Need a little 'sed' help

[awingnut]

I have a series of lines in which there is string data I need to extract that is enclosed in parens "()". There is other "stuff" outside the parens and each set of data is separated with a comma. I cannot seem to come up with a working 'sed' command(s) to effectively delete everything except what is in parens and separating commas. In other words a line might look like:

blah blah blah (good stuff here) maybe some blah, more blah (more good stuff here), ...

Can one of you regexp experts help me with this? TIA.

 

[PHV]

Something like this ?
sed 's!)[^)]*$!)!;s![^(][^(]*(\([^)][^)]*\))!\1,!g;s!,$!!' /path/to/input > output

Hope This Helps, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884 or FAQ181-2886

 

[duncdude]

ouch.

this works in Perl:-

[b]#!/usr/bin/perl[/b]

$_ = 'blah blah blah (good stuff here) maybe some blah, more blah (more good stuff here), ...';

s/[^(]*\(([^)]*)\)[^(]*/$1\n/g;

print;

i am not a sed head - but i hope this helps!

outputs:-

[red]good stuff here
more good stuff here[/red]


Kind Regards
Duncan

 

[awingnut]

PHV:

WOW! It works. Thanks. Yikes, what a cryptic mess. I figure it will take me a month to figure out what you did.

duncdude:

Thanks but I am not writing a perl script. However, looking a yours and PHV's, I'm wondering why.

 

[PHV]

A simplified version:
sed 's![^(]*(\([^)]*\))[^(]*!\1,!g;s!,$!!' /path/to/input > output

Hope This Helps, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884 or FAQ181-2886

 

[duncdude]

Hi awingnut

It gets even easier, in Perl, with an array:-

[b]#!/usr/bin/perl[/b]

$_ = 'blah blah blah (good stuff here) maybe some blah, more blah (more good stuff here), ...';

@matches = $_ =~ m/(\([^)]+\))/g;

print join("\n", @matches);

I realise you are not writing this in Perl - i just thought it would be interesting for you to see the far simpler regex


Kind Regards
Duncan

 

[awingnut]

My I impose for another sed expression?

I have a series of lines that have the word 'case' followed by more data. I need to insert a space between 'case' and whatever follows it.

I used to know regexp but using it in sed seems to be slightly different (or my memory is failing which is more likely). Anyway what I can't get to work is:

sed 's!^case*$!^case *$!'

TIA.

 

[PHV]

You attempt suggests that you're interested on lines beginning with 'case':
sed 's!^case\([^ ]\)!case \1!' /path/to/input > output

Hope This Helps, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884 or FAQ181-2886

 

[awingnut]

Thanks but could I impose for a little tutoring here?

!^case -- I get this, it means the beginning of the line starts with'case'

\( --escape the '('

[^] -- I don't get this. I would expect to use \*$ meaning anything to the eol?

\) -- excape the matching ')'

!case -- puts back case at the beginning

\1 -- ???? A sed thing that refers to what was collected by [^] which explains the above ?????

 

[PHV]

A starting point:
man sed

Hope This Helps, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884 or FAQ181-2886

 

[awingnut]

If you don't have the time to answer a specific question then don't. The sarcasm was unnecessary.

 

[PHV]

Absolutely no sarcasm but a pointer for you to get the answer.

 

[awingnut]

Sorry, I just don't understand why you would think I am unaware of the man pages or would have asked here if I found a clear answer there, in a reasonable time. Are there really shell script writers here that do know know about man pages?

 

[PHV]

Are there really shell script writers here that do know know about man pages?
Unfortunately YES !!!
's!^case\([^ ]\)!case \1!'
\(...\) is a subexpression
\1 is a back-reference to the 1st subexpression
[^ ] match any character but a space

Hope This Helps, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884 or FAQ181-2886

 

[feherke]

Hi

Yes, we use the man pages regularly, but sadly, the [tt]sed[/tt] man page is not complete. The full list of features seems to be on it's web site :

http://www.gnu.org/software/sed/manual/html_mono/sed.html

Beside this, I think you find [tt]sed[/tt]'s man page unclear because you are not enough familiar with regular expressions. Take a look at this too :

http://www.regular-expressions.info/tutorial.html

Feherke.

http://rootshell.be/~feherke/