in reply to Falling through Perl 5.10 switch statement

The Imp of the Perverse still wonders if it is possible to directly "goto" (maybe that's a hint right there...) the next block, skipping the when check.

Well, goto itself won't work in my perl 5.9.4:

use feature qw(say switch);

my $data = 1;
given($data) {
    when (1) { say 'one'; goto two;  }
    when (2) { two: say 'two' ;      }
    default  { say 'something else'; }
}

[download]

This prints:

one
panic: goto at /tmp/test line 5.

[download]

Interesting error message though.

Replies are listed 'Best First'.
Re^2: Falling through Perl 5.10 switch statement
by Trizor (Pilgrim) on Apr 07, 2007 at 20:33 UTC
    Thats because your trying to goto a block that looks like it would be optimized out of existance. Since $data is a constant the optimizer components that look at switches are probably smart enough to notice that only the when (1) case will run and will replace the whole switch in the op tree with simply say 'one'; goto two;.

    Disclaimer: I don't claim to have intimate knowledge of how switches are implemented in the optree or how the optimizer treats them. Take this explanation with a grain of salt, it may sound good, but it may also be flat out wrong.
[reply]
[d/l]
[select]

      I'd maybe buy your argument if the code was 

      given (1)

      [download]

      but I've never known perl to optimize lexicals away, even if those who always hold the same constant value. The following two optrees should be identical if perl optimized constant lexicals away: 

      >perl -MO=Concise -e"my $data = 1; print $data;"
a  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 -e:1) v ->3
5     <2> sassign vKS/2 ->6
3        <$> const[IV 1] s ->4
4        <0> padsv[$data:1,2] sRM*/LVINTRO ->5
6     <;> nextstate(main 2 -e:1) v ->7
9     <@> print vK ->a
7        <0> pushmark s ->8
8        <0> padsv[$data:1,2] l ->9       <-- It uses the lexical,
-e syntax OK                                  not a constant.

>perl -MO=Concise -e"print 1;"
6  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 -e:1) v ->3
5     <@> print vK ->6
3        <0> pushmark s ->4
4        <$> const[IV 1] s ->5            <-- This is a constant.
-e syntax OK

      [download]

      Perl 5.8.8

      Update: Here's another example using code more similar to the OP's: 

      >perl -MO=Concise -e"my $data = 1; if ($data) { print 'moo' }"
c  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 -e:1) v ->3
5     <2> sassign vKS/2 ->6
3        <$> const[IV 1] s ->4
4        <0> padsv[$data:1,4] sRM*/LVINTRO ->5
6     <;> nextstate(main 4 -e:1) v ->7
-     <1> null vK/1 ->c
8        <|> and(other->9) vK/1 ->c      <-- "if" not optimized away.
7           <0> padsv[$data:1,4] s ->8   <-- Lex not optimized away.
-           <@> scope vK ->-
-              <0> ex-nextstate v ->9
b              <@> print vK ->c
9                 <0> pushmark s ->a
a                 <$> const[PV "moo"] s ->b
-e syntax OK

>perl -MO=Concise -e"if (1) { print 'moo' }"
6  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 3 -e:1) v ->3
-     <@> scope vK ->6                  <-- "if" optimized away.
-        <0> ex-nextstate v ->3
5        <@> print vK ->6
3           <0> pushmark s ->4
4           <$> const[PV "moo"] s ->5
-e syntax OK

      [download]
[reply]
[d/l]
[select]
      Thats because your trying to goto a block that looks like it would be optimized out of existance.
      That sounds like a nice explanation; however, the following script panics with the same message if there is a "1" in the command line: 
      use feature qw(say switch);

for my $data (@ARGV) {
    given($data) {
        when (1) { say 'one'; goto two;  }
        when (2) { two: say 'two' ;      }
        default  { say 'something else'; }
    }
}

      [download]
[reply]
[d/l]
Re^2: Falling through Perl 5.10 switch statement
by ysth (Canon) on Apr 08, 2007 at 06:09 UTC
    Still a problem in bleadperl @ 30869.
[reply]
      Time to rollback :p
[reply]

In Section Meditations