Re: sed escaping question

[Stepan Kasal]

Hello,

On Tue, May 23, 2006 at 07:08:43PM +0200, Ralf Wildenhues wrote:
> Regarding this: (info Autoconf "Limitations of Usual Tools"):
> 
> | Portable `sed' regular expressions should use `\' only to escape
> | characters in the string `$()*.0123456789[\^n{}'.  For example,
> | alternation, `\|', is common but Posix does not require its
> | support, so it should be avoided in portable scripts.  Solaris
> | `sed' does not support alternation; e.g., `sed '/a\|b/d'' deletes
> | only lines that contain the literal string `a|b'.  Similarly, `\+'
> | and `\?' should be avoided.
> 
> Does this mean that `]' in a sed regex should not be escaped if it is to
> match a literal `]'?

Definitly.  `]' is not a special character in BREs (nor in EREs).
You can choose between `]' and `[]]', but `\]' is undefined.

> Conversely to the second half of the paragraph, can we be certain that
>   sed 's|a\|b||'
> 
> does what I think it should do, namely remove a literal `a|b' from the
> code, and not invoke alternation?  Or should a different delimiter be
> preferred for safety?

I read the Posix definition right now, and I can confirm that it
seems to define this meaning.  (I don't know anything about actual
implementations.)

My understanding is that the above paragraph is not speaking about
this; it's a special exception that \<delimiter> denotes the literal
character <delimiter> in the `s' command.  (I'm not sure, but perhaps
a clarification could be added there, in paretheses).

Have a nice day,
        Stepan