sed Suggestion

[Jason Van Cleve]

Hi,

I've been trying to contact someone for weeks regarding an issue ran
into with sed.  I was trying to use sed to strip whitespace from the
ends of lines in my source code tree, using the regex "$" symbol.  Some
of those files have DOS newlines, and sed was quietly failing those.  I
learned the hard way what was going wrong, and eventually I employed the
dos2unix tool to work around it.

The trouble is, the hard way was the only way for me, since I could not
divine that "$" in unix regex'es means unix newlines only, not DOS or
Mac newlines.  I have found no documentation addressing this obscure
little gotcha, other than a passing mention of the DOS version of sed
supporting DOS newlines, etc., and I understand the historical reasons
for this.  But the advance of networking has made DOS-formatted files
more ubiquitous on unix-like systems, and I feel the developers of sed
really ought to address the problem in some way.

My preference would be the addition of a command-line option that will
enable "$" to match all three newline sequences at once.  This would be
most expedient, since I wouldn't need an intermediate dos2unix
invocation, and it would preserve backward compatibility.

Barring that, I suggest you add just a small paragraph in the man page
pointing out this pitfall, which would have saved me some hours of
research and discussion (not to mention some face).  I sent a draft of
this text to someone in a email now submerged in my archives.  If you
like, I'll dig that up and pass it on, but I'm not even sure I'm talking
to the right people yet.  I do hope someone is still maintaining sed,
considering entire books on it are still in print.

Thanks for your attention,

Jason Van Cleve