From: Tim Hammerquist <tim@...>
Date: 2002-05-20T08:48:44+09:00
Subject: Re: Ruby regex question

Dossy graced us by uttering:
> On 2002.05.19, Tim Hammerquist <tim@vegeta.ath.cx> wrote:
>> Basically, if the match fails, you don't have to check any values.
> 
> Yikes.
> 
>   def foo(something)
>     something =~ /.../
>     or_other = $1
>     bar(or_other)
>   end
> 
>   def bar(or_other)
>     ...
>   end
> 
> Are you saying that if you test the success or failure of the match in
> foo(), that bar() doesn't need to test its inputs?

No, I'm saying that foo() should check _its_ inputs, and bar() should
check its inputs. Depending on foo() to check bar()'s input would be a
violation of encapsulation, ne?

And no. To clarify, I'm saying that if you check if the entire regex
/(foo) (bar) (baz)/ matches against "foo bar bas", you don't have to
check $1, $2, or $3.

>> Most Perl code from gurus checks the match success immediately.
>> 
>> Most Ruby idioms do the same.
> 
> Really?

In my experience, yes.

> -r--r--r--  1 root  root  217854 Oct  5  2001 /usr/lib/perl5/5.6.1/CPAN.pm
> 
>    3086     # read header
>    3087     my($line_count,$last_updated);
>    3088     while (@lines) {
>    3089         my $shift = shift(@lines);
>    3090         last if $shift =~ /^\s*$/;
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^
If $shift is all whitespace or empty, the entire loop is skipped.
Success of the match is check directly _before_ acting upon it.

>    3091         $shift =~ /^Line-Count:\s+(\d+)/ and $line_count = $1;
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The binary operation 'and' checks the boolean status of the previous
expression, in this case, a regex match.  This is a variation of:

    if ($shift =~ /match/) { $line_count = $1; }

Success of the match is checked before use of $1.

>    3092         $shift =~ /^Last-Updated:\s+(.+)/ and $last_updated = $1;
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Same here.

Success of the match is checked before use of $1.

> Maybe this portion of CPAN.pm wasn't written by a Perl guru.

Andreas Koenig is a very capable Perl hacker, and a frequent poster to
clpm. He also checks match success immediately in this excerpt.  I fail
to understand why you include it.

> You are right though, that it is more common to find regex success
> immediately with the regex evaluation.  Perhaps it's because with
> Perl, you can't know if $1 is valid (applies to the regex you just
> evaluated) _unless_ you test if the regex succeeded, which _forces_
> you to test for the regex success.  Ruby doesn't force you to do
> this, so the only reason for it to be a Ruby idiom is because it's a
> carry-over from Perl.

(There are quite a few features of Ruby that were deliberately borrowed
from Perl, but...)

I only hope you don't place the success of the entire regex in the
contents of $3 in the following match.

    "foo bar " =~ /(foo) (bar) (baz)?/

The result of the match is true (0, to be exact, but boolean true), but
$3 is nil. (tested; 1.6.7)

> In Ruby, which is more OO than Perl, I'd imagine that the idiom
> would involve more decoupling between objects.  The object receiving
> the message should be responsible for determing the validity of its
> inputs, not necessarily the sender.

Granted. But _how_ you check the input has little to do with _where_
you check its input.

> -- Dossy

Tim Hammerquist
-- 
"Sometimes these hairstyles are exaggerated beyond the laws of physics."
    -- Unknown narrator on Anime