You will be much better suited by using a proper parser than hoping to do the parsing through a regular expression.

That said, it seems to me that you only want to replace variable names and not function names, so, barring any special/convenient notation for your variable names, I can only suggest that you build a proper regular expression out of your variable names:

my $variables = join "|", map { "\\b$_\\b" } reverse sort keys %cats;
$variables = qr/$variables/;

$texttosub =~ s|($variables)|$cats{ $1 }|g;

# now there are no more variables, so we can evaluate the thing
[download]

You have two logic errors in your code:

While your string still contains variable names, it makes no sense to try to evaluate it.

Your code assumes that variable names should be matched independent of case, but your replacement can only replace the case that was stored in %cats. Your code will never replace blackCat with the value for blackcat.

join "|", map { "\\b$_\\b" } reverse sort keys %cats;
[download]

I'd suggest putting a quotemeta in there, just to play it safe. Although I agree that using a proper parser is better! skooma: See also Building Regex Alternations Dynamically.

use warnings;
use strict;
use Test::More;

my %cats = (blackcat=>5, whitecat=>10,orangecat=>20);

my ($regex) = map { qr/\b($_)\b/ } join '|', map {quotemeta}
    sort { length $b <=> length $a or $a cmp $b } keys %cats;

diag explain $regex;

sub do_replace {
    my $input = shift;
    $input =~ s/$regex/$cats{$1}/g;
    return $input;
}

is do_replace("log10(blackcat)"), "log10(5)";
is do_replace("log10(blackcat)*whitecat*(log10(orangecat))"),
    "log10(5)*10*(log10(20))";

done_testing;

__END__

# qr/\b(orangecat|blackcat|whitecat)\b/
ok 1
ok 2
1..2
[download]

#!/usr/bin/perl

# http://perlmonks.org/?node_id=1212142

use strict;
use warnings;

my %cats = ( blackcat=>5, whitecat=>10,orangecat=>20 );

while(<DATA>)
  {
  print;
  s#([a-zA-Z][A-Za-z_0-9]+)# $cats{$1} // $1 #ge;
  print;
  }

__DATA__
log10(blackcat)*whitecat*(log10(orangecat))
[download]

Outputs:

log10(blackcat)*whitecat*(log10(orangecat))
log10(5)*10*(log10(20))
[download]

$texttosub =~ s/(?!log10)|([a-zA-Z][A-Za-z_0-9]+)/$cats{$1}/i
[download]

But I am getting an infinite loop for some reason.

You don't need the alternation operator |. Lookaround Assertions like (?!...) are zero-width (tutorial). So the regex / (?!foo) | bar /x means "match a zero-length string as long as the next thing isn't foo, or match bar".

However, even with the alternation operator removed, the regex still won't do what you expect: the pattern basically means "any letters or numbers, as long as the next thing isn't log10", so the result will be "l(5)", because the substring og10 matches the pattern and there is no key og10 in %cats!

Although I've already said that I agree with Corion that using a proper parser is better, and I've shown a different solution, just for the sake of completeness and TIMTOWTDI, here are two additional ways to do what you want. First, you can use the word boundary \b to make sure that the string being matched isn't just a portion of a longer identifier: s/(?!log10)(\b[a-zA-Z][A-Za-z_0-9]+)/$cats{$1}/ works on the sample input you've shown.

Second, you could look at all identifiers, and then figure out what they are once you've matched them. Here, I'm taking advantage of the /e modifier to execute the replacement part as Perl code:

use warnings;
use strict;

my %cats = (blackcat=>5, whitecat=>10,orangecat=>20);
my %funcs = map {$_=>1} qw/ log10 sin cos /; # etc.

my $texttosub = "log10(blackcat)*whitecat*(log10(orangecat))";

$texttosub =~ s{(\b[a-zA-Z][a-zA-Z0-9_]+\b)}{
    my $repl;
    if ($cats{$1}) { $repl = $cats{$1} }
    elsif ($funcs{$1}) { $repl = $1 }
    else { die "Unknown identifier '$1'" }
    print "Matched '$1', replacement '$repl'\n"; # Debug
    $repl }eg;

print $texttosub, "\n";

__END__

Matched 'log10', replacement 'log10'
Matched 'blackcat', replacement '5'
Matched 'whitecat', replacement '10'
Matched 'log10', replacement 'log10'
Matched 'orangecat', replacement '20'
log10(5)*10*(log10(20))
[download]

Crossposted to StackOverflow. While crossposting is acceptable, it is considered polite to inform about it so that efforts are not duplicated.

Update: Also, your node title is quite long. Please see How do I compose an effective node title?

My bad. I thought the sites were unrelated.

Edit: Thanks let me check it out.

I thought the sites were unrelated.

They are unrelated — except they're both on the Interwebs! Some folks attend one or the other or both (or neither, of course). Also, please see How do I change/delete my post? for site etiquette and protocol regarding changing your post.

---

Give a man a fish:  <%-{-{-{-<

Appropriating from ideas of several other postings including the OP, tybalt89's example, haukex' s use of \b, and a hopefully plausible approach to Corion's concern about case sensitivity, perhaps the following is simple but still adequate. Lower casing the "cat" may or may not be the exact right approach to being case insensitive.

use strict;
use warnings;

my %cats = ( blackcat=>5, whitecat=>10,orangecat=>20 );
my $texttosub = 'log10(blackcat)*whitecat*(log10(orangeCat))';
$texttosub =~ s/\b(?!log10)([a-zA-Z][A-Za-z_0-9]+)\b/$cats{lc $1}/ige;
print "$texttosub\n";
[download]

Output

log10(5)*10*(log10(20))
[download]