package CGI::Fast;
use strict;
use warnings;
use if $] >= 5.019, 'deprecate';


use CGI;
use CGI::Carp;
use FCGI;
# use vars works like "our", but is compatible with older Perls.
use vars qw(
@ISA = ('CGI');

# workaround for known bug in libfcgi
while (($ignore) = each %ENV) { }

# override the initialization behavior so that
# state is NOT maintained between invocations
sub save_request {
    # no-op

# If ENV{FCGI_SOCKET_PATH} is specified, we maintain a FCGI Request handle
# in this package variable.
use vars qw($Ext_Request $socket $socket_perm $queue);

sub import {
    my ($package,@import) = @_;
    # check imports for this class then pass on
    # imports to SUPER class
    for (my $i = 0; $i < scalar( @import ); $i++) {
        if ( $import[$i] eq 'socket_path' ) {
            $socket = $import[$i+1];
        } elsif ( $import[$i] eq 'socket_perm' ) {
            $socket_perm = $import[$i+1];
        } elsif ( $import[$i] eq 'listen_queue' ) {
            $queue = $import[$i+1];

sub _create_fcgi_request {
    my ( $in_fh,$out_fh,$err_fh ) = @_;
    # If we have a socket set, explicitly open it
    if ($ENV{FCGI_SOCKET_PATH} or $socket) {
        my $path    = $ENV{FCGI_SOCKET_PATH}  || $socket;
        my $perm    = $ENV{FCGI_SOCKET_PERM}  || $socket_perm;
        my $backlog = $ENV{FCGI_LISTEN_QUEUE} || $queue || 100;
        my $socket  = FCGI::OpenSocket( $path, $backlog );
        if ($path !~ /^:/ && defined $perm) {
            chmod $perm, $path or croak( "Couldn't chmod($path): $!" );
        return FCGI::Request(
            ( $in_fh  || \*STDIN ),
            ( $out_fh || \*STDOUT ),
            ( $err_fh || \*STDERR ),
    else {
        return FCGI::Request(
            ( $in_fh  || \*STDIN ),
            ( $out_fh || \*STDOUT ),
            ( $err_fh || \*STDERR ),

    my ( $in_fh,$out_fh,$err_fh );

    sub file_handles {
        my ($self, $handles) = @_;

        if ( ref( $handles ) eq 'HASH' ) {
            $in_fh  = delete( $handles->{fcgi_input_file_handle} );
            $out_fh = delete( $handles->{fcgi_output_file_handle} );
            $err_fh = delete( $handles->{fcgi_error_file_handle} );

    sub new {

		# the interface to the ->new method is unfortunately somewhat
		# overloaded as it can be passed:
		#         nothing
		#         an upload hook, "something", 0
		#         an initializer, an upload hook, "something", 0
		# these then get passed through to the SUPER class ( that
		# also has a constructor that can take various order of args
        my ($self, @args) = @_;

        if (
			! $args[0]
			|| (
				ref( $args[0] )
				&& UNIVERSAL::isa( $args[0],'CODE' )
				&& ! $args[3]
		) {
            $Ext_Request ||= _create_fcgi_request( $in_fh,$out_fh,$err_fh );
			my $accept = $Ext_Request->Accept;
            return undef unless ( defined $accept && $accept >= 0 );
        $self->_setup_symbols(@CGI::SAVED_SYMBOLS) if @CGI::SAVED_SYMBOLS;
        return $CGI::Q = $self->SUPER::new(@args);


=head1 NAME

CGI::Fast - CGI Interface for Fast CGI

=for html
<a href=''><img src='' alt='Build Status' /></a>
<a href=''><img src='' alt='Coverage Status' /></a>


    use CGI::Fast
        socket_path  => '9000',
        socket_perm  => 0777,
        listen_queue => 50;

    use CGI qw/ :standard /;

    $COUNTER = 0;

    # optional, will default to STDOUT, STDERR
        fcgi_output_file_handle => IO::Handle->new,
        fcgi_error_file_handle  => IO::Handle->new,

    while ($q = CGI::Fast->new) {


CGI::Fast is a subclass of the CGI object created by  It is
specialized to work with the FCGI module, which greatly speeds up CGI
scripts by turning them into persistently running server processes.
Scripts that perform time-consuming initialization processes, such as
loading large modules or opening persistent database connections, will
see large performance improvements.

Note that as CGI::Fast is based on it is no longer advised as
a way to write Perl web apps. See L<>
for more information about this


In order to use CGI::Fast you'll need the FCGI module.  See for details.


FastCGI scripts are persistent: one or more copies of the script
are started up when the server initializes, and stay around until
the server exits or they die a natural death.  After performing
whatever one-time initialization it needs, the script enters a
loop waiting for incoming connections, processing the request, and
waiting some more.

A typical FastCGI script will look like this:

    use CGI::Fast;
    while ($q = CGI::Fast->new) {

Each time there's a new request, CGI::Fast returns a
CGI object to your loop.  The rest of the time your script
waits in the call to new().  When the server requests that
your script be terminated, new() will return undef.  You can
of course exit earlier if you choose.  A new version of the
script will be respawned to take its place (this may be
necessary in order to avoid Perl memory leaks in long-running
scripts).'s default CGI object mode also works.  Just modify the loop
this way:

    while (CGI::Fast->new) {

Calls to header(), start_form(), etc. will all operate on the
current request.


See the FastCGI developer's kit documentation for full details.  On
the Apache server, the following line must be added to srm.conf:

    AddType application/x-httpd-fcgi .fcgi

FastCGI scripts must end in the extension .fcgi.  For each script you
install, you must add something like the following to srm.conf:

    FastCgiServer /usr/etc/httpd/fcgi-bin/file_upload.fcgi -processes 2

This instructs Apache to launch two copies of file_upload.fcgi at
startup time.


Any script that works correctly as a FastCGI script will also work
correctly when installed as a vanilla CGI script.  However it will
not see any performance benefit.


FastCGI supports a TCP/IP transport mechanism which allows FastCGI scripts to run
external to the webserver, perhaps on a remote machine.  To configure the
webserver to connect to an external FastCGI server, you would add the following
to your srm.conf:

    FastCgiExternalServer /usr/etc/httpd/fcgi-bin/file_upload.fcgi -host sputnik:8888

Two environment variables affect how the C<CGI::Fast> object is created,
allowing C<CGI::Fast> to be used as an external FastCGI server. (See C<FCGI>
documentation for C<FCGI::OpenSocket> for more information.)

You can set these as ENV variables or imports in the use CGI::Fast statement.
If the ENV variables are set then these will be favoured so you can override
the import statements on the command line, etc.


=item FCGI_SOCKET_PATH / socket_path

The address (TCP/IP) or path (UNIX Domain) of the socket the external FastCGI
script to which bind an listen for incoming connections from the web server.

=item FCGI_SOCKET_PERM / socket_perm

Permissions for UNIX Domain socket.

=item FCGI_LISTEN_QUEUE / listen_queue

Maximum length of the queue of pending connections, defaults to 100.


For example:

    use CGI::Fast
        socket_path  => "sputnik:8888",
        listen_queue => "50"

    use CGI qw/ :standard /;


    while ($q = CGI::Fast->new) {


    use CGI::Fast;
    use CGI qw/ :standard /;


    $ENV{FCGI_SOCKET_PATH} = "sputnik:8888";

    while ($q = CGI::Fast->new) {

Note the importance of having use CGI after use CGI::Fast as this will
prevent any CGI import pragmas being overwritten by CGI::Fast. You can
use CGI::Fast as a drop in replacement like so:

    use CGI::Fast qw/ :standard /


FCGI defaults to using STDOUT and STDERR as its output filehandles - this
may lead to unexpected redirect of output if you migrate scripts from
to CGI::Fast. To get around this you can use the file_handles method, which
you must do B<before> the first call to CGI::Fast->new. For example using

        fcgi_output_file_handle => IO::Handle->new,
        fcgi_error_file_handle  => IO::Handle->new,

    while (CGI::Fast->new) {

Overriding STDIN using the C<fcgi_input_file_handle> key is also possible,
however doing so is likely to break at least POST requests.

=head1 CAVEATS

I haven't tested this very much.

=head1 LICENSE

Copyright 1996-1998, Lincoln D. Stein.  All rights reserved. Currently
maintained by Lee Johnson

This library is free software; you can redistribute it and/or modify
it under the same terms as Perl itself.

Address bug reports and comments to:

=head1 BUGS

This section intentionally left blank.

=head1 SEE ALSO

L<CGI::Carp>, L<CGI>