Table Of Contents

Previous topic

Input Validation

Next topic

Implementing Transports and Protocols

This Page

Working with RPC Metadata

This section builds on User Manager section. If you haven’t done so, we recommended you to read it first.

In most of the real-world scenarios, an RPC request comes with additional baggage like authentication headers, routing history, and similar information. Spyne comes with rich mechanisms that lets you deal with both protocol and transport metadata.

At the protocol level, the input and the output of the rpc function itself are kept in ctx.in_object and ctx.out_object attributes of the spyne.MethodContext whereas the protocol metadata reside in ctx.in_header and ctx.out_header attributes.

You can set values to the header attributes in the function bodies or events. You just need to consider the order the events are fired, so that you don’t overwrite data.

If you want to use headers in a function, you must denote it either in the decorator or the spyne.service.ServiceBase child that you use to expose your functions.

A full example using most of the available metadata functionality is available here: https://github.com/plq/spyne/blob/master/examples/authenticate/server_soap.py

Protocol Headers

As said before, the protocol headers are available in ctx.in_header and ctx.out_header objects. You should set the ctx.out_header to the native value of the declared type.

Header objects are defined just like any other object:

class RequestHeader(ComplexModel):
    user_name = Mandatory.Unicode
    session_id = Mandatory.Unicode

They can be integrated to the rpc definition either by denoting it in the service definition:

class UserService(ServiceBase):
    __tns__ = 'spyne.examples.authentication'
    __in_header__ = RequestHeader

    @rpc(_returns=Preferences)
    def some_call(ctx):
        # (...)

Or in the decorator:

@rpc(_in_header=RequestHeader, _returns=Preferences)

It’s generally a better idea to set the header types in the ServiceBase child as it’s likely that all methods will use it. This will avoid cluttering the service definition with header declarations. The header declaration in the decorator will overwrite the one in the service definition.

Transport Headers

There is currently no general transport header api – transport-specific apis should be used for setting headers. The only transport that supports headers right now is Http, and you can use ctx.transport.resp_headers which is a dict where keys are field names and values are field values. Both must be str instances.

Exceptions

Here’s a sample custom public exception:

class PublicKeyError(Fault):
    __type_name__ = 'KeyError'
    __namespace__ = 'spyne.examples.authentication'

    def __init__(self, value):
        Fault.__init__(self,
                faultcode='Client.KeyError',
                faultstring='Value %r not found' % value
            )

Let’s modify the python dict to throw our own exception class:

class SpyneDict(dict):
    def __getitem__(self, key):
        try:
            return dict.__getitem__(self, key)
        except KeyError:
            raise PublicKeyError(key)

We can now modify the decorator to expose the exception this service can throw:

preferences_db = SpyneDict()

class UserService(ServiceBase):
    __tns__ = 'spyne.examples.authentication'
    __in_header__ = RequestHeader

    @rpc(_throws=PublicKeyError, _returns=Preferences)
    def get_preferences(ctx):
        retval = preferences_db[ctx.in_header.user_name]

        return retval

While this is not really necessary in the world of the dynamic languages, it’d still be nice to specify the exceptions your service can throw in the interface document. Plus, intefacing with your services will just feel more natural with languages like Java where exceptions are kept on a short leash.

What’s next?

With this document, you know most of what Spyne has to offer for application developers. You can refer to the Implementing Transports and Protocols section if you want to implement your own transports and protocols.

Otherwise, please refer to the rest of the documentation or the mailing list if you have further questions.