in reply to In CGI::Application cgiapp_prerun vs. cgiapp_init?

The main difference is that cgiapp_init() is called once by new, while cgiapp_prerun() is called everytime a request comes in. In a standard CGI environment, both will always happen, but in a modperl environment, cgiapp_init() will only happen when the server first starts.

The order of operation in a CGI::Application is basically:

  1. cgiapp_init()
  2. setup()
  3. cgiapp_prerun()
  4. your selected run mode
  5. cgiapp_postrun()
  6. teardown()

cgiapp_init() and setup() are called in the class's new() method, while the remaining hooks are run in the class's run() method.

  • Comment on Re: In CGI::Application cgiapp_prerun vs. cgiapp_init?

Replies are listed 'Best First'.
Re^2: In CGI::Application cgiapp_prerun vs. cgiapp_init?
by weierophinney (Pilgrim) on Dec 12, 2005 at 17:35 UTC
    The order of operations you provide does a better job at explaining this than your statement, "cgiapp_init() is called once by new, while cgiapp_prerun() is called everytime a request comes in".

    Basically, cgiapp_init() is called as the object is first created in order to setup the environment for the application. cgiapp_prerun() is called prior to executing the selected run mode for a given request. I typically use cgiapp_prerun() to check if the run mode is allowed at this point by this user -- i.e. I use cgiapp_prerun() as the place to implement ACLs.

Re^2: In CGI::Application cgiapp_prerun vs. cgiapp_init?
by saberworks (Curate) on Jul 17, 2012 at 20:52 UTC
    I know this is an old thread, but I ran into an issue under mod_perl where cgiapp_init was being called on every request, not just at server (or process) start. I did some searches and found this: http://old.nabble.com/cgi%3A%3Aapplication%3A%3Adispatch-and-modperl-td29916199.html Summary: they're both called on every request. Not sure if it's a change since this thread was posted but the information here doesn't seem to be correct.

      Summary: they're both called on every request. Not sure if it's a change since this thread was posted but the information here doesn't seem to be correct.

      Its neither, its a matter scoping :)

      If this is your mod_perl application

      sub handler { my $app = MyCGIAppSubclass->new; $app->run; }

      cgiapp_init is going to get called with every request, obviously, because you're creating a new object each time -- this is typical Apache::Registry/ModPerl::Registry, running unmodified .cgi's under mod_perl

      But if your mod_perl application is

      my $app = MyCGIAppSubclass->new; sub handler { $app->run; }

      Then init/setup is only done once, and only run is done upon every request.

      So nothing changed, this are working as designed, as they always have.

      See Understanding the application flow of CGI::Application and CGI::Application::Loop

      See also the source of CGI::Application::FastCGI

      package CGI::Application::FastCGI; use strict; use base qw (CGI::Application); use FCGI; use CGI; our $VERSION = '0.02'; sub run { my $self = shift; my $request = FCGI::Request(); $self->fastcgi($request); while ($request->Accept >= 0) { $self->reset_query; $self->SUPER::run; } } sub reset_query { my $self = shift; CGI::_reset_globals(); $self->{__QUERY_OBJ} = $self->cgiapp_get_query; } sub fastcgi { my $self = shift; @_ ? $self->{__FASTCGI} = shift : $self->{__FASTCGI}; } 1;

      If you want to run CGI::Application as mod_perl handlers as I've shown above with persistent object, you will have to reset query, just like CGI::Application::FastCGI

        Ah, it's so blindingly obvious when you point it out, thanks for this. Of course I'm doing exactly what your first example shows, but not for long.