[SOLVED] using bash sed to do the following

ghantauke · 12-02-2010, 07:09 AM

Assume I want to put all occurances of "ha" or "boowa" inside ::. so input "ha" gives output ":ha:".

I could do the following.

Code:

sed 's/\(ha\)/:\1:/g' test
sed 's/\(boowa\)/:\1:/g' test

But I want to use logical or and do it in 1 line of code.
sed 's/\(ha\) LOGICAL OR \(boowa\)/:\1: LOGICAL OR :\2:/g' test
Is there anyway to do that? The [] only only matches characters and not a string.

fordeck · 12-02-2010, 07:27 AM

So will this do what you want?

Code:

$ sed -e 's/\(ha\)/:\1:/g' -e 's/\(boowa\)/:\1:/g' file.txt 
:ha:
:boowa:
boo
foo

Regards,

Fordeck

ghantauke · 12-02-2010, 07:42 AM

Quote:

Originally Posted by fordeck

So will this do what you want?

Code:

$ sed -e 's/\(ha\)/:\1:/g' -e 's/\(boowa\)/:\1:/g' file.txt 
:ha:
:boowa:
boo
foo

Regards,

Fordeck

Thanks. Also how can I do a logical and? For example consider the following data in file called file.txt

Code:

Hey
'nt
've
'thou'
em'
last

Suppose I want to remove single quotes from the word only if it is surrounded by it.
So I'd only want 'thou' to be changed to thou. The following approach doesn't work (obviously)

Code:

sed "s/^'//" file.txt | sed "s/'$//" >result

So what I want is a logical and in the sed command. Thanks in advance.

grail · 12-02-2010, 07:49 AM

So how about:

Code:

sed -r 's/(ha|boowa)/:\1:/g' file.txt

For your second one it would be soomething like:

Code:

sed -r "s/'([^']*)'/\1/g" file.txt

ghantauke · 12-02-2010, 07:53 AM

ignore this accidentally posted it

ghantauke · 12-02-2010, 07:59 AM

Quote:

Originally Posted by grail

So how about:

Code:

sed -r 's/(ha|boowa)/:\1:/g' file.txt

For your second one it would be soomething like:

Code:

sed -r "s/'([^']*)'/\1/g" file.txt

Well thanks m8

angel115 · 12-02-2010, 08:04 AM

This one will do it too:

Code:

sed -e "s/^'\(.*\)'/\1/g" file.txt

grail · 12-02-2010, 08:11 AM

@angel115 - Although the shown input would work with your sed it would fail if more than one quoted item on a line (this is even if you remove the ^ so to look at others in the line)
Also, -r saves you worrying about escaping the brackets

sycamorex · 12-02-2010, 08:18 AM

Sorry I'm at work now on Windows, but wouldn't it work too?:

sed -r 's/ha|boowa/:&:/g' file.txt

or for exact matches (ie. 'ha' and not 'haaaa' or 'aha')

sed -r 's/\<ha\>|\<boowa\>/:&:/g' file.txt

ghantauke · 12-02-2010, 10:27 AM

This is the solution I've come up with.

Code:

sed "s/^'\([a-z']\)'$/\1/"

Seems to work perfectly. Thanks for all your help.