/ 
Config-Framework-1.01
River stage zero No dependents
/ Framework

Config::Framework

This module provides global configuration data to perl programs which live under a standard distribution. For more information on the standard, perldoc or man 'manifesto'.

In addition to providing global configuration data, this module also provides facilities for reading and writing XML files using Data::DumpXML, decrypting and encrypting data via Crypt::CBC, Logging, and sending alerts to a specified email address.

Config::Framework is built around the ability to dump and restore arbitrary perl data structures to and from files. This concept is central to what Config::Framework does, which is essentially, to manage configuration data. There are several perl modules out there which can handle the dumping and restoring of data structures, mainly Data::Storable, which is a fine module, using binary files. However, in the general spirit of keeping things as simple, and open as possible, I decided to go with Data::DumpXML, which uses an XML DTD to dump and restore the data. This is good because in a pinch a human can look at the dump'd file and figure out what's going on.

So, we let Data::DumpXML worry about the file format, and we keep all of our data in a universally defined data structure (see "Configuration Data Structure" below). Using this module, you can load that data into the Config::Framework object. Confused? Ok lemme break it down a bit more. The Config::Framework object is the thingy you create using the New method (see below). Any perl object is really just a blessed reference, meaning that it serves a dual purpose: you can call methods against it, and it also points to some other perl thing, such as a string or an array, or a bit of code, etc. In this case it points to a hash reference. So once you have created a Config::Framework object, you can both use the data stored in the hash it points to, and also call methods against the object. So now you can see what a handy thing this is! Using this module you not only have a place to store any configuration data (or actually any data structure at all), as well as an object which you can call a bunch of handy methods on.

So there's still a big sticky-wicket. Once you've got this magic object, and you've loaded your data into it, how do you find it? What if you've got 20 different config files you need to load? How do you keep things organized? Answer: 'configNamespace'.

"configNamespace" is just a fancy name I made up for a hash key. Sounds cool, huh? Well really it's a bit more than that, but not much more. Each file that you load into the object must define an element in it's root hash (see 'Configuration Data Structure' below) called 'configNamespace'. This tells Config::Framework where to load the data under it's main hash reference. Say that I had config file that I wanted to load that defined configNamespace to be "myProgram". If $Config was my Config::Framework object, I could find the data I just loaded under $Config->{'myProgram'}. Pretty much anything goes when it comes to what to use for your configNamespcace, there is however one reserved word. YOU CANNOT USE 'Secure', WHICH IS RESERVED FOR DATA FROM passwds.xml.

So, what if I try to load two files which have the same 'configNamespace'? Well pretty much what would happen is that the second file would overwrite the first, unless you defined 'protectNamespace' to be a nonzero value in the first file. In that case, Config::Framework would throw an error when you try to load the second file.

Ok so one other thing about configNamespaces. They can be cascading. That is to say, you may have several levels of them. This especially happens in config files which define a 'User Config'. So, a config can define a second config to load into a configNamespace underneath the existing one. It sounds confusing, but isn't.

Configuration Data Structure

This is the data structure that Config::Framework expects to see in configfiles that you attempt to load into the object. Note that Config::Framework can load ANY perl data structre dumped using Data::DumpXML, however, it can't store the data in it's object unless you use this format. The data structure should be a hash at it's root ... after that it's pretty much as arbitrary as you want to make it. The only thing that's REALLY REQUIRED is this:

$hash{configNamespace} = "whatever you want it to be";

Beyond that, practically speaking, you're on your own. However, there are some guidelines which you should probably adhere to. Following is a datastructure that I reccomend that you use. Obviously, everything in this structure does not apply to every program out there, all I ask is that if you need to store any of this data in a config, that you use this format ... that keeps everyone sane.

The Format

%DATA = (
 ## Program Identification #######################
   #name of program (not filename, example: "Skelator" not "skelator.pl")
    'Program Name'			=> $program_name,
   #most recent version of this program
    'Program Version'		=> $version,
   #each key in this hash is the name of something which is required
   #for this program to run. The value of the key would be a brief
   #description of that thing. This might be a list of modules, or a list
   #of configuration files. This is meant to be general purpose as possible.
   #(following are example entries).
    'Dependencies'			=> {
    	'DBD\::Sybase'	=> "perl module for talking to Sybase databases",
    	'/csserve0/loc4/secsup/Prod/config/my_file.txt'	=> "a random file I need"
    },
    
 ## Authors ######################################
   #name of the lead developer for this project
    'Primary Author'		=> $lead_developer,
   #the email of the lead developer so we can bug him when things go wrong
    'Primary Author Email'	=> $lead_developer_email,
   #other people might have also worked on this project, if so this is an
   #array of them
    'Developers'			=> [
        "array",
        "of",
        "other",
        "developers",
    ],
    
 ## Config Meta Data #############################
   #You oughta know what this is by now ... ;-)
    'configNamespace'		=> $configNamespace,
   #throw an error if someone tries to overwrite your 'configNamespace' in 
   #the global config object. (1 throws error)
    'protectNamespace'		=> 1|0,
   #the current version of this config file
    'Config Version'			=> $config_version,
   #some programs define a user-specific configuration file
   #which is stashed in the user's home directory.
    'User Config'				=> {
       #filename to look for in user's home directory
        'FileName'			=> $file_name,
       #if 'FileName' isn't found load this skeleton file and copy it to
       #the user's home directory
        'Skeleton File'		=> $skeleton_file_name
       #if the version of the user's config is less than this number, the 
       #file format is updated from the 'Skeleton File'
        'Require Version'	=> $version_to_require
    },
    
 ## Documentation ################################
    'Documentation'		=> {
       #Program (executable) lives in this directory
        'Path'				=> $path_to_program,
       #The name of the executable file under 'Path'
        'ExecFileName'	=> $filename,
       #a brief synopsis of what the program does
        'Synopsis'		=> $brief_synopsis,
       #a string containing a description of commandline syntax
        'Syntax'		=> $syntax,
       #any additional information that might be useful
        'Aux Info'		=> $more_optional_info
    },
    
 ## Program Specific Keys ########################
   #would go here ...
);

Global Configuration Data

Annother nice thing that Config::Framework does for you is to serve as a place where global configuration data (i.e. not read from a file) is stored. Durring the Makefile.PL process, you are asked a series of questions, where you set global configuration data. This data is available at the root level of any Config::Framework object, without loading any files. At the moment, if you want to add additional global config items you've got to edit the file manually and modify the %GLOB_CONFIG hash at the top of the module. keep in mind that you can override any of these by sending alternate values as the argument to the equivalent option to New. Here's a brief rundown on what's in there by default:

'program'

This is the name of the currently running process. This is pretty much $0, excpet we strip the path off, and just give the filename.

'v_root'

"Virtual Root". To get more familiar with this concept, you should read the SecsupManifesto (perldoc or man 'manifesto'). This is directory under which everything lives. If you are using the standards in the manifesto (you are, aren't you?) then this will be your 'prod' directory. USE THIS. IT MEANS NEVER HAVING A HARD CODED PATH IN ANY OF YOUR PROGRAMS.

'config_loc'

"Configuration Location" This is the directory under 'v_root' which contains all of your configuration files. When you load a config file using LoadConfig or LoadXMLConfig, we presume that the file or path to the file exists under this location. It's good to keep things organized! ;-)

'SYBASE"

The path to the sybase client libraries. If defined, this variable is exported to the process' shell when New is called.

'ORACLE_HOME'

The path to oracle client libraries, also exported to the process' shell upon object creation if defined.

'LD_LIBRARY_PARH'

If this is defined, then this LD_LIBRARY_PATH will be exported to the process' shell at object creation. If you are using the secsup manifesto this will be the 'lib' directory under 'v_root'.

'sendmail'

This is the path to the sendmail binary. We need this for sending alerts to people via AlertAdmin.

'admin'

The email address to send mail to when something goes drastically wrong ... also the default email address to send alerts to via AlertAdmin.

'EnvExportList'

This is an array containing the default set of values to export to the process' shell (if defined). By default, this is "SYBASE", "ORACLE_HOME", "ORACLE_SID", "ARTCPPORT", and "LD_LIBRARY_PATH". You and add to this list using the EnvExportList option to New. You can prevent any values being exported at all by using the EnvExportOverride option to New.

'Crypt'

Use this Crypt::CBC compliant enctrpyion module to decrypt data in passwds.xml. Use it by default for any other encrypted files also.

'Key'

Use this pasphrase to decrypt data in passwds.xml. Use it by default for ant other encrypterd files as well.

To Do

Write Test Programs for 'make test'

Modify Log to use Net::Syslog.

Add hook so that we can specify an external subroutine to interpolate existing data
from user configs which have an out of date version and which need to be updated.

Dependencies

Crypt::CBC Version 2.02+
Data::DumpXML::Parser Version 1.01+
File::Copy Version 2.03+

Methods

New (contructor)

Notes

This is the object constructor. Upon failure, the method will return the undef value, and an error message will be written to $Config::Framework::errstr.

Synopsis

$Object = Config::Framework::New([options]) || die $Config::Framework::errstr, "\n";

Options

EnvExportList

A list of other options you wish to export to the process' shell environment. This should be an array reference to a list of option names. This isn't as straight forward as it sounds, here's an example:

$Object = Config::Framework::New(
	EnvExportList	=> ["Option1", "Option2"],
	Option1			=> "blah",
	Option2			=> "moreBlah"
);

In this case the variables $Option1 and $Option2 would be set to their respective values in the process' shell.

There is a default list of variables and values to export at the top of the module, so if you need something permanent you can edit the module manually, if you want. Someday perhaps I'll build this into the Makefile.PL.

EnvExportOverride

If this option is set to a nonzero value, then no values will be exported to the process' shell environment.

AutoLoadUserCfg

If this option is set to a nonzero value and any of the files loaded via the Files option defines a user config, then the user's config is loaded from the appropriate location. (Default value for this option is 1).

GetSecure

If this option is set to a nonzero value, passwds.xml is decrypted and loaded into the configNamespace 'Secure'.

File

This is a file or list of files containing configuration data (in the Data::DumpXML DTD), which should be loaded at object creation. If there is only one file to be loaded, the argument should be a string containing the file name uneder v_root/config_loc to load. If there is more than one file to load, the argument should be a reference to an array containing the list of filenames. Each file to be loaded MUST DEFINE configNamespace.

AsymetricLogging

if set to a nonzero value, this option will background all of the logging processes with the shell '&'. This is good if you want to do lots of fast, repetative logging, but don't want to need to wait on disk updates. Of course the downside to this is that you can't garantee that your logfile will be in strict chronological order.

LoadConfig

Notes

Load the specified file into the object under the configNamespace defined in the file. If the given file does not define configNamespace, then it may be loaded under an optionally supplied configNamespace option. The specified file must be in the Data::DumpXML DTD.

Synopsis

$Object->LoadConfig([options]) || die $Object->{errstr}, "\n";

Options

File (required)

This is the filename of the file to load under v_root/config_loc.

configNamespace

If the specified file does not define it's own configNamespace, the file will be loaded under this alternate configNamespace, if supplied.

LoadXMLConfig

Notes

This routine loads an XML file in the Data::DumpXML DTD, and returns the data structure therein. All loading of Data::DumpXML DTD documents is done through this routine internally. It's also hanndy as an external routine.

Synopsis

$data = $Object->LoadXMLConfig([options]) || die $Object->{errstr}, "\n";

Options

File (required)

This is the filename of the file to load under v_root/config_loc. If the file is binary, we presume that it's encrypted, and attempt to decrypt it with the given Crypt and Key, if given. If they are not given, we attempt to use the default Crypt and Key defined for passwds.xml

Key

If the given File is binary, we will attempt to use this passphrase to decrypt the data.

Crypt

If the given file is binary, attempt to use this Crypt::CBC compliant encryption module to decrypt the data.

AltLoc

If you want to load a config file which dosen't exist under v_root/config_loc, you can define an aleternate path here.

LoadUserCfg

Notes

Load the user's config for the given configNamespace. If the user's config does not exist, and the given configNamespace defines a skeleton file, copy it to the user's home directory, and then load it. This presumes that the given configNamspace defines a userConfig. The user's config is loaded into specified child configNamespace under the given configNamespace.

Synopsis

$Object->LoadUserCfg([options]) || die $Object->{errstr}, "\n";

Options

configNamespace

Examine this already loaded namespace for a defined userConfig. If one is found, load it.

AlertAdmin

Notes

Send an alert to the address specified in 'admin', and to any additional specified addresses. This is the location where exceptions are handled for host programs. If the Debug option is set on the object, no email will be sent, but the message will be printed to stdout.

Synopsis

$Object->AlertAdmin([options]) || die $Object->{errstr}, "\n";

Options

Message (required)

A string containing the text of the message to send.

To

An optional list of email addresses to copy the message to. This should be an array reference. You may find it helpfull to build the array reference inline with the function call like so: To => ["address1", "address2", "address3"] Don't forget to escape @'s in your email addresses!

Die

If this option is set to a nonzero value, we call die with the given Message after sending the email and logging if defined.

Log

The name of a logfile to log the Message to.

Log

Notes

Log the given message to given logfile. All logging for applications should occur through this interface. At this time, the logging occurs through a primitive system ("echo $message >> $file"), so keep that in mind if you're going to use Message's with quotes in 'em. This will definately be modified in the near future to use a more advanced logging facility, probably Net::Syslog.

Options

Message (required)

A string containing the text of the message to log

Log (required)

The path and filename of the file to log the message to.

Author:

Andrew N. Hicox	<ahicox@uu.net>
http://www.hicox.com