topic Re: awk or sed? in Operating System - HP-UX

awk or sed?

David Burgess — Mon, 22 Nov 2004 08:51:22 GMT

I've got a file that was created from a database extract. Each field has ()'s round it. I want to strip the brackets out. Simple except some of the fields contain brackets inside them and these need to stay. For example :-

(one) (two) (three (test)) ((test1) five)

Each row has a variable number of fields.

Any ideas?

Regards,

Dave.

Re: awk or sed?

Hein van den Heuvel — Mon, 22 Nov 2004 09:01:19 GMT

Sounds like you can safely remove the leading open and trailing close parens.
Furthermore, you can remove any close-space-open combos.
Suggesting the following solution:

perl -pe 's/^$//;s/$$//;s/\) $/ /g' file

Or if you do not need the space between open/close:

perl -pe 's/^\(|$$|\) \(//g' x

Hein.

Re: awk or sed?

Steven E. Protter — Mon, 22 Nov 2004 09:02:13 GMT

Hmmm.

sed could be used to prepare the data by strpping out all brackets.

sed s/(/ /g $varname

same thing for the close bracket.

Then a standard awk command

cat $var | awk `{print $1 $2 $3}`

So my answer with some tweaks is probably both.

cut command might be useful in replacing the awk command.

SEP

Re: awk or sed?

H.Merijn Brand (procura — Mon, 22 Nov 2004 09:10:05 GMT

What if the parens are to be saved if escaped?

(one $)(two)($three) ...

In that case Parse::RecDescent might help

Enjoy, Have FUN! H.Merijn

Re: awk or sed?

David Burgess — Mon, 22 Nov 2004 09:12:23 GMT

Hein,

Looks good. Execept I get a trailing ) after the four.

$ perl -pe 's/^$//;s/$$//;s/\) \(/ /g' test.txt

one two three (test) (test1) four)

Any ideas?

Regards,

Dave.

Re: awk or sed?

Hein van den Heuvel — Mon, 22 Nov 2004 09:21:54 GMT

Probably a space after the final paren.
Those are hard to 'see' in the itrc forums.
You'd have to append a few lines froma real file to analyze that.

If it is just optional trailing space(s) you can capture those with: \)\s*$

so:

perl -pe 's/^$|$\s*$|\) \(//g' x

Hein.

Re: awk or sed?

Muthukumar_5 — Mon, 22 Nov 2004 10:17:13 GMT

use tr command to do this easily as,

# echo "(one) (two) (three (test)) ((test1) five)" | tr -d '()'
one two three test test1 five

HTH.

Re: awk or sed?

Muthukumar_5 — Mon, 22 Nov 2004 10:19:07 GMT

sorry again.
using sed you can do this as,
# echo "(one) (two) (three (test)) ((test1) five)" | sed "s/(//g;s/)//g"
one two three test test1 five

HTH.

Re: awk or sed?

David Burgess — Tue, 23 Nov 2004 08:23:32 GMT

Thanks guys. The DBA's are happy and the day was saved!

Regards,

Dave.