[SOLVED] Need help with SED command

guessity · 01-17-2010, 01:58 AM

Hi,
I have been trying to remove few lines of text in a file -

I basically want to remove the following -

Code:

<?xml version="1.0" encoding = "iso-8859-1"?>
<!--
All Rights Reserved.
-->
<!--
This feed is made by using XML Feed Maker v2.2.
-->

I used this command

Code:

sed '/<?xml\/version"/,/-->/d' article1.txt > u.txt

any idea where I went wrong?

konsolebox · 01-17-2010, 02:51 AM

Code:

sed '/<\?xml\/version"/,/-->/d' article1.txt > u.txt

guessity · 01-17-2010, 03:06 AM

Quote:

Originally Posted by konsolebox

Code:

sed '/<\?xml\/version"/,/-->/d' article1.txt > u.txt

Thats what I had typed. the problem is that the text still remains in the new file.

colucix · 01-17-2010, 03:12 AM

If you want to edit the file in place, use the -i option. Be careful with that: if you are not really sure that the result is exactly what you're trying to do, the loss of the original file is at risk. Anyway, you can use the -i option in the following way to do a safe backup copy:

Code:

sed -i.bck 'blah blah blah' file

this will create a copy of th original file, called file.bck, where .bck is just an example. Actually, you can choose the suffix at your pleasure.

guessity · 01-17-2010, 03:33 AM

Quote:

Originally Posted by colucix

If you want to edit the file in place, use the -i option. Be careful with that: if you are not really sure that the result is exactly what you're trying to do, the loss of the original file is at risk. Anyway, you can use the -i option in the following way to do a safe backup copy:

Code:

sed -i.bck 'blah blah blah' file

this will create a copy of th original file, called file.bck, where .bck is just an example. Actually, you can choose the suffix at your pleasure.

I did this

Code:

sed -i '/<\?xml\/version"/,/-->/d' article1.txt > u.txt

I ended with a blank file u.txt and nothing in article1.txt changed

(

What could be wrong?

David the H. · 01-17-2010, 03:48 AM

What's with all the escaping? I just used the following:

Code:

sed '/<?xml version/,/-->/d'  file

Of course, this only removes the first commented section. To remove the second one, you'd need to modify it. Probably the easiest way to do it is to simply address the number of lines following the first match (assuming gnu sed).

Code:

sed '/<?xml version/,+6d'  file

A useful way to debug a sed expression is to first set it to print only the lines you want to match first. Then once you get it working correctly, reverse it.

Code:

sed -n '/<?xml version/,+6p'  file

PS: I often use the sed faq to help me with my matching needs.

David the H. · 01-17-2010, 04:44 AM

Quote:

Originally Posted by guessity

I did this

Code:

sed -i '/<\?xml\/version"/,/-->/d' article1.txt > u.txt

I ended with a blank file u.txt and nothing in article1.txt changed

(

What could be wrong?

">u.text" redirects standard output to the new filename. Since you used -i to edit the original file in place, there's no output to redirect. And since the expression failed to match, there was nothing to edit in the original file either.

colucix · 01-17-2010, 07:44 AM

As David said!

konsolebox · 01-19-2010, 06:03 AM

Quote:

Originally Posted by David the H.

What's with all the escaping? I just used the following:

Code:

sed '/<?xml version/,/-->/d' file

My mistake. I didn't notice that it was space. Anyway shouldn't '<?' mean an optional '<' and not '<?' ? I haven't tested it though.

David the H. · 01-19-2010, 07:13 AM

Quote:

Originally Posted by konsolebox

Anyway shouldn't '<?' mean an optional '<' and not '<?' ? I haven't tested it though.

In extended regexp mode, yes. But without the -r flag, sed uses only a simplified "basic" regex syntax and "?" loses its special meaning (along with some other characters).

In fact, in basic mode, escaping the question mark actually enables the extended regex functionality!

The grep manpage goes into more detail about the differences between basic and extended regex, since it works the same way.

konsolebox · 01-19-2010, 07:31 AM

i see thanks for the info.

pixellany · 01-19-2010, 07:37 AM

Anyone who can remember all the differences between regular and extended Regexes---AND how those differences vary among programs---is worthy of sainthood.

I have developed a simple approach: Trial and Error

chrism01 · 01-19-2010, 05:37 PM

I highly recommend the OWL book http://regex.info/

Tinkster · 01-19-2010, 07:21 PM

Indeed ... Friedl is a genius :}