...
my $replace = '"$1 Perl"';

$text2 =~ s/$match/$replace/ee;
[download]

my $replace = '$1 Perl'; # use original setting of $replace

$text2 =~ s/$match/'"'.$replace.'"'/ee;
[download]

my $replace = sub { "$1 Perl" };

$text2 =~ s/$match/$replace->()/e;
[download]

That works - thanks :)

But, reading through the regex documentation, I still can't work out exactly why. Hmmm.

so I guess needs to be \\" in the replace string for quotes. Arghhh, double escaping!!

Or how to embed " in the output - but that's not a biggie right now.

You want to produce the code

"$1 \"Perl\""
[download]

so

my $replace = '"$1 \\"Perl\\""';  # Produces: "$1 \"Perl\""
s/$match/eval $replace/e
[download]

Technically, the slashes don't need to be doubled since single quotes are forgiving when it's unambiguous.

To explain what's going on it might be helpful to use underlined bold text to represent strings. So, one can write:

eval 2+2 → 4

and this means if you eval the string 2+2, you'll get the string 4. In perl this is just:

my $x = '2+2';
my $y = eval $x; # $y is the string '4'
[download]

In your original example, what is happening during the substitution s/$match/$replace/e is:

eval $replace → $1 Perl

I.e., the match is replaced with the string $1 Perl, and that is why $text2 contains $1 Perl world.

How about just adding another /e modifier to evaluate the substitution again? Unfortunately this doesn't work because $1 Perl world is not valid perl syntax:

eval eval $replace → eval $1 Perl world → syntax error

When $replace and the substitution is written as:

my $replace = '"$1 Perl"';
$text2 =~ s/$match/$replace/ee;
[download]

the evaluation of the replacement proceeds as follows:

eval eval $replace → eval "$1 Perl"

and because of the double quotes this last eval yields $1 concatenated with a space and the string Perl.

/e treats the contents of the replace operand of the s/// operator as Perl source code. Therefore, $replace (not its content) is treated as source code. An expression consisting of a variable name returns its contents, so s/.../$replace/e replaces the match with the contents of $replace.

To execute Perl code in a variable, you need eval EXPR. Change
s/$match/$replace/e
to
s/$match/eval qq{"$replace"}/e

The latter can also be written as
s/$match/qq{"$replace"}/ee
(note the double "e") but I find that to be rather obfuscated. eval EXPR is dangerous so it shouldn't be hidden.

Update: Added missing code to add quotes to the code to execute.

$text2 =~ s/$match/$replace/;
$text2 =~ s/$match/$replace/e;
[download]

I used this and it works my $replace = "$1 Perl";

What about this replace text, which attempts to lower case the protocol and hostname in a URL string, but leave the path (and anything else) as is:

my $url = 'http://www.FOo.COm/wibbLE';
my $search = '^([^:]+://[^/]+)/?(.*)?$';
my $replace = '\L$1\E/$2';
printf "%s\n", $url if ($url =~ s!$search!$replace!);
[download]

Anyone got a way of not resorting to "/e" or "/ee" or eval to make this work? For me the $replace is coming from a database, and I don't trust the code so I dont want it doing an "unlink /" or similar! :)