Perl \R regex strip Windows newline character

Question

I'm using a Perl script using the following code to remove possible Windows newline characters in input file:

foreach my $line(split /\\r|\\R/)

Executing the same script on two different Linux machines has different results. On machine1 script works as intended, on machine2 every time a capital "R" character is found line is split and result is messed.

I would like to know if the \\R regex is correct and how to make machine2 behave as intended.

Answer 1

In Perl, there are several differences in the way carriage returns can be handled:

\n matches a line-feed (newline) character (ASCII 10)
\r matches a carriage return (ASCII 13)
\R matches any Unicode newline sequence; can be modified using verbs

Windows uses the two characters ASCII 13 + ASCII 10 ( \\r\\n ) and unix uses ASCII 10 ( \\n ). The \\R expression matches any Unicode newline sequence ( \\r , \\n , \\r\\n ).

The likely reason \\R works on one machine and not the other might be differing versions of Perl . \\R was introduced in perl 5.10.0 , so if the other machine is using an older version then updating should solve your issue.

More info :

• Perl Regular Expression Backslash Sequences and Escapes
• Syntax of Regular Expressions in Perl

Answer 2

One of your machines must be using a rather ancient version of Perl.

5.8:

$ perl -wle'print for split /\R/, "QRS\r\nTUV\r\n";'
Unrecognized escape \R passed through at -e line 1.
Q
S
TUV

5.10:

$ perl -wle'print for split /\R/, "QRS\r\nTUV\r\n";'
QRS
TUV

Always use use strict; use warnings; use strict; use warnings; !

Alternatives:

• split /[\\r\\n]/ . This is equivalent to what you are using, but it's probably buggy.
• split /\\n|\\r\\n?/ . This is equivalent to split /\\R/ .
• split /\\r?\\n/ . This matches unix and Windows line endings.
• split /\\r\\n/ . This matches Windows line endings.

I'd use the second last.

Answer 3

I use Perl nearly every day.

However, if all I have to do is convert line endings, then I use

• dos2unix to convert to Linux/Unix line endings
• unix2dos to convert to Windows line endings.