Shell Programming and Scripting

I am having a dificult time trying to replace 2 carriage returns and 1 newline with 1 carriage return and 1 newline in a file. The file may come as CR NL in which case I need to leave it as is. When it comes with CR CR NL then I need to remove one of the CR's.

i tried

sed 's/\r\r\n/\r\n\g' file

but it has no effect on the file

appreciate the help. (I did search and tried many of the examples)

The syntax you posted is incorrect and would only cause an error message since the pattern is missing the last / .

Second, sed never matches \n, since it considers it to mark the end of the line. sed 's/\r\r$/\r/' should do the job.

I tried this and it does not have any effect on the file

sed 's/\r\r$/\r/' filename > o.txt

I opened my filename in vi and I can see one ^M. when I open in ultraedit I see 0D 0D 0A at the end of each line. I want the file to be in DOS format but as I metnioned before I need to insure that if it is already in DOS format that the sed command does not change it in any way.

I think I got it - please confirm my understanding. When I open the file in vi I only see 1 ^M. In ultraedit I see what looks like 2 carriage returns 0D 0D.

tr -d '\015' < filename > o.txt

the above removed the extra carriage return. Then I tried it on the output file and it left the file unchanged.

I tested my solution as follows before posting it:

Code:

$ echo -en '\r\r\n' | hexdump -C
00000000  0d 0d 0a                                          |...|
00000003
$ echo -en '\r\r\n' | sed 's/\r\r$/\r/' | hexdump -C
00000000  0d 0a                                             |..|
00000002
$ $ echo -en '\r\r\n\r\r\n' | sed 's/\r\r$/\r/' | hexdump -C
00000000  0d 0a 0d 0a                                       |....|
00000004

I am using GNU sed, what version are you using? It may be available on your system as gsed.

Whatever sed you're using may behave better if you remove the $. The $ prevents carriage-returns from being removed in the middle of a line, so if you don't need to worry about that, ditch the $. Also, 's/\r\r/\r/g' might also be worth a try.

That trick with tr will remove all carriage returns, every last one, so it may be that your editor is displaying spurious carriage returns for whatever reason.

Lastly, it is possible you're using an odd OS that considers files to have text mode or binary mode.. This may cause different programs to display the file in different ways depending on the mode they open it in. Are you using cygwin?

I am on SUNos so I use sed

The file is created by an external source. The external source is suppoed to create the file in DOS format but for some reason sometimes they are introduce an extra 0D.

I pick up the file from our server and move it to a folder foR my testing using binary mode SSH client. I open the file in a text editor and I can see 0D 0D 0A. I open it in vi and I can only see one ^M. I though ^M and 0D are the same thing - no?

When I run the tr against a DOS file with 0D 0A it does not remove the 0D -which is what I want.

I am not using cygwin.

One thing - when i view the file in the text editor and go to the end of the line the cursor is on the second 0D. So like this

0D 0D 0A (cursor appears on the bolded 0D)