Re: Playing with (macro/source-filter) fire

sleepingsquirrel,

Intruiging idea. I am a fan of LISP style (runtime code expansion) more than C style (text substitution) macros myself. So this node grabbed my attention. But I tried running your code, which worked, but when I added a few print statments to see the guts of it, it looked to me that nothing is really happening.

Here is the output I got:


MACRO (NOT-COMPILED): for ($var1=$cons2;$var3<$cons4;$var5++) {$body6}
-e syntax OK
MACRO (COMPILED & DEPARSED): while ($var3 < $cons4) {
    $body6;
}
continue {
    ++$var5
}

SOURCE (NOT-COMPILED): #use Recur_unroll;

my $x;

for($x=0;$x<8;$x++)
{
    my $f=fact($x);
    print "factorial($x)=$f\n";
}

sub fact
{
    return(1) if ($_[0] <= 0);
    return $_[0]*fact($_[0]-1);
}

1;

-e syntax OK
SOURCE (COMPILED & DEPARSED): my $x;
while ($x < 8) {
    my $f = fact($x);
    print "factorial($x)=$f\n";
}
continue {
    ++$x
}
sub fact {
    $_[0] <= 0 and return 1;
    return $_[0] * fact($_[0] - 1);
}
'???';

REORDER: $var3 = $1;$cons4 = $2;$body6 = $3;$var5 = $4;
REG-EXED MACRO: while \((\$[a-zA-Z][a-zA-Z0-9]*) < (\d+)\) \{
    ([^}]+);
\}
continue \{
    \+\+(\$[a-zA-Z][a-zA-Z0-9]*)
\}

FINAL SOURCE: my $x;
while ($x < 8) {
    my $f = fact($x);
    print "factorial($x)=$f\n";
}
continue {
    ++$x
}
sub fact {
    $_[0] <= 0 and return 1;
    return $_[0] * fact($_[0] - 1);
}
'???';

MACRO (NOT-COMPILED): for ($var1=$cons2;$var3<$cons4;$var5++) {$body6}
-e syntax OK
MACRO (COMPILED & DEPARSED): while ($var3 < $cons4) {
    $body6;
}
continue {
    ++$var5
}

SOURCE (NOT-COMPILED): 
SOURCE (COMPILED & DEPARSED): 
REORDER: $var3 = $1;$cons4 = $2;$body6 = $3;$var5 = $4;
REG-EXED MACRO: while \((\$[a-zA-Z][a-zA-Z0-9]*) < (\d+)\) \{
    ([^}]+);
\}
continue \{
    \+\+(\$[a-zA-Z][a-zA-Z0-9]*)
\}

FINAL SOURCE: 
factorial()=1
factorial(1)=1
factorial(2)=2
factorial(3)=6
factorial(4)=24
factorial(5)=120
factorial(6)=720
factorial(7)=5040
[download]

And here is the altered Loop_unroll module:

package Loop_unroll;

use strict;
use Filter::Simple;

FILTER {

    my $unroll_limit = 10;  #max unrolling depth
    my ($var1, $cons2, $var3, $cons4, $var5, $body6);

    #This is the macro replacement pattern we're looking for...
    my $macro = 'for ($var1=$cons2;$var3<$cons4;$var5++) {$body6}';

    print "MACRO (NOT-COMPILED): $macro\n";

    #run the macro snippet through B::Deparse (better way to grab outp
+ut?)
    my $compiled_macro=`perl -MO=Deparse,-x9 -e'$macro'`; #screws with
+ ', FIX ME.
    
    print "MACRO (COMPILED & DEPARSED): $compiled_macro\n";
    
    print "SOURCE (NOT-COMPILED): $_\n";
    
    #compile the main script
    my $source = `perl -MO=Deparse,-x9 -e'$_'` if $_;     #same proble
+m, Bad dog.

    print "SOURCE (COMPILED & DEPARSED): $source\n";

    #These are crude regex to try to match perl variables and integers
    $var1 = $var3 = $var5 = '(\$[a-zA-Z][a-zA-Z0-9]*)';
    $cons2 = $cons4 = '(\d+)';
    $body6 = '([^}]+)'; #barf, should be something nicer, no nested bl
+ocks

    #deswizzle variable order (mangled by compile)
    #variables should be guarenteed to occur only once because of side
+-effects
    my $i; my $reorder;
    while ($compiled_macro=~/(var1|cons2|var3|cons4|var5|body6)/g)
    {
        $reorder.="\$$1 = \$".++$i.";"; # map vars to capturing parens
    }

    print "REORDER: $reorder\n";

    $compiled_macro=~s/([|(){}\^*+?.\[\]])/\\\\$1/g; #de-meta stuff, n
+ot $ (ick)

    my $regexed_macro;
    eval '$regexed_macro=qq/'."$compiled_macro".'/'; #interpolate patt
+ern match junk
    
    print "REG-EXED MACRO: $regexed_macro\n";    
    
    $source =~ s{$regexed_macro}
                {   eval $reorder;  # better method?
                   ($var1 eq $var3  and
                    $var1 eq $var5  and
                    $cons4 - $cons2 < $unroll_limit) ?   
                        "$var1 = $cons2;".("{$body6;} $var1++;") x ($c
+ons4-$cons2)
                        : 
                        $&;  #Lazy like a fox.
                }eg;
            
    $_ = $source;
    print "FINAL SOURCE: $source\n";
};

1;
[download]

Am I missing something?
(BTW - I am on Mac OS X (10.3.2) with Perl v5.8.0)

-stvn

Comment on Re: Playing with (macro/source-filter) fire Select or Download Code

Replies are listed 'Best First'.
Re: Re: Playing with (macro/source-filter) fire by sleepingsquirrel (Chaplain) on Jan 23, 2004 at 20:52 UTC
Hmmm. I don't see what went wrong, but here's the output of the filter I get with use Unroll_loop after uncommenting the last `print STDERR $source;` my $x; $x = 0;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;{my $f = fact($x); print "factorial($x)=$f\n";} $x++;sub fact { $_[0] <= 0 and return 1; return $_[0] * fact($_[0] - 1); } [download] And here's the result of Recur_unroll... `my $x; $x = 0; while ($x < 8) { my $f = fact($x); print "factorial($x)=$f\n"; } continue { ++$x } sub fact { $_[0] <= 0 and return 1; return $_[0] * eval{local @_=$_[0] - 1; $_[0] <= 0 and return 1; return $_[0] * fact($_[0] - 1); }; }` [download]	[reply] [d/l] [select]
Re: Re: Re: Playing with (macro/source-filter) fire by stvn (Monsignor) on Jan 23, 2004 at 21:15 UTC
Still not working for me. I tried downloading your code again (in case it was me), and ran both the Loop and Recurse filters. I have Filter::Simple version 0.79 (i upgraded this too), which uses Text::Balanced which I have version 1.89. Again, Mac OS X 10.3.2 with the default installed Perl 5.8.0. Should I upgrade anything? What platform are you on? I mean, the code looks like it should work. I dont see why its not though. -stvn	[reply]
strange differences... by sleepingsquirrel (Chaplain) on Jan 23, 2004 at 21:34 UTC
After running your version of the code on my machine I notice a couple of differences between your results and mine. When you print out the results of the compiled and deparsed macro, you get... `MACRO (COMPILED & DEPARSED): while ($var3 < $cons4) { $body6; } continue { ++$var5 }` [download] while I get... `MACRO (COMPILED & DEPARSED): $var1 = $cons2; while ($var3 < $cons4) { $body6; } continue { ++$var5 }` [download] ...I notice that the first line `$var1 = $cons2;` is missing in your report. And strangely enough there's a line is missing from your compilied and deparsed source code ... `SOURCE (COMPILED & DEPARSED): my $x; while ($x < 8) { my $f = fact($x); print "factorial($x)=$f\n"; } continue { ++$x } sub fact { $_[0] <= 0 and return 1; return $_[0] * fact($_[0] - 1); } '???';` [download] vs. my machine `SOURCE (COMPILED & DEPARSED): my $x; $x = 0; while ($x < 8) { my $f = fact($x); print "factorial($x)=$f\n"; } continue { ++$x } sub fact { $_[0] <= 0 and return 1; return $_[0] * fact($_[0] - 1); }` [download] ...the `$x = 0;` has gone missing. So this causes `$reorder` to differ... `REORDER: $var3 = $1;$cons4 = $2;$body6 = $3;$var5 = $4;` [download] vs. `REORDER: $var1 = $1;$cons2 = $2;$var3 = $3;$cons4 = $4;$body6 = $5;$va +r5 = $6;` [download] and is probably what causes the substitution match failure and thus you get the original code back. I don't know why lines would suddenly dissapear. Is it a buffering problem with the back-tick operator? update:I'd try just running... `perl -MO=Deparse,-x9 -e'for($i=0;$i<10;$i++){$x++}'` ...all by itself and see if it matches... `$i = 0; while ($i < 10) { ++$x; } continue { ++$i }` [download] That's straight from 3rd edition camel p. 480. Anyone else have any ideas? BTW, I'm running perl 5.6.1 under Slackware Linux	[reply] [d/l] [select]
Re: strange differences... by stvn (Monsignor) on Jan 23, 2004 at 21:58 UTC
Hmm, I re-downloaded all your code, tried it again, no go. My guess is that B::Deparse on OS X is different that Linux. My version of Deparse is one behind the latest, so that might play a role in this. Cool stuff none the less, thanks for showing your output, so I could see the guts of what it was doing though. -stvn	[reply]