Re: Perl Variables Being Retained

[tobyink]

I think it's likely that you do not have IIS configured to run via CGI, but instead have it configured to run via FastCGI or some other persistent mechanism.

That's generally considered a good thing. It means that your Perl script can save doing a bunch of initialization stuff for every single request (loading modules, connecting to databases, etc) and instead do it just once. This could allow you to process many more requests per second.

However, it does mean that you've got to be careful not to persist stuff you don't want to persist. Generally making sure that variables are lexically scoped (i.e. my $var) should be sufficient to take care of that. The other thing you'll need to be careful about is leaking memory. This can happen if you have cyclical references pointing to each other, e.g.:

my $a = [];
my $b = [];
push @$a, $b;
push @$b, $a;

... these two arrays now each contain a reference to each other, so Perl doesn't ever realise it can free that memory, and doesn't free it until the end of the process. If every web request creates these two arrays, then the process ends up eating a little more memory for each request.

But anyway, although you need to be a little more careful than you do with CGI, generally speaking a persistent Perl interpreter is a good way of serving web content.

[sundialsvc4]

Specifically in terms of “Perl not cleaning up the memory,” also note the weaken in Scalar::Util.   If you have data structures referring to one-another such that a circular reference is formed, memory does not clean up but rather “leaks.”   You avoid this by weakening one of the links in the chain.   Test::Memory::Cycle is another tool that is designed specifically to look for these circular reference-chains.