In this code golf answer, aross gives a solution that presumably uses IBM-850 encoding as parameter values in PHP"
echo preg_filter("/^(.*?[aeiouy]+)/i","$1 $1 $0",$argn);
echo preg_filter(~ðíÎÐı└ñ×ÜûÉèåóÈÍðû,~█╬▀█╬▀█¤,$argn); # Encoded
How does this work? Why are the parameters not quoted? How come only the parameters are encoded and not the rest of the code?
It hasn't so much to do with IBM-850, that's just a codepage filling out the 8th bit beyond ASCII to give a representation to the bytes you'll end up with.
The key here is the bitwise not operator ~ which flips all the bits - 0 becomes 1, and 1 becomes 0. If you dump ~"/^(.*?[aeiouy]+)/i" to a file and open it up as 850 it'll look like:
ðíÎÐı└ñ×ÜûÉèåóÈÍðû
And likewise ~"$1 $1 $0" looks like:
█╬▀█╬▀█¤
So you see where this is headed.
In PHP an undefined constant is assumed to have a string value matching its name. For example:
var_dump(foo);
Outputs string(3) "foo" (as well as the notice "Use of undefined constant foo - assumed 'foo'", if notices are on.)
When either of the two gibberish strings above are put in a PHP script without quotes they're taken as undefined constants with their names assumed for their values as well.
Now prepend each with ~ to flip their bits back and you've got the original regex and replacement strings:
preg_filter("/^(.*?[aeiouy]+)/i","$1 $1 $0",$argn)
Only those parameters had their bits flipped because they're the only string literals, which is what this trick applies to. For each string it's shaving off a pair of quotes in exchange for taking on only a single tilde.
The bit flipping had to be done because either of the original strings on their own without quotes would've landed parse errors.
Clever way to net two bytes.