Very Computer

I want to use some grep command to match lines that contain
some word and NOT followed by another.
Examples:
    [1]    ^foo$
    [2]    ^foobar$
    [3]    ^foo foobar$
    [4]    ^fooba$

I want to grep on 'foo' NOT followed by 'bar'. So line 1, 3 and
4 would match. (line 3 has a foo NOT followed by bar)

Question:
    How should I do this?
    I tried a few things but did not succeed. (I rtfm, the Unix FAQ
    and O'Reilly's Unix in a Nutshell I believe)
    This may not matter but I try to do this on AIX 3.2.5

Cheers,
Vincent Zocca

PS: I noteced that our nn is freaked up (21 hours no news and I am
    no longer sysadmin) so an email would help me quicker I believe.

    VZ> I want to grep on 'foo' NOT followed by 'bar'. So line 1, 3
    VZ> and 4 would match. (line 3 has a foo NOT followed by bar)

Negating a regular expression is tricky business.  For this relatively
easy example, you need something like:

egrep 'foo([^b]|b[^a]|ba[^r])' baz.txt  #Those parens need to be escaped, really

As you can see, the regexp will quickly get very complicated.  Here's
one for regular, not followed by expression:

egrep 'regular ([^e]|e[^x]|ex[^p]|exp[^r]|expr[^e]|expre[^s]|expres[^s]|express[^i]|expressi[^o]|expressio[^n])' baz.txt

(An elisp macro makes short work of it.  If you use emacs, let me
know)

--
                                                         =[ cooper ]=

It's much easier to grep twice:

grep foo filename | grep -v foobar

Barny ;-{)

--
----

 - save a copy of the input line.
 - replace every 'foobar' by a space.
 - check whether there is any 'foo' left.
 - if so, print the saved line.

In sed that becomes:

sed -e '
h
s/foobar/ /g
/foo/ {
g
p

Quote:}

The perl equivalent:

perl -ne '
$save = $_;
s/foobar/ /g;
if (/foo/) {print $save;}
'

Quote:> PS: I noteced that our nn is freaked up (21 hours no news and I am
>     no longer sysadmin) so an email would help me quicker I believe.

--
Uwe Waldmann, Max-Planck-Institut fuer Informatik
Im Stadtwald, D-66123 Saarbruecken, Germany

grep ^foo filename | grep -v ^foobar

Soren
--
Soren Dayton

Quote:>I want to use some grep command to match lines that contain
>some word and NOT followed by another.
>Examples:
>    [1]    ^foo$
>    [2]    ^foobar$
>    [3]    ^foo foobar$
>    [4]    ^fooba$

>I want to grep on 'foo' NOT followed by 'bar'. So line 1, 3 and
>4 would match. (line 3 has a foo NOT followed by bar)

--
Lloyd Zusman            01234567 <-- The world famous Indent-o-Meter.

  To get my PGP public key automatically mailed to you, please send me email
  with the following string either as the subject or on a line by itself:
                        mail-request: public key

egrep 'foo($|[^b]|b$|b[^a]|ba$|ba[^r])' file.txt

(Elaborated)

Vincent

A single 'grep -v foobar' will do as well. Which is almost what he wants.

Try:

        grep -v '^foobar$' Examples

On to the next wave...

        Ed Allen

sed -ne '/foo/{h;ba;}
         d
         :a
         s/foobar/+/g
         /foo/{x;p;}'

does the job just fine.  Amazing what sed will do for you...

- Harald

#!/bin/ksh

echo foo         > tt
echo foobar     >> tt
echo foo foobar >> tt
echo fooba      >> tt
echo abc foobar >> tt
echo abc        >> tt

gawk '{ if ( $0 ~ /foo/ ) \
           { \
           if ( $0 !~ /foobar/ ) \
              { print $0 } \
           else { \
              line = $0; \
              i = 1; \

              if ( line ~ /foo/ ) { print $0 } \
              } \
           } \
      }' tt

R. Corr

| Seems to me you could do it with sed:
|
| sed -ne '/foo/{h;ba;}
|          d
|          :a
|          s/foobar/+/g
|          /foo/{x;p;}'

Good method.  If you're laboring with a sed that doesn't support semicolons,
though, you can write the same logic this way:

  sed 'h
       s/foobar/+/g
       /foo/!d
       x'

You could also use 'g' instead of 'x' at the end.

|    The solution is :
|
|  sed 's/foobar/hooops/g' <filename> | grep foo | sed 's/hooops/foobar/g'

You can do it in one call to sed and fork one program instead of three:

  sed -e 's/foobar/hooops/g' -e '/foo/!d' -e 's/hooops/foobar/g'

assuming, as your suggestion does, that "hooops" doesn't occur in the input.
To get around even that assumption,

  sed -e h -e 's/foobar/hooops/g' -e '/foo/!d' -e g

 sed 's/foobar/hooops/g' <filename> | grep foo | sed 's/hooops/foobar/g'

^splat foobar$

here is an awk statement which does what you want:

awk '$0 ~ /foo/{
        if(($0 ~ /foobar/) && ($0 ~ /foo[^b]/)) print $0;
        if($0 !~ /foobar/) print $0;
        }' testgrep

a bit unwieldy, though, I admit.

I do not know if grep can accomodate conditionals, which is the problem
we run into.

is there regexp syntax similar to the use of [^c] which matches all
non-c's?  

if such syntax existed, the following statement would work:

grep '/foo\[^bar]\/' testgrep

where \[ and \] are grouping symbols like \( and \) and ^ tells it to
look for anything which is not bar that follows foo.

but I suppose I just have highly irregular dreams...  :)

Janet

--

   sed -n -e '/foobar/d' -e '/foo/p' file >out