in vi, if i type ctrl-J, i can shift the next line up to the current line.
I need to do this whenever a specific string exists (in particular ^M) ,however i have not been able to find a way to do this in vi, sed, or awk

seems simple enough.

can someone help me?

thanks in advance

You must mean shift-J, not control J.

In vi you can do this:
:g/^M/j

But to type the ^M, you need to type cntl-V followed by cntl-M.

thanks, but it does not join all of the ^Ms

240|^M\
^M\
^M\
Old Port Marketing order recd $62,664.- to ship 6/22/99^M\
^M\
^M\
^M\
^MCompleted by: C. Malczynski|


becomes:

240|^M\ ^M\
^M\ Old Port Marketing order recd $62,664.- to ship 6/22/99^M\
^M\ ^M\
^M\ ^MCompleted by: C. Malczynski|

but should be:

240|^M\ ^M\^M\ Old Port Marketing order recd $62,664.- to ship 6/22/99^M\^M\ ^M\^M\ ^MCompleted by: C. Malczynski|


this is what i typed :g/^M\\/j ----- (ctrl-V, ctrl-M)


any suggestions??

p.s. thanks, this has been stumping me for awhile

:g/^M\\$/j
and run the command a 2nd or third time until no more trailing ^M\ remain.

thanks, this has been really helpful, but i need to run this on several files that have >1000000 records, which i can break down into smaller files if need be (migrating data in case you haven't guessed), hence the reason i was trying to find a one time fell swoop, the solution you provided is much better than me scrolling for ^M and hitting Shft-J five or six times, but also doesn't lend itself to being run unattended, do you know of a way to peform the same task with sed?

if not, the help you have given me has been great, and i appreciate your quick responses.

thanks again

This is a sed script. It does not use a shell. You must get all 11 lines exactly as I have them. Call it combiner or something. Run it like:
combiner < oldfile > newfile

The paste command can be used for serial merging, e.g...

sweet! looks like voodoo to me, if you have the patience to explain what that script did, i would love to learn more about it.

thank you very, much!!!!

I'll try, but it's not real easy to explain...

s/^M\\$/^M\\/ changes cntl-m backslash to cntl-m backslash. The $ makes this happen only if the pattern is at end of line. This results in no change, but it set a success flag that can be tested later. That is how we know if we want to join the next line.

H appends the pattern area to the hold area. The pattern area is where lines arrive as they are read.

t test to see if that previous s commnand worked or not. If if did, we jump to the end of the script. Well not the last line. This means that we are finished with the current line. So we will read a new line and restart the script.

x exchange. What I really want to do is to retrieve the hold area. Since we are here, the test failed. So we have a line that did not end with cntl-m backslash.

s/\n//g That hold area that I just retrieved is a collection of lines that I want to join. So I delete the newlines.

p print the line. Since there was a -n on the sed command line, no lines are printed automatically.

s/.*// Remove all characters. This empties the pattern area. Thre is a d command, but it has side effects. I usually go with this as my delete.

h copies pattern area to hold area. Now it's empty too.

i think i get the gist, i was able to use this script to clean up a few other anomalies as well.

thank you very much, i was to the point of asking the users if they really needed those records... which, invariably, they answer yes... you have saved me much headache.

thanks again.