in reply to Re^2: "when" and replacements
in thread "when" and replacements

There's currently a thread on p5p where some people argue that the non-aliasing behaviour of given is a bug. I kind of like the idea that since one can use when clauses in a for, given doesn't alias. If you want aliases, write for; otherwise, write given. OTOH, I don't think I've ever written code where I wanted given to alias, nor that I wanted it to not-alias. That is, each time I use given, whether $_ aliases or not, it does not matter.

Replies are listed 'Best First'.
Re^4: "when" and replacements
by LanX (Saint) on May 07, 2011 at 11:25 UTC
    >I kind of like the idea that since one can use when clauses in a for, given doesn't alias. If you want aliases, write for; otherwise, write given.

    Yeah sounds like perfect orthogonal design. Though one has to keep in mind that leaving the block demands different statements. 

    use feature qw(switch);
 
given (1) {
  print;
  break;
  print
}
 
for (1) {
  print;
  last;
  print
}

    [download]

    Cheers Rolf

    UPDATE:

    confusingly this works well:

    $\="\n";
 
for ("012") {
  if   (/0/) {print "0 matched"};   # prints 0
  when (/1/) {print "1 matched"};   # prints 1
  when (/2/) {print "2 matched"};   # nix
  if   (/3/) {print "3 matched"};   # nada
}

    [download]

    but the perlsyn says Every "when" block is implicitly ended with a "break".

    Consequently "break" should also leave loop-blocks, or the documentation should be updated.

    IMHO break is needless and should be abandoned in favor of last. (which linguistically still makes sense)

[reply]
[d/l]
[select]
      IMHO break is needless and should be abandoned in favor of last. (which linguistically still makes sense)
      But break doesn't act as last. It acts as next. But using next in a given/when to end the given would be odd - I'd expect next to act as continue does now.
[reply]
[d/l]
[select]
        > It acts as next.

        good catch! :)

        > But using next in a given/when to end the given would be odd

        Yes and no. Your implying that last (which doesn't look odd) couldn't be used because next is possible.

        There are many grammatically correct combinations in perl that look odd, so nobody uses them.

        Cheers Rolf

        UPDATE: consequently full orthogonality would imply allowing multiple args in given(LIST)

[reply]
[d/l]
[select]
Re^4: "when" and replacements
by John M. Dlugosz (Monsignor) on May 08, 2011 at 01:40 UTC
    "least surprise" is a powerful consideration.

    In Perl 6, you can declare the iteration variable of the for to have the desired attributes. For a given you could use an expression to cause a copy to be made: given (0+$x) {... and if you write a function that just returns its argument, and you call that function copy, then you could write given (copy $x) ... nice and sugary. It would be harder to impose aliasing if the normal case didn't.

    But since it's been this way for a while now, I think any "fix" will be to use a pragma to choose the behavior.

[reply]
[d/l]
[select]
      "least surprise" is a powerful consideration.
      Then, IMO, it should not do aliasing. Aliasing (in for/foreach/map/grep/values) still surprises people, and it causes bugs. Except for the few 1000s of people who are really, really into Perl, most programmers don't know about aliases, and it will catch them off-guard.

      As for given, it will take a long time before given is used by "the masses" (that is, the 995,000 Perl programmers that aren't deeply connected to it). Most of them will never make the connection that given shares features with for. Even if they know about aliasing in for/foreach, I very much doubt they'll be surprised if given doesn't have it.

[reply]
      After some meditation I think this attempt to transplant Perl6 features with exactly the same syntax is the source of problems.

      Perl6 handles some fundamental stuffs like sigils differently and I'm not sure if its possible within P6 to use "when"-clauses within "foreach"-blocks.

      IMHO enforcing every detail leads to anti reactions in the perl5 system. I'd rather prefer a semantical correspondence of features than introducing more breaks in symmetry and logical confusion.

      For example nobody seems to care that c-style "for" is now "loop" in Perl6...same semantic but different syntax.

      Cheers Rolf

[reply]

In Section Seekers of Perl Wisdom