Make chomp return the chomped string instead of chomping in place

[DarwinAwardWinner]

The behavior of chomp has always been kind of weird, especially when it forces you to write:

chomp(my $line = <INPUT>);

instead of:

my $line = chomp <INPUT>;

Shouldn't chomp work more like trim?

[avar]

FWIW the reason this is done is that chomp() is much faster in void context. Instead of creating a new string it just modifies its length (it doesn't even reallocate).

And further the reason chomp() still returns the value that it does is historical. Some code out there depends on it.

The latter could be munged by perl5i. But it's a core operator so overriding it means overidding pp_chomp.

Also note that you could call chomp like this:

chomp(@strings);

[schwern]

chomp() can be overridden by defining CORE::GLOBAL::chomp. It can be made to behave lexically the perl5i way or the normal way. The trouble is, the wrapper will come with a performance penalty that effects all uses of chomp().

Fortunately, autoboxing lets us get around that. We can do whatever we like with the autoboxed version. while( $line = <$fh>->chomp ) { .. } would be interesting.

Once again I wish for the Ruby style method! naming convention to indicate methods which act in place.

[DarwinAwardWinner]

See also: http://dev.perl.org/perl6/rfc/58.html

[schwern]

There are two thing which trouble me about that Perl 6 RFC. The proposed special case for chomp() in while() is, well, a special case. Second, that chomp() would work differently in void and non-void contexts. Similar problems came up on p5p when discussing trim().

The dichotomy between the efficiency of in-place manipulation and the convenience of returning the modified structure plagues Perl ops.

[DarwinAwardWinner]

Aren't COW strings supposed to solve this in Perl 6? Though obviously this doesn't help for Perl 5.

[schwern]

Since chomp() is altering the string before returning, I don't see how it would. Unless it does something Extremely Clever like sharing the underlying C string between the two scalars, each using a different length. But that's not generally applicable.

[avar]

It could. That's how the core datastructures of Clojure and similar functional "strings are immutable" languages work, you can create lots of copies of 10MB strings and only store the differences.

There's COW support in p5, but I don't know its status.

[daxim]

In favour of keeping chop/chomp as is. Add functions chopped/chomped from http://p3rl.org/Text::Chomped instead which DTRT.

[schwern]

It just occurred to me that we already have trim which does what we want, it just needs to be given an argument.

my $line = <$fh>->trim("\n")

Is it worthwhile to write a special method just for trim("\n")? Or is it sufficient to just mention the above in the FAQ?

If so, what would it be called that would be more convenient to type and as obvious what it's doing? chomped fails the second case as it says nothing about how it's different from chomp.

[doherty]

Yes, it is worthwhile.

use perl5i::2;
use Test::More tests => 2;

my $string = " hello \n";
is($string->trim,  'hello', 'trim DTRT');

$string = " goodbye \n";
is($string->chomp, ' goodbye ', 'chomp, wtf'); # I expect this to work too