NAME
Net::Spooler - A Perl extension for writing spooling daemons
SYNOPSIS
# Create a subclass of Net::Spooler
use Net::Spooler;
package MySpooler;
@MySpooler::ISA = qw(Net::Spooler);
# Inherit everything, except a single method:
sub ProcessFile {
my $self = shift; my $file = shift;
# Try to process the file here
...
# Raise an exception, if something went wrong:
die "Failed: $!" unless Something();
# Return to indicate sucess.
}
# Create and run the spooler
package main;
my $spooler = Net::Spooler->new(
'spool-dir' => '/var/myspooler'
);
$spooler->Bind();
DESCRIPTION
This package contains a spooling daemon, in other words a
process, that accepts files from an outside source (currently a
Unix or TCP/IP socket), stores them in a spooling directory and
processes them.
The package is implemented as an abstract base class: It is not
usefull in itself, but you can get your spooling daemon easily
by deriving a concrete subclass from `Net::Spooler'. In the best
case you can inherit everything and overwrite just a single
method, the *ProcessFile* method, which attempts to process a
single file from the spooling directory.
`Net::Spooler' is in turn derived from the `Net::Daemon'
package, thus it borrows class design, in particular methods and
attributes, from `Net::Daemon'. See the Net::Daemon manpage for
details on this superclass.
However, there are a few additions to `Net::Daemon':
Attributes
Like in `Net::Daemon', attributes can be set via the command
line, in the config file or as constructor arguments (order
descending from most important). And remember, that you can use
the `Net::Daemon' attributes too! See the Net::Daemon manpage.
*admin* (B<--admin=<email>)
The administrators email address. From time to time it
may happen, that the admin receives an email in case of
problems.
*expiretime* (--expiretime=<time>)
If processing a file fails repeatedly, the file may
finally expire. This means that the file will be removed
from the spool directory and a message is sent to the
administrator.
The default value are 432000 seconds (5 days). A value
of 0 means that expiration never happens.
Example: Expire after 3 days.
--expiretime=259200
*maxsixe* (--maxsize=<maxsize>)
By default the maximum size of a file is restricted to
100000 bytes and larger files will be rejected. This
option is changing the size, a value of 0 means
disabling the limitation.
Example: Disable max size
--maxsize=0
*processtimeout* (--processtimeout=<timeout>)
If processing a single file may result in an endless
loop, or simply run too long, then you may specify a
timeout. The daemon will raise a signal after the given
amount of seconds and stop processing the file, as if
the method `ProcessFile' raised an exception.
The default value is 0 seconds which means that no
timeout is used.
Example: Use a timeout of 30 seconds.
--processtimeout=30
*loop-timeout* (--loop-timeout=<time>)
If processing a file failed, the spooler will reprocess
the file later by forking a child process after the
given amount of seconds, by default 300 seconds (5
minutes). This child process will run through all
scheduled file
*spool-dir* (--spool-dir=<dir>)
If the daemon accepts files, they are stored in the
*spool directory*. There's no default, you must set this
attribute.
Example: Use /var/myspooler as a spool directory.
--spool-dir=/var/myspooler
*tmpfiles*
This attribute is for internal use only. It contains an
hash ref, the keys being temporary file names to be
removed later.
Methods
As already said, the `Net::Spooler' package inherits from
`Net::Daemon'. All methods of the superclass are still valid in
`Net::Spooler', in particular access control and the like. See
the Net::Daemon manpage for details.
Processing a file
$self->ProcessFile($file)
(Instance method) Called for processing a single file.
This is typically the only method you have to overwrite.
The method raises an exception in case of errors. If an
exception is raised, the scheduler will later retry to
process the file until it expires. See the *queuetime*
and *expiretime* attributes above.
If processing a file exceeds the *processtimeout* (see
above), then the scheduler will cancel processing the
method and continue as if it raised an exception. (This
timeout can be disabled by setting it to 0, the default
value.)
If the method returns without raising an exception, then
the scheduler assumes that the file was processed
successfully and remove it from the spool directory.
Choosing file names
my $sfile = $self->SequenceFile();
my $seq = $self->Sequence($sfile);
my $dfile = $self->DataFile($seq);
my $cfile = $self->ControlFile($seq);
(Instance methods) If the daemon receives a new file, it
has to choose a name for it. These names are constructed
as follows:
First of all, a so-called sequence number is generated
by calling the method *Sequence*. By default these are
the numbers 1, 2, 3, ... in 8 hex digits (00000001,
00000002, 00000003, ...). The last generated sequence
number is always stored in the sequence file (by default
$spool-dir/.sequence, set by calling the *SequenceFile*
method).
Two files are generated for processing the file: The
*data file* is the unmodified file, as received by the
client. The *control file* contains information used
internally by `Net::Spooler', for example the time and
date of spooling this file. By default the names $spool-
dir/$seq.dat and $spool-dir/$seq.ctl are used, generated
by calling the methods *DataFile* and *ControlFile*.
Temporary file names are derived by adding the suffix
.tmp.
Typically you rarely need to overwrite these methods.
Accepting a file from the client
$self->ReadFile($socket, $fh, $file, $control);
(Instance method) This method is actually reading the file $file
from the socket $socket. The file is already opened and the
method must use the file handle $fh for writing into $file. (The
file name is passed for creating error messages only.)
The method may store arbitrary data in the hash ref $control:
This hash ref is stored in the control file later.
The default implementation is accepting a raw file on the
socket. You should overwrite the method, if you are accepting
structured data, for example 4 bytes of file size and then the
raw file. However, if you do overwrite this method, you should
consider the *maxsize* attribute. (See above.)
A Perl exception is raised in case of problems.
Creating the control file
$self->ControlFile($fh, $file, $control);
(Instance method) Creates the control file $file by writing the
hash ref $control into the open file handle $fh. (The file name
$file is passed for use in error messages only.)
The default implementation is using the `Data::Dumper' module
for serialization of $control and then writing the dumped hash
ref into $fh.
A Perl exception is raised in case of problems; nothing is
returned otherwise.
Reading the control file
my $ctrl = $self->ReadControlFile($file);
(Instance method) This method reads a control file, as created
by the *ControlFile* method and creates an instance of
*Net::Spooler::Control*.
The default implementation does a simple require (in a Safe
compartment for security reasons, see the the Safe manpage
package for details) for loading the hash ref from the file. The
hash ref is then blessed into the package corresponding to
$self: The package name of $self is taken by appending the
string ::Control.
The method returns nothing, a Perl exception is thrown in case
of trouble.
AUTHOR AND COPYRIGHT
This package is
Copyright (C) 1999 Jochen Wiedmann
Am Eisteich 9
72555 Metzingen
Germany
E-Mail: joe@ispsoft.de
All rights reserved.
You may distribute under the terms of either the GNU General
Public License or the Artistic License, as specified in the Perl
README file.
SEE ALSO
L<Net::Daemon(3)>