Parsing the pod¶

How does Pod::WSDL work? If you instantiate a Pod::WSDL object with the name of the module (or the path of the file, or an open filehandle) providing the web service like this

  my $pwsdl = new Pod::WSDL(source => 'My::Module', 
        location => 'http://my.services.location/on/the/web');

Pod::WSDL will try to find "My::Module" in @INC, open the file, parse it for WSDL directives and prepare the information for WSDL output. By calling

  $pwsdl->WSDL;

Pod::WSDL will output the WSDL document. That's it.

When using Pod::WSDL, the parser expects you to do the following:

• Put the pod directly above the subroutines which the web service's client is going to call. There may be whitespace between the pod and the sub declaration but nothing else.
• Use the "=begin"/"=end" respectively the "=for" directives according to standard pod: anything between "=begin WSDL" and "=end" will be treated as pod. Anything composing a paragraph together with "=for WSDL" will be treated as pod.

Any subroutine not preceded by WSDL pod will be left unmentioned. Any standard pod will be ignored (though, for an exception to this, see the section on own complex types below).

The individual instructions for Pod::WSDL always begin with a keyword, like "_RETURN" or "_DOC" or "_FAULT". After this different things may follow, according to the specific type of instruction. The instruction may take one or more lines - everything up to the next line beginning with a keyword or the end of the pod is belonging to the current instruction.

Describing Methods¶

How do we use Pod::WSDL? In describing a web service's method we have to say something about parameters, return values and faults. In addition you might want to add some documentation to these items and to the method itself.

Parameters

WSDL differentiates between in-, out- and inout-parameters, so we do that, too. A different matter is the question, if the client can do this too, but now we are talking about possibilities, not actualities.

The pod string describing a parameter has the structure

  (_IN|_OUT|_INOUT) NAME ($|@)TYPE DESCRIPTION

like

  _IN foo $string This is a foo

or

  _INOUT bar @bar An array of bars

You will easily guess what "_IN", "_OUT" and "_INOUT" stand for so we can move on. "NAME" is the name of your parameter. It does not have any real function (the order of the parameters being the only important thing) but it is nice to have it since in a WSDL document the parameters need to have names. So instead of having Pod::WSDL automatically generate cryptic names (it cannot do that right now) be nice to the client and use some sensible name. The "TYPE" of the parameters can be any of the xsd (schema) standard types (see [5]) or a type of your own creation. The "$" resp. "@" symbols tell Pod::WSDL and your client if it is a scalar or array parameter. Everything following the type up to the next instruction is treated as the parameter's documentation. If you call the constructor of Pod::WSDL with the argument "withDocumentation => 1", it will be added to the WSDL.

Return Values

Return values work like parameters but since in WSDL there is provision for only one return value (you have (in)out parameters, or can return arrays if that isn't enough), you do not need to give them a name. Pod::WSDL will automatically call them 'Return' in the WSDL document. So, the structure of "_RETURN" instructions is

  _RETURN ($|@)TYPE DESCRIPTION

as in

  _RETURN $string Returns a string

The pod for one method may only have one "_RETURN" instruction. If you don't specify a "_RETURN" instruction, Pod::WSDL will assume that you return void. Of course the perl subroutine still will return something, but your web service won't. To make this clear Pod::WSDL generates an empty response message for this.

If you want some method to be a one way operation (see [4], ch. 2.4.1), say so by using the instruction "_ONEWAY" in the pod. In this case no response message will be generated and a "_RETURN" instruction will be ignored.

Faults

SOAP faults are usually translated into exceptions in languages like Java. If you set up a web service using SOAP::Lite, SOAP will trap your dying program and generate a generic fault using the message of "die". It is also possible to access SOAP::Lite's SOAP::Fault directly if you want more control - but this is not our issue. If you want to use custom-made fault messages of your own, define them in "_FAULT" instructions, which look like this:

  _FAULT TYPE DESCRIPTION

An example could be the following:

  _FAULT My::Fault If anything goes wrong

Since you probably won't return an array of fault objects, you do not need to use the "($|@)" tokens. Just say that you return a fault, declare its type and add an optional description.

As with parameters (but in contrary to "_RETURN" instructions) you can declare as many "_FAULT" instructions as you like, providing for different exception types your method might throw.

Method Documentation

Method documentation is easily explained. Its structure is

  _DOC Here comes my documentation ...

That's it. Use several lines of documentation if you like. If you instantiate the Pod::WSDL object with the parameter "withDocumentation => 1", it will be written into the WSDL document.

Describing Modules - Using Own Complex Types¶

Quite often it will be the case that you have to use complex types as parameters or return values. One example of this we saw when talking about faults: you might want to create custom fault types (exceptions) of your own to fullfill special needs in the communication between web service and client. But of course you also might simply want to pass a complex parameter like a address object containing customer data to your application. WSDL provides the means to describe complex types borrowing the xsd schema syntax. Pod::WSDL makes use of this by allowing you to add WSDL pod to your own types. Assuming you have some own type like

  package My::Type;
  sub new {
    bless {
      foo => 'foo',
      bar => -1
    }, $_[0];
  }
  1;

simply describe the keys of your blessed hash like this.

  =begin WSDL
    _ATTR foo $string A foo
    _ATTR bar $integer And a bar
  =end WSDL

Put this pod anywhere within the package My::Type. Pod::WSDL will find it (if it is in @INC), parse it and integrate it into the WSDL document. The "_ATTR" instruction works exactly as the "_IN", "_OUT" and "_INOUT" instructions for methods (see above).

If you initialize the Pod::WSDL object using "withDocumentation => 1", Pod::WSDL will look for standard pod in the module, parse it using Pod::Text and put it into the WSDL document.

new¶

Instantiates a new Pod::WSDL.

Parameters

• source - Name of the source file, package of the source module or file handle on source file for which the WSDL shall be generated. This source must contain specialized Pod tags. So, if your source is '/some/directory/modules/Foo/Bar.pm' with package declaration 'Foo::Bar', source may be '/some/directory/modules/Foo/Bar.pm' or 'Foo::Bar' (in which case '/some/directory/modules' has to be in @INC) or an open file handle on the file. Right?
• location - Target namespace for the WSDL, usually the full URL of your webservice's proxy.
• pretty - Pretty print WSDL, if true. Otherwise the WSDL will come out in one line. The software generating the client stubs might not mind, but a person reading the WSDL will!
• withDocumentation - If true, put available documentation in the WSDL (see "Pod Syntax" above). For used own complex types ('modules') this will be the output of Pod::Text on these modules. The software generating the client stubs might give a damn, but a person reading the WSDL won't!

WSDL¶

Returns WSDL as string.

Parameters

• pretty - Pretty print WSDL, if true. Otherwise the WSDL will come out in one line. The software generating the client stubs might not mind, but a person reading the WSDL will!
• withDocumentation - If true, put available documentation in the WSDL (see "Pod Syntax" above). For used own complex types ('modules') this will be the output of Pod::Text on these modules. The software generating the client stubs might give a damn, but a person reading the WSDL won't!

addNamespace¶

Adds a namespace. Will be taken up in WSDL's definitions element.

Parameters

1.

URI of the namespace

2.

Declarator of the namespace

Describe Several Signatures for one Method¶

Of course, one subroutine declaration might take a lot of different sets of parameters. In Java or C++ you would have to have several methods with different signatures. In perl you fix this within the method. So why not put several WSDL pod blocks above the method so the web service's client can handle that.

Implement a Better Parsing of the pod¶

Right know, the pod is found using some rather complex regular expressions. This is evil and will certainly fail in some situations. So, an issue on top of the fixme list is to switch to regular parsing. I'm not sure if I can use Pod::Parser since I need the sub declaration outside the pod, too.

Handle Several Package Declarations in One File¶

So far, Pod::WSDL assumes a one to one relation between packages and files. If it meets several package declarations in one file, it will fail some way or the other. For most uses, one package in one file will presumably suffice, but it would be nice to be able to handle the other cases, too.

Handle Array based blessed References¶

Array based blessed references used for complex types are something of a problem.

Get Information on Complex Types from Somewhere Else¶

If you use complex types for parameters that are not your own (we assume, that the module containing the web service always is your own), you might not be able to put the WSDL pod into the module files. So why not fetch it from somewhere else like a configuration file?

Integrate Pod::WSDL with SOAP::Lite¶

With Axis, you simply call the web service's URL with the parameter '?wsdl' and you get the WSDL document. It would be nice to be able to do this with SOAP::Lite, too.

Implement Non RPC Style Messages¶

Pod::WSDL writes WSDL documents in encoded RPC style. It should be able to generate literal RPC and document styles, too.