You'll need to use an if statement:

my $split_pattern //= getOpt();

my @x;
unless( defined $split_pattern ) {
    @x = split ' ', $line;
}
else
    @x = split $split_pattern, $line;
}
[download]

Not elegant, but the only way. The special behaviour of split ' ', is triggered by the use of exactly ' ' in the source code.

Substituting a variable set to a space, or even a conditional statement where one branch is ' ', will not do the same thing.

UPDATE: Pls ignore this post. Thanks to LanX above I now understand what the problem is.

Can you explain what the supposedly special behavior of split ' ' is? I cannot find any documentation saying anything about it. For me this works:

use strict;
use warnings;
use Data::Dumper;

my $str = "I want a split pattern";
my $pattern = $ARGV[0] // ' ';

my @pieces = split /$pattern/, $str;
print Dumper \@pieces;
[download]

hdb$ perl split4.pl
$VAR1 = [
          'I',
          'want',
          'a',
          'split',
          'pattern'
        ];
hdb$ perl split4.pl a
$VAR1 = [
          'I w',
          'nt ',
          ' split p',
          'ttern'
        ];
[download]

You may also want to use quotemeta. e.g.

ski@anito:~$ perl -e 'use strict; use warnings; my $x = "a.*b.*c.*d"; my $pattern = ".*"; my @x = split /$pattern/, $x; print join " - ",@x'
ski@anito:~$

vs.

ski@anito:~$ perl -le 'use strict; use warnings; my $x = "a.*b.*c.*d"; my $pattern = quotemeta(".*"); my @x = split /$pattern/, $x; print join " - ",@x'
a - b - c - d
ski@anito:~$

perldoc -f quotemeta.

Search for "specifying a PATTERN of space" in perldoc -f split. The difference relates to leading empty strings.

If you want repeated whitespaces to be ignored, then simply tell split to do so.

  DB<113> split ' ', 'abc  def'
 => ("abc", "def")

  DB<114> $del=qr/ /
 => qr/ /

  DB<115> split $del, 'abc  def'
 => ("abc", "", "def")

  DB<116> $del=qr/\s+/
 => qr/\s+/

  DB<117> split $del, 'abc  def'
 => ("abc", "def")
[download]

update

split :

               As a special case, specifying a PATTERN of space (' ') 
+will
               split on white space just as "split" with no arguments 
+does.
               Thus, "split(' ')" can be used to emulate awk’s default
               behavior, whereas "split(/ /)" will give you as many nu
+ll
               initial fields as there are leading spaces.  A "split" 
+on
               "/\s+/" is like a "split(' ')" except that any leading
               whitespace produces a null first field.  A "split" with
+ no
               arguments really does a "split(' ', $_)" internally.
[download]

I'd like the split ' ' behavior, unless changed from the command line option. Something like this:

my $split_pattern = ' '; # default
...
# $split_pattern might get changed here from a command line option
$split_pattern = $foo if $bar;
...
$line = "   one two three   \n";
my @words = split $split_pattern, $line;
for my $i (0..$#words) {
    print "$i: ($words[$i])\n";
}

# Expected output for default case
0: one
1: two
2: three

# Actual output for default case
0: ()
1: ()
2: ()
3: (one)
4: (two)
5: (three)
6: ()
7: ()
8: (
)
[download]

-QM
--
Quantum Mechanics: The dreams stuff is made of

Now, I understand what you are looking for, it wasn't clear to me so far. If you store ' ' into a variable, you don't get the expected behavior as if you hard code:

my @fields = split ' ', $line;

Using "qr/\s+/;" as a split pattern improves the result but still does not yields what you want (you still get an extra empty element at the beginning of the array).

Well, then I don't see any real simple direct solution. Even something like this:

my @fields = split $pattern? $pattern : ' ', $line;

does not do what you are looking for. The only solutions I can think of are either to use the above "qr/\s+/;" and shift the first element if empty, or use a if else construct, or yet this equivalent construct:

my @fields  = defined $pattern? split $pattern, $line : split ' ', $line;

Many thanks. I had not seen that bit.

Update: I have just given a way to define the default pattern that you asked for for use if no other pattern has been defined. But you might consider using something else than ' ', because of its special behavior. For example:

See Re^2: How to get split $var to work like split ' '?.

-QM
--
Quantum Mechanics: The dreams stuff is made of

There's always the eval solution.

my $sep = ' ';
$, = ', ';       
$\ = $/;
my $string = "          Hello World";

print split ' ', $string;
print split $sep, $string; # This does not work
print eval qq< split '$sep', \$string >;
[download]

Hello, World
, , , , , , , , , , Hello, World
Hello, World
[download]

Yes, that occurred to me later.

Still, to follow the Principle of Least Astonishment, I'd like to be able to specify a regex. To avoid further if then's, I'd probably set the default like so:

my $sep = q/' '/;
$sep = shift @ARGV if @ARGV; # you get the idea
my $line = "   one two three   \n";
my @words = eval qq(split $sep, \$line);
[download]

Then something evaluating to a regex must be used, such as:

$sep = q(/\s+/);
$sep = q(//);
$sep = q(/two|\s+/);
$sep = q(m|/|);
[download]

But the delimiters must be part of the string.

(I'm ignoring for now the security issue of letting the user specify a regex, as this is a script just for me. And I'm not up to untainting the full regex language, so I can't see letting this into the wild hands of colleagues.)

-QM
--
Quantum Mechanics: The dreams stuff is made of

See my reply at Re^2: How to get split $var to work like split ' '?.

-QM
--
Quantum Mechanics: The dreams stuff is made of