mjhjmtu

I need to swap these lines, only where the matching strings are in both lines:

before:

REF*CE*-------------------------
REF*1W*-------------------------

after:

REF*1W*-------------------------
REF*CE*-------------------------

I tried this, it didn't work:

ed -s testfile.txt <<<$'/REF*CE*/-0,/REF*CE*/+0m/REF*1W*/nwnq'

sed -e :a -e '$!N;s/^(REF*CE.*)n(REF*1W.*)/2n1/;ta' -e 'P;D' <testfile.txt 

1. If we're not on the last line then append next line.


2. Do a substitution on the current line that only occurs if it matches substring containing pattern 1 + newline + substring containing pattern 2. The substitution flips the two sub-strings. After substitution go back to label :a.


3. If there was no match Print the pattern space as is. Then Delete the pattern space and start the cycle again.

Sample with some surrounding lines...

In:

 XEF*CE*------------------------- 
 REF*CE*------------------------- 
 REF*1W*------------------------- 
 REF*2W*------------------------- 

Out:


 XEF*CE*------------------------- 
 REF*1W*------------------------- 
 REF*CE*------------------------- 
 REF*2W*------------------------- 

More generally for any pattern1 and pattern2

sed -e :a 
 -e "$!N; s/^(.*$pattern1.*)n(.*$pattern2.*)/2n1/;ta" 
 -e 'P;D' < inputfile

We can get the desired effect by putting the matched line into hold buffer, reading next line and printing it, then swapping the hold buffer with pattern buffer and printing again.

bash-4.3$ sed -n '/^REF*CE/!p;/^REF*CE/h;n;p;x;p' input.txt
some line here
REF*BB*106497026---------------
REF*1W*723266637---------------
REF*CE*NEW JERSEY--------------
SVC*HC^S5102*78.5*78.5**1------
another line there

A general solution for how to swap two (possibly distant) lines that match particular regular expressions with ed:

1. 
   Copy one line to after the second line.


2. 
   Move the second line to after the original first line.


3. 
   Delete the first original line.

Or, with ed editing commands:

1. /pat1/t/pat2/


2. ?pat2?m/pat1/


3. ?pat1?d

Example with the file

CLP*815900102*2*489.8*101.5*82.29*13*PVJLS03YP0000*13*7
AMT*AU*489.8
REF*6R*00000000002
DTM*472*20160528
CAS*OA*23*306.01
CAS*PR*2*82.29
SVC*HC:99212:25*489.8*101.5**1
AMT*B6*411.43

in which we'd like to swap the first AMT line with the second CAS line. pat1 will be ^AMT*AU and pat2 will be ^CAS*PR. Note that we need to escape the * for it to be treated literally in the regular expression.

I have annotated the changes below so that it's easier to see them. The XXX indicates the current position in the file after each operation.

1. 
   

   /^AMT*AU/t/^CAS*PR/ produces

   
   
   

   CLP*815900102*2*489.8*101.5*82.29*13*PVJLS03YP0000*13*7
   AMT*AU*489.8 <-- Line copied *from* here
   REF*6R*00000000002
   DTM*472*20160528
   CAS*OA*23*306.01
   CAS*PR*2*82.29
   AMT*AU*489.8 <-- Line copied *to* here (XXX)
   SVC*HC:99212:25*489.8*101.5**1
   AMT*B6*411.43
   

   
   


2. 
   

   ?^CAS*PR?m/^AMT*AU/ produces

   
   
   

   CLP*815900102*2*489.8*101.5*82.29*13*PVJLS03YP0000*13*7
   AMT*AU*489.8
   CAS*PR*2*82.29 <-- line moved here (XXX)
   REF*6R*00000000002
   DTM*472*20160528
   CAS*OA*23*306.01
   AMT*AU*489.8 <-- line previous to this deleted
   SVC*HC:99212:25*489.8*101.5**1
   AMT*B6*411.43
   

   
   


3. 
   

   ?^AMT*AU?d produces

   
   
   

   CLP*815900102*2*489.8*101.5*82.29*13*PVJLS03YP0000*13*7
   CAS*PR*2*82.29 <-- the line before this was removed (XXX)
   REF*6R*00000000002
   DTM*472*20160528
   CAS*OA*23*306.01
   AMT*AU*489.8
   SVC*HC:99212:25*489.8*101.5**1
   AMT*B6*411.43
   

   
   

As an easy-to-remember "one-liner":

pat1='^AMT*AU'; pat2='^CAS*PR'; printf '/%s/t/%s/n?%s?m/%s/n?%s?dnwqn' "$pat1" "$pat2" "$pat2" "$pat1" "$pat1" | ed -s file

Note that this is reversible by running the exact thing a second time, i.e. it doesn't matter which pattern is the first or second.

I have simple solution like this:
Say, If you want swap a line that has "XXXX" with line that has "YYYY" in afile, providing only one line has "XXXX" and only one has "YYYY"

Example file is like this:
*ssss dddd
aaaa ffff
ddd rrrr
ddddd XXXX
ddddde
ffff
ffff
fff
eeee YYYY
ghghgh
hhhhh

My 'sed' command is sed 's/XXXX/ZZZZ/g;s/YYYY/XXXX/g;s/ZZZZ/YYYY/g'
Remember "ZZZZ" string should not be in the file anyware.