Using the Message Class---Data Transfer in Service Contracts
The Message class is fundamental to Windows Communication Foundation (WCF). All communication between clients and services ultimately results in Message instances being sent and received.
You would not usually interact with the Message class directly. Instead, WCF service model constructs, such as data contracts, message contracts, and operation contracts, are used to describe incoming and outgoing messages. However, in some advanced scenarios you can program using the Message class directly. For example, you might want to use theMessage class:
- When you need an alternative way of creating outgoing message contents (for example, creating a message directly from a file on disk) instead of serializing .NET Framework objects.
- When you need an alternative way of using incoming message contents (for example, when you want to apply an XSLT transformation to the raw XML contents) instead of deserializing into .NET Framework objects.
- When you need to deal with messages in a general way regardless of message contents (for example, when routing or forwarding messages when building a router, load-balancer, or a publish-subscribe system).
Before using the Message class, familiarize yourself with the WCF data transfer architecture in Data Transfer Architectural Overview.
A Message is a general-purpose container for data, but its design closely follows the design of a message in the SOAP protocol. Just like in SOAP, a message has both a message body and headers. The message body contains the actual payload data, while the headers contain additional named data containers. The rules for reading and writing the body and the headers are different, for example, the headers are always buffered in memory and may be accessed in any order any number of times, while the body may be read only once and may be streamed. Normally, when using SOAP, the message body is mapped to the SOAP body and the message headers are mapped to the SOAP headers.
Using the Message Class in Operations
You can use the Message class as an input parameter of an operation, the return value of an operation, or both. If Messageis used anywhere in an operation, the following restrictions apply:
- The operation cannot have any out or ref parameters.
- There cannot be more than one input parameter. If the parameter is present, it must be either Message or a message contract type.
- The return type must be either void, Message, or a message contract type.
The following code example contains a valid operation contract.
[ServiceContract] public interface IMyService { [OperationContract] Message GetData(); [OperationContract] void PutData(Message m); }
Creating Basic Messages
The Message class provides static CreateMessage factory methods that you can use to create basic messages.
All CreateMessage overloads take a version parameter of type MessageVersion that indicates the SOAP and WS-Addressing versions to use for the message. If you want to use the same protocol versions as the incoming message, you can use theIncomingMessageVersion property on the OperationContext instance obtained from the Current property. MostCreateMessage overloads also have a string parameter that indicates the SOAP action to use for the message. Version can be set to None to disable SOAP envelope generation; the message consists of only the body.
Creating Messages from Objects
The most basic CreateMessage overload that takes only a version and an action creates a message with an empty body. Another overload takes an additional Object parameter; this creates a message whose body is the serialized representation of the given object. Use the DataContractSerializer with default settings for serialization. If you want to use a different serializer, or you want the DataContractSerializer configured differently, use the CreateMessage overload that also takes an XmlObjectSerializer parameter.
For example, to return an object in a message, you can use the following code.
public class MyService1 : IMyService { public Message GetData() { Person p = new Person(); p.name = "John Doe"; p.age = 42; MessageVersion ver = OperationContext.Current.IncomingMessageVersion; return Message.CreateMessage(ver, "GetDataResponse", p); } public void PutData(Message m) { // Not implemented. } } [DataContract] public class Person { [DataMember] public string name; [DataMember] public int age; }
Creating Messages from XML Readers
There are CreateMessage overloads that take an XmlReader or an XmlDictionaryReader for the body instead of an object. In this case, the body of the message contains the XML that results from reading from the passed XML reader. For example, the following code returns a message with body contents read from an XML file.
public class MyService2 : IMyService { public Message GetData() { FileStream stream = new FileStream("myfile.xml",FileMode.Open); XmlDictionaryReader xdr = XmlDictionaryReader.CreateTextReader(stream, new XmlDictionaryReaderQuotas()); MessageVersion ver = OperationContext.Current.IncomingMessageVersion; return Message.CreateMessage(ver,"GetDataResponse",xdr); } public void PutData(Message m) { // Not implemented. } }
Additionally, there are CreateMessage overloads that take an XmlReader or an XmlDictionaryReader that represents the entire message and not just the body. These overloads also take an integer maxSizeOfHeaders parameter. Headers are always buffered into memory as soon as the message is created, and this parameter limits the amount of buffering that takes place. It is important to set this parameter to a safe value if the XML is coming from an untrusted source to mitigate the possibility of a denial of service attack. The SOAP and WS-Addressing versions of the message the XML reader represents must match the versions indicated using the version parameter.
Creating Messages with BodyWriter
One CreateMessage overload takes a BodyWriter instance to describe the body of the message. A BodyWriter is an abstract class that can be derived to customize how message bodies are created. You can create your own BodyWriterderived class to describe message bodies in a custom way. You must override the BodyWriter.OnWriteBodyContentsmethod that takes an XmlDictionaryWriter; this method is responsible for writing out the body.
Body writers can be buffered or non-buffered (streamed). Buffered body writers can write out their contents any number of times, while streamed ones can write out their contents only once. The IsBuffered property indicates whether a body writer is buffered or not. You can set this for your body writer by calling the protected BodyWriter constructor that takes anisBuffered boolean parameter. Body writers support creating a buffered body writer from a non-buffered body writer. You can override the OnCreateBufferedCopy method to customize this process. By default, an in-memory buffer that contains the XML returned by OnWriteBodyContents is used. OnCreateBufferedCopy takes a maxBufferSize integer parameter; if you override this method, you must not create buffers larger than this maximum size.
The BodyWriter class provides the WriteBodyContents and CreateBufferedCopy methods, which are essentially thin wrappers around OnWriteBodyContents and OnCreateBufferedCopy methods, respectively. These methods perform state checking to ensure that a non-buffered body writer is not accessed more than once. These methods are called directly only when creating custom Message derived classes based on BodyWriters.
Creating Fault Messages
You can use certain CreateMessage overloads to create SOAP fault messages. The most basic of these takes aMessageFault object that describes the fault. Other overloads are provided for convenience. The first such overload takes aFaultCode and a reason string and creates a MessageFault using MessageFault.CreateFault using this information. The other overload takes a detail object and also passes it to CreateFault together with the fault code and the reason. For example, the following operation returns a fault.
public class MyService3 : IMyService { public Message GetData() { FaultCode fc = new FaultCode("Receiver"); MessageVersion ver = OperationContext.Current.IncomingMessageVersion; return Message.CreateMessage(ver,fc,"Bad data","GetDataResponse"); } public void PutData(Message m) { // Not implemented. } }
Extracting Message Body Data
The Message class supports multiple ways of extracting information from its body. These can be classified into the following categories:
- Getting the entire message body written out at once to an XML writer. This is referred to as writing a message.
- Getting an XML reader over the message body. This enables you to later access the message body piece-by-piece as required. This is referred to as reading a message.
- The entire message, including its body, can be copied to an in-memory buffer of the MessageBuffer type. This is referred to as copying a message.
You can access the body of a Message only once, regardless of how it is accessed. A message object has a State property, which is initially set to Created. The three access methods described in the preceding list set the state to Written, Read, and Copied, respectively. Additionally, a Close method can set the state to Closed when the message body contents are no longer required. The message body can be accessed only in the Created state, and there is no way to go back to the Created state after the state has changed.
Writing Messages
The WriteBodyContents method writes out the body contents of a given Message instance to a given XML writer. TheWriteBody method does the same, except that it encloses the body contents in the appropriate wrapper element (for example, <soap:body>). Finally, WriteMessage writes out the entire message, including the wrapping SOAP envelope and the headers. If SOAP is turned off (Version is MessageVersion.None), all three methods do the same thing: they write out the message body contents.
For example, the following code writes out the body of an incoming message to a file.
public class MyService4 : IMyService { public void PutData(Message m) { FileStream stream = new FileStream("myfile.xml",FileMode.Create); XmlDictionaryWriter xdw = XmlDictionaryWriter.CreateTextWriter(stream); m.WriteBodyContents(xdw); xdw.Flush(); } public Message GetData() { throw new NotImplementedException(); } }
Two additional helper methods write out certain SOAP start element tags. These methods do not access the message body and so they do not change the message state. These include:
- WriteStartBody writes the start body element, for example, <soap:Body>.
- WriteStartEnvelope writes the start envelope element, for example, <soap:Envelope>.
To write the corresponding end element tags, call WriteEndElement on the corresponding XML writer. These methods are rarely called directly.
Reading Messages
The primary way to read a message body is to call GetReaderAtBodyContents. You get back an XmlDictionaryReader that you can use to read the message body. Note that the Message transitions to the Read state as soon asGetReaderAtBodyContents is called, and not when you use the returned XML reader.
The GetBody method also enables you to access the message body as a typed object. Internally, this method usesGetReaderAtBodyContents, and so it also transitions the message state to the Read state (see the State property).
It is good practice to check the IsEmpty property, in which case the message body is empty and GetReaderAtBodyContentsthrows an InvalidOperationException. Also, if it is a received message (for example, the reply), you may also want to checkIsFault, which indicates whether the message contains a fault.
The most basic overload of GetBody deserializes the message body into an instance of a type (indicated by the generic parameter) using a DataContractSerializer configured with the default settings and with the MaxItemsInObjectGraph quota disabled. If you want to use a different serialization engine, or configure the DataContractSerializer in a non-default way, use the GetBody overload that takes an XmlObjectSerializer.
For example, the following code extracts data from a message body that contains a serialized Person object and prints out the person’s name.
public class MyService5 : IMyService { public void PutData(Message m) { Person p = m.GetBody<Person>(); Console.WriteLine(p.name); } public Message GetData() { throw new NotImplementedException(); } } } namespace Samples2 { [ServiceContract] public interface IMyService { [OperationContract] Message GetData(); [OperationContract] void PutData(Message m); } [DataContract] public class Person { [DataMember] public string name; [DataMember] public int age; }
Copying a Message into a Buffer
Sometimes it is necessary to access the message body more than once, for example, to forward the same message to multiple destinations as part of a publisher-subscriber system. In this case, it is necessary to buffer the entire message (including the body) in memory. You can do this by calling CreateBufferedCopy. This method takes an integer parameter that represents the maximum buffer size, and creates a buffer not larger than this size. It is important to set this to a safe value if the message is coming from an untrusted source.
The buffer is returned as a MessageBuffer instance. You can access data in the buffer in several ways. The primary way is to call CreateMessage to create Message instances from the buffer.
Another way to access the data in the buffer is to implement the IXPathNavigable interface that the MessageBuffer class implements to access the underlying XML directly. Some CreateNavigator overloads allow you to create System.Xml.XPathnavigators protected by a node quota, limiting the number of XML nodes that can be visited. This helps prevent denial of service attacks based on lengthy processing time. This quote is disabled by default. Some CreateNavigator overloads allow you to specify how white space should be handled in the XML using the XmlSpace enumeration, with the default beingXmlSpace.None.
A final way to access the contents of a message buffer is to write out its contents to a stream using WriteMessage.
The following example demonstrates the process of working with a MessageBuffer: an incoming message is forwarded to multiple recipients, and then logged to a file. Without buffering, this is not possible, because the message body can then be accessed only once.
[ServiceContract] public class ForwardingService { private List<IOutputChannel> forwardingAddresses; [OperationContract] public void ForwardMessage (Message m) { //Copy the message to a buffer. MessageBuffer mb = m.CreateBufferedCopy(65536); //Forward to multiple recipients. foreach (IOutputChannel channel in forwardingAddresses) { Message copy = mb.CreateMessage(); channel.Send(copy); } //Log to a file. FileStream stream = new FileStream("log.xml",FileMode.Append); mb.WriteMessage(stream); stream.Flush(); } }
The MessageBuffer class has other members worth noting. The Close method can be called to free resources when the buffer contents are no longer required. The BufferSize property returns the size of the allocated buffer. TheMessageContentType property returns the MIME content type of the message.
Accessing the Message Body for Debugging
For debugging purposes, you can call the ToString method to get a representation of the message as a string. This representation generally matches the way a message would look on the wire if it were encoded with the text encoder, except that the XML would be better formatted for human readability. The one exception to this is the message body. The body can be read only once, and ToString does not change the message state. Therefore, the ToString method might not be able to access the body and might substitute a placeholder (for example, “…” or three dots) instead of the message body. Therefore, do not use ToString to log messages if the body content of the messages is important.
Accessing Other Message Parts
Various properties are provided to access information about the message other than its body contents. However, these cannot be called once the message has been closed:
- The Headers property represents the message headers. See the section on “Working with Headers” later in this topic.
- The Properties property represents the message properties, which are pieces of named data attached to the message that do not generally get emitted when the message is sent. See the section on "Working with Properties" later in this topic.
- The Version property indicates the SOAP and WS-Addressing version associated with the message, or None if SOAP is disabled.
- The IsFault property returns true if the message is a SOAP fault message.
- The IsEmpty property returns true if the message is empty.
You can use the GetBodyAttribute method to access a particular attribute on the body wrapper element (for example,<soap:Body>) identified by a particular name and namespace. If such an attribute is not found, null is returned. This method can be called only when the Message is in the Created state (when the message body has not yet been accessed).
Working with Headers
A Message can contain any number of named XML fragments, called headers. Each fragment normally maps to a SOAP header. Headers are accessed through the Headers property of type MessageHeaders. MessageHeaders is a collection ofMessageHeaderInfo objects, and individual headers can be accessed through its IEnumerable interface or through its indexer. For example, the following code lists the names of all the headers in a Message.
public class MyService6 : IMyService { public void PutData(Message m) { foreach (MessageHeaderInfo mhi in m.Headers) { Console.WriteLine(mhi.Name); } } public Message GetData() { throw new NotImplementedException(); } }
Adding, Removing, Finding Headers
You can add a new header at the end of all existing headers using the Add method. You can use the Insert method to insert a header at a particular index. Existing headers are shifted for the inserted item. Headers are ordered according to their index, and the first available index is 0. You can use the various CopyHeadersFrom method overloads to add headers from a different Message or MessageHeaders instance. Some overloads copy one individual header, while others copy all of them. The Clear method removes all headers. The RemoveAt method removes a header at a particular index (shifting all headers after it). The RemoveAll method removes all headers with a particular name and namespace.
Retrieve a particular header using the FindHeader method. This method takes the name and namespace of the header to find, and returns its index. If the header occurs more than once, an exception is thrown. If the header is not found, it returns -1.
In the SOAP header model, headers can have an Actor value that specifies the intended recipient of the header. The most basic FindHeader overload searches only headers intended for the ultimate receiver of the message. However, another overload enables you to specify which Actor values are included in the search. For more information, see the SOAP specification.
A CopyTo method is provided to copy headers from a MessageHeaders collection to an array of MessageHeaderInfoobjects.
To access the XML data in a header, you can call GetReaderAtHeader and return an XML reader for the specific header index. If you want to deserialize the header contents into an object, use GetHeader or one of the other overloads. The most basic overloads deserialize headers using the DataContractSerializer configured in the default way. If you want to use a different serializer or a different configuration of the DataContractSerializer, use one of the overloads that take anXmlObjectSerializer. There are also overloads that take the header name, namespace, and optionally a list of Actor values instead of an index; this is a combination of FindHeader and GetHeader.
Working with Properties
A Message instance can contain an arbitrary number of named objects of arbitrary types. This collection is accessed through the Properties property of type MessageProperties. The collection implements the IDictionary interface and acts as a mapping from String to Object. Normally, property values do not map directly to any part of the message on the wire, but rather provide various message processing hints to the various channels in the WCF channel stack or to the CopyTo service framework. For an example, see Data Transfer Architectural Overview.
Inheriting from the Message Class
If the built-in message types created using CreateMessage do not meet your requirements, create a class that derives from the Message class.
Defining the Message Body Contents
Three primary techniques exist for accessing data within a message body: writing, reading, and copying it to a buffer. These operations ultimately result in the OnWriteBodyContents, OnGetReaderAtBodyContents, and OnCreateBufferedCopy methods being called, respectively, on your derived class of Message. The base Message class guarantees that only one of these methods is called for each Message instance, and that it is not called more than once. The base class also ensures that the methods are not called on a closed message. There is no need to track the message state in your implementation.
OnWriteBodyContents is an abstract method and must be implemented. The most basic way to define the body contents of your message is to write using this method. For example, the following message contains 100,000 random numbers from 1 to 20.
public class RandomMessage : Message { override protected void OnWriteBodyContents(XmlDictionaryWriter writer) { Random r = new Random(); for (int i = 0; i <100000; i++) { writer.WriteStartElement("number"); writer.WriteValue(r.Next(1,20)); writer.WriteEndElement(); } } //code omitted
The OnGetReaderAtBodyContents and OnCreateBufferedCopy methods have default implementations that work for most cases. The default implementations call OnWriteBodyContents, buffer the results, and work with the resulting buffer. However, in some cases this may not be enough. In the preceding example, reading the message results in 100,000 XML elements being buffered, which might not be desirable. You might want to override OnGetReaderAtBodyContents to return a custom XmlDictionaryReader derived class that serves up random numbers. You can then override OnWriteBodyContentsto use the reader that the OnGetReaderAtBodyContents property returns, as shown in the following example.
public override MessageHeaders Headers { get { throw new Exception("The method or operation is not implemented."); } } public override MessageProperties Properties { get { throw new Exception("The method or operation is not implemented."); } } public override MessageVersion Version { get { throw new Exception("The method or operation is not implemented."); } } } public class RandomMessage2 : Message { override protected XmlDictionaryReader OnGetReaderAtBodyContents() { return new RandomNumbersXmlReader(); } override protected void OnWriteBodyContents(XmlDictionaryWriter writer) { XmlDictionaryReader xdr = OnGetReaderAtBodyContents(); writer.WriteNode(xdr, true); } public override MessageHeaders Headers { get { throw new Exception("The method or operation is not implemented."); } } public override MessageProperties Properties { get { throw new Exception("The method or operation is not implemented."); } } public override MessageVersion Version { get { throw new Exception("The method or operation is not implemented."); } } } public class RandomNumbersXmlReader : XmlDictionaryReader { //code to serve up 100000 random numbers in XML form omitted
Similarly, you might want to override OnCreateBufferedCopy to return your own MessageBuffer derived class.
In addition to providing message body contents, your message derived class must also override the Version, Headers, andProperties properties.
Note that if you create a copy of a message, the copy uses the message headers from the original.
Other Members that Can Be Overridden
You can override the OnWriteStartEnvelope, OnWriteStartHeaders, and OnWriteStartBody methods to specify how the SOAP envelope, SOAP headers, and SOAP body element start tags are written out. These normally correspond to<soap:Envelope>, <soap:Header>, and <soap:Body>. These methods should normally not write anything out if theVersion property returns MessageVersion.None.
 Note: |
|---|
| The default implementation of OnGetReaderAtBodyContents calls OnWriteStartEnvelope and OnWriteStartBodybefore calling OnWriteBodyContents and buffering the results. Headers are not written out.
|
Override the OnWriteMessage method to change the way the entire message is constructed from its various pieces. TheOnWriteMessage method is called from WriteMessage and from the default OnCreateBufferedCopy implementation. Note that overriding WriteMessage is not a best practice. It is better to override the appropriate On methods (for example,OnWriteStartEnvelope, OnWriteStartHeaders, and OnWriteBodyContents.
Override OnBodyToString to override how your message body is represented during debugging. The default is to represent it as three dots ("…"). Note that this method can be called multiple times when the message state is anything other than Closed. An implementation of this method should never cause any action that must be performed only once (such as reading from a forward-only stream).
Override the OnGetBodyAttribute method to allow access to attributes on the SOAP body element. This method can be called any number of times, but the Message base type guarantees that it is only called when the message is in the Created state. It is not required to check the state in an implementation. The default implementation always returns null, which indicates that there are no attributes on the body element.
If your Message object must do any special cleanup when the message body is no longer required, you can overrideOnClose. The default implementation does nothing.
The IsEmpty and IsFault properties can be overridden. By default, both return false.