1

I need a command / script to get the CVS merge conflicts. Basically, I need a SED or AWK command to print only lines that are between the two set patterns

Example:

Pattern1="RCS file:"
Pattern2="conflicts during merge"

When I try a simple SED command like:

sed -n '/RCS file:/,/conflicts during merge /p' INPUT.txt

I am not getting the expected output. I want to capture the details for "filename2 and filename3" only (which had conflicts during merge).

Can someone please help?

INPUT.txt

RCS file: /hello/filename1
retrieving revision 1.4.2.1.18.2.2.1
retrieving revision 1.4.2.1.18.2.2.1.4.2
Merging differences between 1.4.2.1.18.2.2.1 and 1.4.2.1.18.2.2.1.4.2 into filename1

RCS file: /hello/filename2
retrieving revision 1.4.2.1.18.2.2.1
retrieving revision 1.4.2.1.18.2.2.1.4.2
Merging differences between 1.4.2.1.18.2.2.1 and 1.4.2.1.18.2.2.1.4.2 into filename2
rcsmerge: warning: conflicts during merge

RCS file: /hello/filename3
retrieving revision 1.6.18.1.2.1.2.1
retrieving revision 1.6.18.1.2.1.2.1.4.3
Merging differences between 1.6.18.1.2.1.2.1 and 1.6.18.1.2.1.2.1.4.3 into filename3
rcsmerge: warning: conflicts during merge

RCS file: /hello/filename4
retrieving revision 1.4.2.1.18.2.2.1
retrieving revision 1.4.2.1.18.2.2.1.4.2
Merging differences between 1.4.2.1.18.2.2.1 and 1.4.2.1.18.2.2.1.4.2 into filename4

(Expected) OUTPUT.txt

RCS file: /hello/filename2
retrieving revision 1.4.2.1.18.2.2.1
retrieving revision 1.4.2.1.18.2.2.1.4.2
Merging differences between 1.4.2.1.18.2.2.1 and 1.4.2.1.18.2.2.1.4.2 into filename2
rcsmerge: warning: conflicts during merge

RCS file: /hello/filename3
retrieving revision 1.6.18.1.2.1.2.1
retrieving revision 1.6.18.1.2.1.2.1.4.3
Merging differences between 1.6.18.1.2.1.2.1 and 1.6.18.1.2.1.2.1.4.3 into filename3
rcsmerge: warning: conflicts during merge
fedorqui
  • 275,237
  • 103
  • 548
  • 598
katewadi
  • 11
  • 1
  • 2

4 Answers4

1

Your sed command is starting to output whenever it sees 'RCS file' and stops when it sees the 'conflicts' marker. So it pretty much outputs everything. It's possible to do what you want with sed, but it's complicated. Awk is much simpler:

awk -v RS= '/conflicts/ {print $0}' INPUT.txt

Use awk's notion of records with a blank line separating them, and essentially grep each record. So this doesn't print lines between two patterns, rather it prints each block of lines that matches a particular pattern.

evil otto
  • 10,348
  • 25
  • 38
  • When I try the AWk command, I get the following error...
    awk: Input line RCS file: /was_apps/ cannot be longer than 3,000 bytes.     – katewadi Apr 01 '13 at 23:25
0

Here you have a hint:

awk '{if ($1=="HO") i=1}; {if ($1=="JOU") i=0}; i{print}' file

Example:

$ cat file
HI
HO
JE
JOU
LA

So,

Pattern1="HO"
Pattern2="JOU"
awk -v p1=${Pattern1} -v p2=${Pattern2} '{if ($1==p1) i=1}; {if ($1==p2) i=0}; i{print}' file
HO
JE

Based on your case,

Pattern1="RCS file:"
Pattern2="conflicts during merge"

$ awk -v p1=$(Pattern1) -v p2=${Pattern2} '{if ($1==p1) i=1}; {if ($1==p2) i=0}; i{print}' file

Edit

If you want to look for the string containing this text, you can do it like this:

Pattern1="RCS file:"
Pattern2="conflicts during merge"

$ awk -v p1=$(Pattern1) -v p2=${Pattern2} '{if ($1 ~ p1) i=1}; {if ($1 ~ p2) i=0}; i{print}' file
fedorqui
  • 275,237
  • 103
  • 548
  • 598
  • This does exactly what the submitter asked for, but not what he wanted. It will start printing when it sees the first pattern, but it doesn't know until it gets to the end of a message if it should have printed it. – evil otto Mar 22 '13 at 23:06
  • Yes, now I see @evilotto. I have updated my answer so that it gets into account that the line *contains* this text. – fedorqui Mar 22 '13 at 23:17
0

Bro ,

use egrep ..... its simple , no need to worry on SED and AWK. Its a fixed pattern of lines that you want here ....

 egrep +B4 conflict INPUT.txt

here B option prints the 4 lines before the match along with the line that matches the pattern ( i m using GNU egrep)

egrep output : 

RCS file: /hello/filename2
retrieving revision 1.4.2.1.18.2.2.1
retrieving revision 1.4.2.1.18.2.2.1.4.2
Merging differences between 1.4.2.1.18.2.2.1 and 1.4.2.1.18.2.2.1.4.2 into filename2
rcsmerge: warning: conflicts during merge
--
RCS file: /hello/filename3
retrieving revision 1.6.18.1.2.1.2.1
retrieving revision 1.6.18.1.2.1.2.1.4.3
Merging differences between 1.6.18.1.2.1.2.1 and 1.6.18.1.2.1.2.1.4.3 into filename3
rcsmerge: warning: conflicts during merge
Nitin4873
  • 16,804
  • 1
  • 13
  • 15
0

Here you go:

sed -n '/RCS file:/ !{H;d}; /RCS file:/ {x; /conflict/p}; $ {x; /conflict/p}' input.txt

For every line that doesn't match 'RCS file:' - append it to hold space. If it matches the line (or is the end of file) - swap the pattern space with hold space and check if it matches 'conflit'. Print if it does.

Could be simpler if we use branching.

Like this:

sed -n '$ba; /RCS file:/ba; H; d; :a; x; /conflict/p' input.txt
aragaer
  • 17,238
  • 6
  • 47
  • 49
  • 1
    `sed -n '/^$/!{H;$!d};x;/conflicts during merge$/p'` – jthill Mar 26 '13 at 06:06
  • sed: Function /RCS file:/ !{H;d}; /RCS file:/ {x; /conflict/p}; $ {x; /conflict/p} cannot be parsed. – katewadi Apr 01 '13 at 23:24
  • @jthill
    when I run the SED command, I am getting the followin error...
    – katewadi Apr 01 '13 at 23:31
  • That could be a limitation of hpux version of sed. You can try to write the sed script into a file (replacing all semicolons with newlines) then run it using `-f` key like this: `sed -n -f my_script.sed input.txt` – aragaer Apr 02 '13 at 03:11
  • I found a manpage of HP-UX sed and I don't see the semicolon there. So it's either a multi-line file or multiple `-e` arguments. Like this: `sed -n -e'/^$/!{H' -e'$!d' -e'}' -ex -e'/conflicts during merge$/p' input.txt` – aragaer Apr 02 '13 at 03:14
  • ... btw @katewadi - I haven't seen anybody banging on RCS that hard in production use before. Those rev numbers are impressive beasts. – jthill Apr 02 '13 at 03:24