man CGI::XMLApplication () - CGI::XMLApplication \-\- Object Oriented Interface for CGI Script Applications


CGI::XMLApplication \-\- Object Oriented Interface for CGI Script Applications


  use CGI::XMLApplication;

  $script = new CGI::XMLApplication;
  $script\->setStylesheetPath( "the/path/to/the/stylesheets" );

  # either this for simple scripts
  # or if you need more control ...
  $script\->run(%context_hash); # or a context object


CGI::XMLApplication is a CGI application class, that intends to enable perl artists to implement CGIs that make use of XML/XSLT functionality, without taking too much care about specialized errorchecking or even care too much about XML itself. It provides the power of the XML::LibXML/ XML::LibXSLT module package for content deliverment.

As well CGI::XMLApplication is designed to support project management on code level. The class allows to split web applications into several simple parts. Through this most of the code stays simple and easy to maintain. Throughout the whole runtime of a script CGI::XMLApplication tries to keep the application stable. As well a programmer has not to bother about some of XML::LibXML/ XML::LibXSLT transformation pitfalls.

The class module extends the CGI class. While all functionality of the original CGI package is still available, it should be not such a big problem, to port existing scripts to CGI::XMLApplication, although most functions used here are the access function for client data such as param().

CGI::XMLApplication, intended to be an application class should make writing of XML enabled CGI scripts more easy. Especially because of the use of object orientated concepts, this class enables much more transparent implemententations with complex functionality compared to what is possible with standard CGI-scripts.

The main difference with common perl CGI implementation is the fact, that the client-output is not done from perl functions, but generated by an internally build XML DOM that gets processed with an XSLT stylesheet. This fact helps to remove a lot of the HTML related functions from the core code, so a script may be much easier to read, since only application relevant code is visible, while layout related information is left out (commonly in an XSLT file).

This helps to write and test a complete application faster and less layout related. The design can be appended and customized later without effecting the application code anymore.

Since the class uses the OO paradigma, it does not force anybody to implement a real life application with the complete overhead of more or less redundant code. Since most CGI-scripts are waiting for events, which is usually the code abstraction of a click of a submit button or an image, CGI::XMLApplication implements a simple event system, that allows to keep event related code as separated as possible.

Therefore final application class is not ment to have a constructor anymore. All functionality should be encapsulated into implicit or explicit event handlers. Because of a lack in Perl's OO implementation the call of a superclass constructor before the current constructor call is not default behavior in Perl. For that reason I decided to have special events to enable the application to initialize correctly, excluding the danger of leaving important variables undefined. Also this forces the programmer to implement scripts more problem orientated, rather than class or content focused.

Another design aspect for CGI::XMLApplication is the strict differentiation between CODE and PRESENTATION. IMHO this, in fact being one of the major problems in traditional CGI programming. To implement this, the XML::LibXML and XML::LibXSLT modules are used by default but may be replaced easily by any other XML/XSLT capable modules. Each CGI Script should generate an XML-DOM, that can be processed with a given stylesheet.

Pay attention: In this Document XML-DOM means the DOM of XML::LibXML and not XML::DOM!

Programflow of a CGI::XMLApplication

The following Flowchart illustratrates how CGI::XMLApplication behaves during runtime. Also chart shows where specialized application code gets control during script runtime.

  ------- CGI Script ------->|<--------- CGI::XMLApplication --------
   .---------------------.    .--------------------.
   | app-class creation  |--- | event registration |
   `---------------------'    | registerEvents()   |
   .------------------------.            |
   | context initialization |------------'
   |     ( optional )       |
   .-----------------------.  .------------------------.
   | run() function called |--| application initialize |
   `-----------------------'  |      event_init()      |
                                / event parameter found? \_
                                \       testEvent()      / \
                                 `--------.,------------'   |
                                          |                 |
                                      yes |              no |
                                          |                 |
                               .------------.  .------------------.
                               | call event |  | call             |
                               |  event_*() |  |  event_default() |
                               `------------'  `------------------'
                                          |                |
                               .---------------------.     |
                               | application cleanup |-----'
                               |     event_exit()    |
                              _/ avoid XML serialization \
                             / \   skip_serialization()  /
                            |   `---------.,------------'
                            |             |
                        yes |          no |
                            |             |
                            |  .--------------------------.
                            |  | XML generation, XSLT     |
                            |  | serialization and output |
                            |  |     serialization()      |
                            |  `--------------------------'
    .---------------.       |             |
    |      END      |-------+-------------'

What are Events and how to catch them

Most CGI Scripts handle the result of HTML-Forms or similar requests from clients. Analouge to GUI Programming, CGI::XMLApplication calls this an event. Spoken in CGI/HTML-Form words, a CGI-Script handles the various situations a clients causes by pushing a submit button or follows a special link. Because of this common events are thrown by arguments found in the CGI's query string.

An event of CGI::XMLApplication has the same name as the input field, that should cause the event. The following example should illustrate this a little better:

    <!-- SOME HTML CODE -->
    <input type="submit" name="dummy" value="whatever" />
    <!-- SOME MORE HTML :) -->

If a user clicks the submitbutton and you have registered the event name dummy for your script, CGI::XMLApplication will try to call the function BIevent_dummy(). The script module to handle the dummy event would look something like the following code:

 # Application Module
 package myApp;

 use CGI::XMLApplication;
 @ISA = qw(CGI::XMLApplication);

 sub registerEvents { qw( dummy ); } # list of event names

 # ...

 sub event_dummy {
     my ( $self, $context ) = @_;

     # your event code goes here

     return 0;

During the lifecircle of a CGI script, often the implementation starts with ordinary submit buttons, which get often changed to so called input images, to fit into the UI of the Website. One does not need to change the code to make the scripts fit to these changes; CGI::XMLApplication already did it. The code has not to be changed if the presentation of the form changes. Therefore there is no need to declare separate events for input images. E.g. an event called evname makes CGI::XMLApplication tests if evname or evname.x exist in the querystring.

So a perl artist can implement and test his code without caring if the design crew have done their job, too ;-)

In many cases an web application is also confronted with events that can not be represented in with querystring arguments. For these cases CGI::XMLApplication offers the possibility to send special events from the BIevent_init() function for example in case of application errors. This is done with the BIsendEvent() Function. This will set a new parameter to the CGI's querystring after removing all other events. One can only send events that are already registred!.

Although a sendEvent function exists, CGI::XMLApplication doesn't implement an event queqe. For GUI programmers this seems like a unnessecary restriction. In terms of CGI it makes more sense to think of a script as a program, that is only able to scan its event queqe only once during runtime and stopped before the next event can be thrown. The only chance to stop the script from handling a certain event is to send a new event or delete this (or even all) events from inside the event_init() function. This function is always called at first from the run method. If another event uses the sendEvent function, the call will have no effect.

method registerEvents
This method is called by the class constructor \- namely CGI::XMLApplication's BInew() function . Each application should register the events it likes to handle with this function. It should return an array of eventnames such as eg. 'remove' or 'store'. This list is used to find which event a user caused on the client side.
method run
Being the main routine this should be the only method called by the script apart from the constructor. All events are handled inside the method BIrun(). Since this method is extremly simple and transparent to any kind of display type, there should be no need to override this function. One can pass a context hash or context object, to pass external or prefetched information to the application. This context will be available and acessable in all events and most extra functions. This function does all event and serialization related work. As well there is some validation done as well, so catched events, that are not implemented, will not cause any harm.

The Event System

A CGI::XMLApplication is split into two main parts: 1) The executable script called by the webserver and 2) the application module which has to be loaded, initialized and called by the script.

Commonly applications that make use of CGI::XMLApplication, will not bother about the run function too much. All functionality is kept inside event- and (pseudo-)callback functions. This forces one to implement much more strict code than common perl would allow. What first looks like a drawback, finally makes the code much easier to understand, maintain and finally to extend.

CGI::XMLApplication knows two types of event handlers: implicit events, common to all applications and explicit events, reflecting the application logic. The class assumes that implicit events are implemented in any case. Those events have reserved names and need not be specified through registerEvents. Since the class cannot know something about the application logic by itself, names of events have to be explicitly passed to be handled by the application. As well all event functions have to be implemented as member methods of the application class right now. Because of perls OO interface a class has to be written inside its own module.

An event may return a integer value. If the event succeeds (no fatal errors, e.g. database errors) the explicit or common event function should return a value greater or eqal than 0. If the value is less than 0, CGI::XMLApplication assumes an application panic, and will not try to generate a DOM or render it with a stylesheet.

There are 4 defined panic levels:

Stylesheet missing
Stylesheet not available
Event not implemented
Application panic

Apart from Application Panic the panic levels are set internally. An Application Panic should be set if the application catches an error, that does not allow any XML/XSLT processing. This can be for example, that a required perl module is not installed on the system.

To make it clear: If CGI::XMLApplication throws a panic, the application is broken, not completely implemented or stylesheets are missing or broken. Application panics are ment for debugging purposes and to avoid Internal Server Errors. They are not ment as a replacement of a propper error handling!

But how does CGI::XMLApplication know about the correct event handler?

One needs to register the names of the events the application handles. This is done by implmenting a registerEvents() function that simply returns an array of event names. Through this function one prepares the CGI::XMLApplication to catch the listed names as events from the query string the client browser sends back to the script. CGI::XMLApplication tries to call a event handler if a name of a registred event is found. The coresponding function-name of an event has to have the following format:


E.g. event_init handles the init event described below.

Each event has a single Parameter, the context. This can be an unblessed hash reference or an object, where the user can store whatever needed. This context is useful to pass scriptwide data between callbacks and event functions around. The callback is even available and useable if the script does not initialize the application context as earlier shown in the program flow chart.

If such a function is not implemented in the application module, CGI::XMLApplication sets the Event not implemented panic state.

All events have to return an integer that tells about their execution state as already described.

By default CGI::XMLApplication does not test for other events if it already found one. The most significant event is the first name of an event found in the query string \- all other names are simply ignored. One may change this behaviour by overriding the BItestEvent() function.

But still it is a good idea to choose the event names carefully and do not mix them with ordinary datafield names.

function testEvent
If it is nesseccary to check which event is relevant for the current script one can use this function to find out in event_init(). If this function returns undef, the default event is active, otherwise it returns the eventname as defined by registerEvents. In case one needs a special algorithm for event selection one can override this function. If done so, one can make use of the application context inside this function since it is passed to BItestEvent() by the BIrun() function.
method sendEvent SCALAR
Sometimes it could be neccessary to send an event by your own (the script's) initiative. A possible example could be if you don't have client input but path_info data, which determinates how the script should behave or session information is missing, so the client should not even get the default output. This can only be done during the event_init() method call. Some coders would prefer the constructor, which is not a very good idea in this case: While the constructor is running, the application is not completely initialized. This can be only ashured in the event_init function. Therefore all application specific errorhandling and initializing should be done there. sendEvent only can be called from event_init, because any CGI::XMLApplication will handle just one event, plus the init and the exit event. If sendEvent is called from another event than BIevent_init() it will take not effect. It is possible through sendEvent() to keep the script logic clean. Example:
  package myApp;
  use CGI::XMLApplication;
  @ISA = qw(CGI::XMLApplication);
  sub registerEvents { qw( missing ... ) ; }
  # event_init is an implicit event
  sub event_init {
     my ( $self, $context ) = @_;
     if ( not ( defined $self\->param( $paraname ) && length $self\->param( $paramname ) ) ){
        # the parameter is not correctly filled
        $self\->sendEvent( 'missing' );
     else {
    ... some more initialization ...
     return 0;
  ... more code ...
  # event_missing is an explicit event.
  sub event_missing {
     my ( $self , $context ) = @_;
     ... your error handling code goes ...
     return \-4 if $panic;  # just for illustration
     return 0;

Implicit Events

CGI::XMLApplication knows three implicit events which are more or less independent to client responses: They are 'init', 'exit', and 'default'. These events already exist for any CGI::XMLApplication. They need not to be implemented separatly if they make no sense for the application.

The init event is set before the CGI::XMLApplication tries to evaluate any of script parameters. Therefore the event_init method should be used to initialize the application.
The event_exit method is called after all other events have been processed, but just before the rendering is done. This should be used, if you need to do something independend from all events before the data is send to the user.
This event is called as a fallback mechanism if CGI::XMLApplication did not receive a stylesheet id by an other event handler; for example if no event matched.

the XML Serialization

The presentation is probably the main part of a CGI script. By using XML and XSLT this can be done in a standartised manner. From the application view all this can be isolated in a separate subsystem as well. In CGI::XMLApplication this subsystem is implemented inside the BIserialize() function.

For XML phobic perl programmers it should be cleared, that CGI::XMLApplication makes real use of XML/XSLT functionalty only inside this function. For all code explained above it is not required to make use of XML at all.

The XML serialization subsystem of CGI::XMLApplication tries to hide most of non application specific code from the application programmer.

This method renders the data stored in the DOM with the stylesheet returned by the event handler. You should override this function if you like to use a different way of displaying your data.

If the serialization should be skipped, CGI::XMLApplication will not print any headers. In such case the application is on its own to pass all the output.

The algorithm used by serialization is simple:

* request the appplication DOM through BIgetDOM()
* test for XML passthru
* get the stylesheet the application preferes through BIselectStylesheet()
* parse the stylesheet
* transform the DOM with the stylesheet
* set Content-Type and headers
* return the content to the client

If errors occour on a certain stage of serialization, the application is stopped and the generated error messages are returned.

CGI::XMLApplication provides four pseudo-callbacks, that are used to get the application specific information during serialization. In order of being called by CGI::XMLApplication::serialization() they are:

* getDOM
* setHttpHeader
* getStylesheet
* getXSLTParameter

In fact only getStylesheet has to be implemented. In most cases it will be a good idea to provide the getDOM function as well. The other functions provider a interface to make the CGI output more generic. For example one can set cookies or pass XSL parameters to XML::LibXSLT's xsl processor.

These methods are used by the serialization function, to create the content related datastructure. Like event functions these functions have to be implemented as class member, and like event funcitons the functions will have the context passed as the single parameter.

getDOM() should return the application data as XML-DOM. CGI::XMLApplication is quite lax if this function does not return anything \- its simply assumed that an empty DOM should be rendered. In this case a dummy root element is created to avoid error messages from XML::LibXSLT.
setHttpHeader should return a hash of headers (but not the Content-Type). This can be used to set the nocache pragma, to set or remove cookies. The keys of the hash must be the same as the named parameters of's header method. One does not need to care about the output of these headers, this is done by CGI::XMLApplication automaticly. The content type of the returned data is usually not required to be set this way, since the XSLT processor knows about the content type, too.
If the getStylesheet is implemented the CGI::XMLApplication will assume the returned value either as a filename of a stylesheet or as a XML DOM representation of the same. If Stylesheets are stored in a file accessable from the , one should set the common path for the stylesheets and let CGI::XMLApplication do the parsing job. In cases the stylesheet is already present as a string (e.g. as a result of a database query) one may pass this string directly to CGI::XMLApplication. selectStylesheet is an alias for getStylesheet left for compatibility reasons. If none of these stylesheet selectors succeeds the Stylesheet missing panic code is thrown. If the parsing of the stylesheet XML fails Stylesheet not available is thrown. The latter case will also give some informations where the stylesheet selection failed. BIselectStylesheet() has to return a valid path/filename for the stylesheet requested.
This function allows to pass a set of parameters to XML::LibXSLT's xsl processor. The function needs only to return a hash and does not need to encode the parameters. The function is the last callback called before the XSLT processing is done.

Flow Control

Besides the sendEvent() function does CGI::XMLApplication provide to other functions that allow to controll the flow of the application.

These two functions are related to the XML serialization and have not affect to the event handling.

Originally for debugging purposes CGI::XMLApplication supports the passthru argument in the CGI query string. It can be used to directly pass the stringified XML-DOM to the client. Since there are cases one needs to decide from within the application if an untransformed XML Document has to be returned, this function was introduced. If is called without parameters BIpassthru() returns the current passthru state of the application. E.g. this is done inside BIserialization(). Where TRUE (1) means the XML DOM should be passed directly to the client and FALSE (0) marks that the DOM must get XSL transformed first. Optional the function takes a single parameter, which shows if the function should be used in set rather than get mode. The parameter is interpreted as just described. If an application sets passthru by itself any external 'passthru' parameter will be lost. This is usefull if one likes to avoid, someone can fetch the plain (untransformed) XML Data.
To avoid the call of BIserialization() one should set skipSerialization.
   event_default {
      my $self = shift;
      # avoid serialization call
      $self\->skipSerialization( 1 ); # use 0 to unset
      # now you can directly print to the client, but don't forget the
      # headers.
      return 0;

Helperfunctions for internal use

function checkPush LIST
This function searches the query string for a parameter with the passed name. The implementation is imagesave meaning there is no change in the code needed, if you switch from input.type=submit to input.type=image or vv. The algorithm tests wheter a full name is found in the querystring, if not it tries tests for the name expanded by a '.x'. In context of events this function interprets each item part in the query string list as an event. Because of that, the algorithm returns only the first item matched. If you use the event interface on this function, make sure, the HTML-forms pass unique events to the script. This is neccessary to avoid confusing behaviour. This function is used by testEvent() so if it is required to change the way CGI::XMLApplication selects events, override that function.
method panic SCALAR
This a simple error message handler. By default this function will print some information to the client where the application failed. While development this is a useful feature on production system this may pass vunerable informations about the system to the outside. To change the default behaviour, one may write his own panic method or simply set $CGI::XMLApplication::Quiet to 1. The latter still causes the error page but does not send any error message. The current implementation send the 404 status to the client if any low level errors occour ( e.g. panic levels > \-4 aka Application Panic). Commonly this really shows a Not Found on the application Level. Application Panics will set the 500 error state. This makes this implementation work perfect with a mod_perl installation. In case mod_perl is used to handle the script one likes to set CGI::XMLApplication::Quiet to 2 which will cause CGI::XMLApplication just to return the error state while mod_perl does the rest. This useful method, helps to pass more specific error messages to the user. Currently this method is not very sophisticated: if the method is called twice, only the last string will be displayed.
function getPanicMsg
This method returns the panic message set by setPanicMsg().

CGI Extras

The following functions are some neat features missing in

function checkFields LIST
This is an easy way to test wether all required fields are filled out correctly. Called in array context the function returns the list of missing parameter. (Different to param() which returns all parameter names). In scalar context the function returns a boolean value.
function getParamHash LIST
This function is a bit better for general data processing as the standard CGI::Vars function. While Vars sets a keys for each parameter found in the query string, getFieldsAsHash returns only the requested fields (as long they aren't NULL). This is useful in scripts where the script itself handles different kind of data within the same event. Since the function relies on Vars the returned data has the same structure Vars returns.

some extra functions for stylesheet handling

The getStylesheet() function should return either a filename or a stringnyfied XSL-DOM. For the firstcase it can be a restriction to return the fully qualified path. The following functions allow to set the stylesheetpath systemwide.

method setStylesheetDir DIRNAME
alias for setStylesheetPath
method setStylesheetPath DIRNAME
This method is for telling the application where the stylesheets can be found. If you keep your stylesheets in the same directory as your script \-\- IMHO a bad idea \-\- you might leave this untouched.
function getStylesheetPath
This function is only relevant if you write your own BIserialization() method. It returns the current path to the application stylesheets.


CGI, perlobj, perlmod, XML::LibXML, XML::LibXSLT


Christian Glahn,