" You use both "platform specific line-ending" and "network line-ending." Is the use of two different phrases accidental or purposeful?"

Very purposeful.

The sub will receive strings for transmission across the network. The caller may have terminated the string with "\n" (which will be the platform-sepecific line-ending dependant upon what the current platform is), or they may not have done so.

Before transmission, I need to:

1. Add a network-specific line-ending (\015\012) if no line-ending is present.
2. Convert whatever (platform-specific) line-ending is present, to the network line-ending is a line-ending, if one is present.
3. Leave the string alone, if the correct line-ending is already present.

It seems like a single regex that used look-behind assertions properly would be able to do this, thereby avoiding messy conditional logic. It also seems like this is an oft-called for requirement and a well tried solution is probably already known.

I've had a couple of attempts at constructing the regex, but in each case, it falls over in one of the above three cases. Either adding an extra, unnecessary line-ending or omitting to add one.

I hoped someone in the know would point me at the correct regex to use?

I would do that this way:

s/\015\012|\012\015|\012|\015/\015\012/g;
[download]

Update: I got that order wrong, as BrowserUk suggests. I had the shorter versions on the left, and the longer on the right. The above is fixed.

It should help you with things like this to take a good, hard look at the perlre section of the standard Perl docs. You should have a copy of the standard Perl docs on your system, but if not, perlre is on search.cpan.org too.



Christopher E. Stith

Update: Deleted this and put it as a response to the main question. Do I need to do something to get this reaped or can I just delete the text and the title?

I'm not the OP, but I took the two phrases to mean very different things.

"platform specific line-ending" could be "\012" ("\l") for Unix, "\015\012" ("\r\l") or "\012\015" ("\l\r") for DOS, or "\015" ("\r") for at least older Macs. All of these, on their own platforms, are represented by the C library as "\n" (newline).

"Network line-ending" means "\015\012", which is the standard for things like Telnet and HTTP.

Any telnet or web implementation on any platform uses "\015\012" (carriage return followed by linefeed) as the end-of-line marker.

See RFC 1945 for HTTP and RFC 854 (STD 0008) for Telnet for examples.



Christopher E. Stith

precisely my point.

That's great! Now, what have you already tried?

Personally, I would attack the problem through the following two steps:

1. Remove any line ending, whatever it is. perldoc -f chomp and perldoc perlvar for $/ might be interesting reading for you.
2. Add \015\012 to the string.

Something like this?

#! perl -slw
use strict;

sub fixLE {
    my $string = shift;
    $string =~ s[(?:\n?\r?\n?)$][\015\012];
    return $string;
    
}

print join('|', unpack 'C*', substr $_         , -8 ), $/,
      join('|', unpack 'C*', substr fixLE( $_ ), -8 ), $/ for 
    'No line ending needs one',
    "unix style needs fixing\x0a",
    "mac style needs fixing\x0d",
    "windows style needs fixing\x0d\x0a",
    "network style: leave alone\015\012";

__END__
P:\test>junk
101|101|100|115|32|111|110|101
100|115|32|111|110|101|13|10

32|102|105|120|105|110|103|10
102|105|120|105|110|103|13|10

32|102|105|120|105|110|103|13
102|105|120|105|110|103|13|10

102|105|120|105|110|103|13|10
102|105|120|105|110|103|13|10

32|97|108|111|110|101|13|10
32|97|108|111|110|101|13|10
[download]

Ideally all in a script you post in "code" tags. Then it would be easier for someone on PM to write one regex that does everything you want to do.

Celsius to Fahrenheit using Regexp::Common has some simple code that would get you started reformulating your question in a more helpful way.