[SOLVED] Limited newline replacement

paullanders · 06-03-2011, 10:45 AM

I have a file where I would like to replace a newline with a space only if it follows an anchor and 2 digits:

infile:

Quote:

12
This is a line of text
1987
13
This is another line
12/14/2001

outfile:

Quote:

12 This is a line of text
1987
13 This is another line
12/14/2001

I tried the following with sed and it does not work:

Quote:

cat infile.txt | sed -r 's/(^[0-9][0-9])$/\1 /gm'

Could someone please advise on a tool to do this?

Thanks!

krizzz · 06-03-2011, 11:30 AM

sed -e 'N;N;s/$^[0-9][0-9]$\n/\1 /g'

This is it

paullanders · 06-03-2011, 11:43 AM

This seems to partly work, but it matches only every other occurance. For example:

infile:

Quote:

12
This is a line of text
1987
13
This is another line
12/14/2001
14
This is a line of text
1987
15
This is another line
12/14/2001

outfile:

Quote:

12 This is a line of text
1987
13
This is another line
12/14/2001
14 This is a line of text
1987
15
This is another line
12/14/2001

It matched 12 and 14, but not 13 or 15.

grail · 06-03-2011, 12:07 PM

Code:

awk 'ORS = /^[0-9][0-9]$/?" ":"\n"' file

krizzz · 06-03-2011, 12:53 PM

Quote:

Originally Posted by paullanders

This seems to partly work, but it matches only every other occurance.

Yes, I edited the post right after posting it (noticed the error) buy you were quicker

This is the right syntax. You don't have to use cat, pass the file directly to the sed:

sed -e 'N;N;s/$^[0-9][0-9]$\n/\1 /g' filename.txt

Tested and it works.

crts · 06-03-2011, 01:10 PM

Quote:

Originally Posted by krizzz

sed -e 'N;N;s/$^[0-9][0-9]$\n/\1 /g' filename.txt

This will still fail if the data looks like this:

Code:

1987
12
This is a line of text
13
This is another line
12/14/2001
14
This is a line of text
1987
15
This is another line
12/14/2001

Alternative:

Code:

sed -r '/^[0-9]{2}$/ {N;s/\n/ /}' file

krizzz · 06-03-2011, 01:14 PM

Code:

root@chris-desktop:/tmp# cat test.txt 
12
This is a line of text
1987
13
This is another line
12/14/2001
14
This is a line of text
1987
15
This is another line
12/14/2001
root@chris-desktop:/tmp# sed -e 'N;N;s/\(^[0-9][0-9]\)\n/\1 /g' test.txt 
12 This is a line of text
1987
13 This is another line
12/14/2001
14 This is a line of text
1987
15 This is another line
12/14/2001
root@chris-desktop:/tmp#

paullanders · 06-03-2011, 01:51 PM

Works perfectly. Thanks so much!

I'm not familiar with the curly brackets and N in sed. Why is it:

Quote:

sed -r '/^[0-9]{2}$/ {N;s/\n/ /}' file

rather than:

Quote:

sed -r '/^[0-9]{2}$/ s/\n/ /' file

crts · 06-03-2011, 03:45 PM

Quote:

Originally Posted by paullanders

Works perfectly. Thanks so much!

I'm not familiar with the curly brackets and N in sed.

Well, the first set of {} specifies a quantifier, i.e. sed will only match digits if there are exactly two of them.

The second pair of {} simply groups the commands 'N' and 's', i.e. only if a match occurs will those two commands be executed. Normally, one would have to escape the first set of {}. However, the '-r' option activates sed's extended RegEx mechanism. Therefore it can determine the meaning of {} by context.

The 'N' command simply reads the next line and appends it to the pattern-buffer. E.g., this is how your pattern-buffer looks like before and after 'N' is executed:

Code:

# before 'N' is executed
12

# afeter 'N' is executed
12\nThis is a line of text

Hope this clears things up a bit.
If you are interested in more details, here is one of the very best online tutorials on 'sed':
http://www.grymoire.com/Unix/Sed.html