2

I have a huge cvs log file which, cleaned from the useless info, reads something like 

Working file: unmodifiedfile1.c
================
Working file: modifiedfile1.h
----------------------------------
revision 1.3
Fixed some bug
================
Working file: unmodifiedfile2.h
================
Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature
================
Working file: unmodifiedfile3.h

I would like to clean the lines related to unmodified files:

Working file: modifiedfile1.h
----------------------------------
revision 1.3
Fixed some bug
================
Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature
================

The pattern to match is 

Working file: FILENAME
================

What i've been able to do up to now is the following:

sed '/Working file:/ N ; s/\n/PLACEHOLDER/' changelog.txt |
grep -v 'PLACEHOLDER===' |
sed 's/PLACEHOLDER/\n/

I'm sure however there is a cleaner solution that my sed ignorance precludes me... ( also, a bonus would be being able to erase the very latest line if necessary)

P.S.

An output ending with:

================
Working file: unmodifiedfile3.h

is also acceptable

| improve this question | |
  • can you add the expected output for clarity? add a separate one if it will be different for the bonus question – Sundeep Sep 8 '16 at 13:34
1
sed '/Working file:/ N ; s/\n/PLACEHOLDER/' changelog.txt |
grep -v 'PLACEHOLDER===' |
sed 's/PLACEHOLDER/\n/

can indeed be shortened to:

$ sed '/Working file:/{N;/===/d}' changelog.txt 
Working file: modifiedfile1.h
----------------------------------
revision 1.3
Fixed some bug
================
Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature
================
Working file: unmodifiedfile3.h


  • To remove all lines containing Working file: and following line if it contains === as well as final line if it contains Working file:

Thanks @ilkkachu for the suggestion. If the pattern needs to be matched at beginning of line, use ^Working file:

$ cat ip.txt 
Working file: 123
================
Working file: f1
----------------------------------
revision 1.3
Fixed some bug
================
Working file: abc
================
Working file: file
----------------------------------
revision 1.1
Added some feature
================
Working file: xyz

$ sed '/Working file:/{N;/===/d}' ip.txt | sed '${/Working file:/d}' 
Working file: f1
----------------------------------
revision 1.3
Fixed some bug
================
Working file: file
----------------------------------
revision 1.1
Added some feature
================
| improve this answer | |
  • Your initial sed solution was exactly what I was after. However the tweaks using 'unmodified' are wrong, the filenames are not made like that, I wasn't probably clear enough. – Davide Sep 12 '16 at 10:56
  • @Davide, I would then suggested to undo this answer as accepted, and edit your question with a better example... that way it helps others trying to solve a similar problem – Sundeep Sep 12 '16 at 11:00
  • 1
    I think you could use ${/^Working file/d}; in the beginning of the sed to remove the final line if it starts with Working file. I'd put the beginning-of-line anchor to the other /Working file/ pattern, too, though the sed would still match that if the final line of an actual commit message started with Working file... – ilkkachu Sep 12 '16 at 11:12
3

sed

This should come close to what you are after:

<cvslog sed -n '/Working file/ { N; /\n=\+$/b; :a; N; /\n=\+$/!ba; p; }'

Output:

Working file: modifiedfile1.h
----------------------------------
revision 1.3
Fixed some bug
================
Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature
================

Explanation

Here is the same sed script with comments:

/Working file/ {
  N                 # append next line to pattern space
  /\n=\+$/b         # is it a file separator -> next file
  :a
  N                 # append next line to pattern space
  /\n=\+$/!ba       # isn't it a file separator -> read next line
  p                 # otherwise print accumulated text
}

awk

If you tell awk to use the file separator line as the record separator (RS), it becomes fairly straightforward to define a sensible selection criteria:

<cvslog awk 'NF>2' RS='\n=+\n' FS='\n' ORS='\n\n'

Output:

Working file: modifiedfile1.h  
----------------------------------
revision 1.3
Fixed some bug

Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature

bash and coreutils

Just for fun:

csplit cvslog '/=\{16\}/1' '{*}'
wc -l xx* | 
head -n-1 | 
while read n f; do 
  if (( n > 2 )); then 
    cat $f
  fi
done

Output:

Working file: modifiedfile1.h
----------------------------------
revision 1.3
Fixed some bug
================
Working file: modifiedfile2.h
----------------------------------
revision 1.1
Added some feature
================
| improve this answer | |
  • Haha I guess you like text processing. – Just a learner Sep 9 '16 at 4:40
  • @OgrishMan: who doesn't? :-) – Thor Sep 9 '16 at 12:11
  • Very good answer, but for the time being I accepted the other one as to me the sed script is easier to understand – Davide Sep 12 '16 at 10:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.