sed: conditional merge of multiple lines

问题

I need to merge line of a file using sed based on a pattern. eg:

Input File:

X1 A B C D E F

\+ G H I J 1 

\+ LK T PP E OO 2

X2 DDF F Y 

\+ J W Q 

....

OutPut Expected:

X1 A B C D E F G H I J 1 LK T PP E OO 2

X2 DDF F Y J W Q 

..

---

I would like to equivalent of wat is possible in vi editor (:%s/\n+/ /g)

Searching the web I found a solution, which logically should have worked

sed -e '{:a; N; s/\n+/ /g; ta}' infile

But this command defies my understanding and logic, and has produced output

X1 A B C D E F

\+ G H I J 1 LK T PP E OO 2
X2 DDF F Y 

\+ J W Q 

....

Any ideas are welcome, & Thanks in advance

Srisurya

回答1:

This might work for you:

sed ':a;$!N;s/\n+//;ta;P;D' file

explanation:

• :a is a loop placeholder
• $!N means if not end-of-file append next line to current line.
• s/\n+// means delete an end on line followed by a plus sign
• ta means if last substitution worked branch to a loop placeholder
• P print upto the first newline.
• D delete upto and including the first newline.

回答2:

an alternative awk oneliner:

awk 'BEGIN{RS="  "}{gsub(/\n\n\+/,"")}1' yourFile

回答3:

potong's answer didn't work for me, but something similar works:

sed -e :a -e '$!N;s/\n+//;ta' -e 'P;D' file

Good sed docs here: http://sed.sourceforge.net/sedfaq3.html

来源：https://stackoverflow.com/questions/11290616/sed-conditional-merge-of-multiple-lines