Prev: search pattern and display lines around the pattern
Next: Processing stdin in blocks
From: Ed Morton on 28 Feb 2010 23:06
On 2/28/2010 11:48 AM, August Karlstrom wrote:
> Hi,
>
> How do I escape all meta characters in $foo in the command
>
> sed s/$foo/bar/
>

Hang on, isn't that the same as asking "How do I replace one STRING (as opposed
to a REGEXP) with another?"? It seems like all you want is just to have $foo
treated as a string so instead of using a RE replacement mechanism like the
above you could just use awk functions that operate on strings as opposed to
REs, e.g.:

awk -v old="$foo" -v new="bar" '
strt=index($0,old){ $0 = substr($0,1,strt-1) new substr($0,strt+length(old)) }1'

Regards,

Ed.
From: Ed Morton on 1 Mar 2010 14:48
On Feb 28, 3:51 pm, August Karlstrom <fusionf...(a)gmail.com> wrote:
> Teemu Likonen wrote:
> > * 2010-02-28 18:48 (+0100), August Karlstrom wrote:
>
> >> How do I escape all meta characters in $foo in the command
>
> >> sed s/$foo/bar/
>
> >     #!/bin/sh
> >     quote-regexp () {
> >         printf '%s\n' "$1" | sed -e 's,[]/[\^$.*],\\&,g'
> >     }
>
> >     foo=something
> >     foo_escaped=$(quote-regexp "$foo")
> >     sed -e "s/$foo_escaped/bar/"
>
> > Note that the function escapes "/" character too because the character
> > will be used in the final sed's s/// command.
>
> Thanks Teemu. I thought there would be a standard command for quoting
> regular expressions.
>
> August- Hide quoted text -
>
> - Show quoted text -

The standard way of not treating an expression as an RE is to not use
RE operators on it (e.g. seds "//" or awks "*sub()" functions).

Unfortunately I can't think of a counter-example but I don't think
Teemu's approach of sticking a backslash in front of BRE
metacharacters will work for all possible BREs and of course it
certainly won't for EREs (which I think some seds support these days
and all awks support).

You should really consider just using a tool that can treat a string
like a string such as the awk example I posted or, apparently, perl.

Ed.


From: August Karlstrom on 2 Mar 2010 13:33
Ed Morton wrote:
> Hang on, isn't that the same as asking "How do I replace one STRING (as
> opposed to a REGEXP) with another?"? It seems like all you want is just
> to have $foo treated as a string so instead of using a RE replacement
> mechanism like the above you could just use awk functions that operate
> on strings as opposed to REs, e.g.:
>
> awk -v old="$foo" -v new="bar" '
> strt=index($0,old){ $0 = substr($0,1,strt-1) new
> substr($0,strt+length(old)) }1'

Thanks Ed. I still find it hard to see why such a common task should
require all that syntax.


August
From: Janis Papanagnou on 2 Mar 2010 13:47
August Karlstrom wrote:
> Ed Morton wrote:
>> Hang on, isn't that the same as asking "How do I replace one STRING
>> (as opposed to a REGEXP) with another?"? It seems like all you want is
>> just to have $foo treated as a string so instead of using a RE
>> replacement mechanism like the above you could just use awk functions
>> that operate on strings as opposed to REs, e.g.:
>>
>> awk -v old="$foo" -v new="bar" '
>> strt=index($0,old){ $0 = substr($0,1,strt-1) new
>> substr($0,strt+length(old)) }1'
>
> Thanks Ed. I still find it hard to see why such a common task should
> require all that syntax.

This is, actually, a task that I find not very well suported by awk.
(Too many examples like that may repel people, I fear - well, with
the exception of php programmers, maybe. ;-)

Janis

>
>
> August
From: Ed Morton on 2 Mar 2010 15:09
On Mar 2, 12:33 pm, August Karlstrom <fusionf...(a)gmail.com> wrote:
> Ed Morton wrote:
> > Hang on, isn't that the same as asking "How do I replace one STRING (as
> > opposed to a REGEXP) with another?"? It seems like all you want is just
> > to have $foo treated as a string so instead of using a RE replacement
> > mechanism like the above you could just use awk functions that operate
> > on strings as opposed to REs, e.g.:
>
> > awk -v old="$foo" -v new="bar" '
> > strt=index($0,old){ $0 = substr($0,1,strt-1) new
> > substr($0,strt+length(old)) }1'
>
> Thanks Ed. I still find it hard to see why such a common task should
> require all that syntax.
>
> August

I agree. Its a mystery why just replacing one string with another
should be harder to code than replacing an RE with a string. Sounds
like perl may have done something about that but I don't know the
details.

Ed.
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: search pattern and display lines around the pattern
Next: Processing stdin in blocks