a non-backslash followed by a
>> closing bracket', then
>> - the rightmost closing bracket is *outside* of any class (the class
>> [^\] has been closed), thus
>>     
>
> Right.
>
>   
>> - it is (should be) a *metacharacter*, and and as such
>>     
>
> Wrong.
>
>   
>> - it must be taken as *closing a class*, which is not there and it
>> neither should match the closing bracket in the data (which it does in
>> some of the other examples)
>>     
>
> The ']' character is only special when part of a bracket expression.  In
> all other contexts it just matches itself.
>
>   


oh my.  right, thanks, this did not occur to me.  i suspected this is
what happens, but from what i read about bre, it was clear that both [
and ] are special characters, and thus the idea that an unpaired ]
outside of any class be taken literally seemed wrong.

this is surely documented, but i could only find, in various sources,
statements and examples to the effect that both [ and ] are
metacharacters, and that ] is taken literally when it's the first
character within a class.

it could probably be taken as obvious that an unpaired ] is literal, but
then the following seems a bit incoherent:

egrep -o ']' <<< ']'
# ]
egrep 'o ')' <<< ')'
# egrep: unmatched ) or \)

why couldn't unmatched ) be taken literally too?

thanks for bearing with my silly questions.
vQ