$line = qq{10 PRINT "<29>This should be dim.<30>EThis should be revers
+ed.<28>This is bold."};

my %fix = ( 
    "<29>" => '";@(-38);"',
    "<28>" => '";@(-40);"', 
    "<27>" => '";@(-35);"',
    "<30>" => "<30>",
);
    
$line =~ s/"([^"]*)"/my $s = $1;$s =~ s{(<\d+>)}{$fix{$1}}g;'"'.$s.'"'
+/ge;

#or
# $line =~ s/("[^"]*")/my $s = $1;$s =~ s{(<\d+>)}{$fix{$1}}g;$s/ge;
[download]

This is a regexp inside regexp. The outer one "extracts" all strings from a line, the second replaces the <\d+> thingies inside. This way it's safe even in the unlikely case that a <\d+> appears outside a string in the code.

Another thingie is that I keep the whole replacements in the hash, including the double quotes and semicolons, and replace <30> by itself. And an unknown <\d+> by nothing.

If you wanted to leave unknown alone you'd change the regregexpexp to

$line =~ s/"([^"]*)"/my $s = $1;$s =~ s{(<\d+>)}{$fix{$1}||$1}ge;'"'.$
+s.'"'/ge;
[download]

or

$line =~ s/("[^"]*")/my $s = $1;$s =~ s{(<\d+>)}{$fix{$1}||$1}ge;$s/ge
+;
[download]

Then you would not have to include the <30> in the %fix hash.

I assume that to include a doublequote in a string (in that version of Basic) you have to double it.

  Jenda