Clean line ends

[Scarborough]

I have a problem with text files being passed between systems. I have a text file from a COBOL program which is t ransfered to a legacy datastore. The COBOL guy is not willing to look at it but the line endings are shot meaning the datastore rejects it. Looking in an editor which displays line end charecters I have found this

SOME DATA\cr\lf
\lf
MORE DATA\ccr\lf
\lf
Yet more data\cr\lf

I need to replace the \lf characters with \cr\lf on the empty lines. Any ideas?
Thanks

[davorg]

Look for lines that start with a line feed (ctrl-J) and convert it to a carriage return (ctrl-M) followed by a line feed.

The following is a command line script to do it. It edits your file in place and renames the original version with a .bak extension.

perl -i.bak -pe 's/^\cJ/\cM\cJ/' your_file_name_here

[shmem]

Replace all lfs not preceeded by cr with cr+lf:

perl -pi.bak -e 's/(?<!\r)\n/\r\n/' cobolfile

See perlre.

--shmem

update: as davorg pointed out, it is better to use control chars for newline conversions. The look-behind assertion above replaces bogus newlines elsewhere, not only at blank lines.

[davorg]

I'm sure that works.

But I find I sometimes get confused about how Perl inteprets \n (sometimes it's \012 and sometimes it's \015\012 - see perlport for the details). In order to minimise my confusion I find it helpful to always refer to the control characters (\cJ and \cM) or the numeric character codes (\012 and \015) when I'm converting newline characters.

[smithers]

I also found David Cross' explanation of differing line end chars very helpful. See his book "Data Munging w/Perl", page 88-89. The topic is very well explained and it was an eye-opener to me when I was confused about new lines for different platforms. Sorry to point you to a book but I think David's book is excellent and his coverage of this particular topic is helpful.