SOAP II: Data Encoding

Marlon Pierce, Bryan Carpenter, Geoffrey Fox

Community Grids Lab
Indiana University
[email protected]
http://www.grid2004.org/spring2004
Review: SOAP Message Payloads
• SOAP has a very simple structure:
– Envelopes wrap body and optional header elements.
• SOAP body elements may contain any sort of XML
– Literally, use <any> wildcard to include other XML.
• SOAP does not provide specific encoding restrictions.
• Instead, provides conventions that you can follow for
different message styles.
– RPC is a common convention.
• Remember: SOAP designers were trying to design it to
be general purpose.
– SOAP encoding and data models are optional
SOAP Data Models
 SOAP’s Abstract Data Model
• SOAP data may be optional
represented using Node-Edge
Graphs.
• Edges connect nodes
– Have a direction Node 1
– An edge is labeled with an
XML QName.
• A node may have 0 or more Edge
inbound and outbound edges.
• Implicitly, Node 2 describes
Node 1.
• A few other notes:
– Nodes may point to
themselves. Node 3 Node 2
– Nodes may have inbound
edges originating from more
than one Node.
 Nodes and Values
• Nodes have values.
– Values may be either simple Node 1
(lexical) or compound.
• A simple value may be (for
example) a string.
– It has no outgoing edges
• A complex value is a node with Node 2
both inbound and outbound (complex)
edges.
• For example, Node 1 has a
value, Node 2, that is
structured.
• Complex values may be either Node 3 Node 4
structs or arrays.
Complex Types: Structs and Arrays
• A compound value is a graph node with zero or
more outbound edges.
• Outbound edges may be distinguished by either
labels or by position.
• Nodes may be one of two sorts:
– Struct: all outbound edges are distinguished solely by
labels.
– Array: all outbound edges are distinguished solely by
position order.
• Obviously we are zeroing in on programming
language data structures.
 Abstract Data Models
• The SOAP Data Model is an abstract model
– Directed, labeled graph
• It will be expressed in XML.
• The graph model implies semantics about data
structures that are not in the XML itself.
– XML describes only syntax.
• Implicitly, nodes in the graph model resemble
nouns, while the edges represent predicates.
• We will revisit this in later lectures on the
Semantic Web.
 Graphs to XML
• SOAP nodes and edges are not readily apparent
in simple XML encoding rules.
– Normally, an XML element in the SOAP body acts as
both the edge and the node of the abstract model.
• However, SOAP does have an internal
referencing system.
– Use it when pointing from one element to another.
– Here, the XML-to-graph correspondence is more
obvious.
SOAP Encoding
 Intro: Encoding Conventions
• SOAP header and body tags <xs:element name="Body"
can be used to contain type="tns:Body" />
arbitrary XML <xs:complexType name="Body">
– Specifically, they can contain <xs:sequence>
an arbitrary sequence of tags,
replacing the <any> tag. <xs:any
– These tags from other namespace="##any"
schemas can contain child processContents="lax"
tags and be quite complex. minOccurs="0“
– See body definition on the maxOccurs="unbounded"
right. />
• And that’s all it specifies.   </xs:sequence>
• SOAP thus does not impose a   <xs:anyAttribute
content model. namespace="##other"
• Content models are defined by processContents="lax" />
convention and are optional. </xs:complexType>
 Encoding Overview
• Data models such as the SOAP graph model are abstract.
– Represented as graphs.
• For transfer between client and server in a SOAP message, we
encode them in XML.
• We typically should provide encoding rules along with the message
so that the recipient knows how to process.
• SOAP provides some encoding rule definitions:
– http://schemas.xmlsoap.org/soap/encoding/
– But these rules are not required and must be explicitly included.
– Note this is NOT part of the SOAP message schema.
• Terminology:
– Serialization: transforming a model instance into an XML instance.
– Deserialization: transforming the XML back to the model.
 Specifying Encoding
• Encoding is specified using the <soapenv:Body>
encodingStyle attribute. <ns1:echo
– This is optional soapenv:encodingStyle="http://
– There may be no encoding style schemas.xmlsoap.org/soap/enc
• This attribute can appear in the oding/"
envelope, body, or headers. xmlns:ns1=“…">
– The example from previous <!--
lecture puts it in the body.
The rest of the payload
– The value is the standard SOAP
encoding rules. -->
• Thus, each part may use different </soapenv:Body>
encoding rules.
– If present, the envelope has the
default value for the message.
– Headers and body elements may
override this within their scope.
 Encoding Simple Values
• Our echo service exchanges strings. The actual
message is encoded like this:
– <in0 xsi:type="xsd:string">Hello World</in0>
• xsi:type means that <in0> will take string values.
– And string means explicitly xsd:string, or string from
the XML schema itself.
• In general, all encoded elements should provide
xsi:type elements to help the recipient decode
the message.
 Simple Type Encoding Examples
Java examples SOAP Encoding
• int a=3; • <a xsi:type=“xsd:int”>
10
• float pi=3.14
</a>
• String s=“Hello”; • <pi xsi:type=“xsd:float”>
3.14
</pi>
• <s xsi:type=“xsd:string”>
Hello
</s>
 Explanation of Simple Type
Encoding
• The XML snippets have two namespaces (would
be specified in the SOAP envelope typically).
– xsd: the XML schema. Provides definitions of
common simple types like floats, ints, and strings.
– xsi: the XML Schema Instance. Provides the
definition of the type element and its possible values.
• Basic rule: each element must be given a type
and a value.
– Types come from XSI, values from XSD.
• In general, all SOAP encoded values must have
a type.
 XML Schema Instance
• A very simple supplemental XML schema that
provides only four attribute definitions.
• Type is used when an element needs to
explicitly define its type rather than implicitly,
through a schema.
– The value of xsi:type is a qualified name.
• This is needed when the schema may not be
available (in case of SOAP).
– May also be needed in schema inheritance
• See earlier XML schema lectures on “Polymorphism”
 Example for Encoding Arrays in
SOAP 1.1
• Java Arrays
– int[3] myArray={23,10,32};

• Possible SOAP 1.1 Encoding:

<myArray xsi:type=“SOAP-ENC:Array
SOAP-ENC:arrayType=“xsd:int[3]”>
<v1>21</v1>
<v2>10</v2>
<v3>32</v3>
</myArray>
 An Explanation
• We started out as before, mapping the Java
array name to an element and defining an
xsi:type.
• But there is no array in the XML schema data
definitions.
– XSD doesn’t preclude it, but it is a complex type to be
defined elsewhere.
– The SOAP encoding schema defines it.
• We also made use of the SOAP encoding
schema’s arrayType attribute to specify the type
of array (3 integers).
• We then provide the values.
Encoding a Java Class in SOAP
• Note first that a general Java class (like a Vector
or BufferedReader) does not serialize in XML.
• But JavaBeans (or if you prefer, Java data
objects) do serialize.
– A bean is a class with accessor (get/set) methods
associated with each of its data types.
– Can be mapped to C structs.
• XML Beans and Castor are two popular Java-to-
XML converters.
Example of Encoding a Java Bean
• Java class
class MyBean {
String Name=“Marlon”;
public String getName() {return Name;}
public void setName(String n) {Name=n;}
}
• Possible SOAP Encoding of the data (as a
struct)
<MyBean>
<name xsi:type=“xsd:string”>Marlon</name>
</MyBean>
 Structs
• Structs are defined in the <xs:element name="Struct"
type="tns:Struct" />
SOAP Encoding schema <xs:group name="Struct">
as shown. <xs:sequence>
  <xs:any namespace="##any"
• Really, they just are used minOccurs="0"
to hold yet more maxOccurs="unbounded"
processContents="lax" />
sequences of arbitrary   </xs:sequence>
XML. </xs:group>
• Struct elements are <xs:complexType name="Struct">
intended to be accessed   <xs:group ref="tns:Struct"
minOccurs="0" />
by name   <xs:attributeGroup
– Rather than order, as ref="tns:commonAttributes" />
Arrays. </xs:complexType>
 SOAP 1.1 Arrays
• As stated several times, <xs:complexType
SOAP encoding includes name="Array">
rules for expressing <xs:group ref="tns:Array"
arrays. minOccurs="0" />
• These were significantly   <xs:attributeGroup
revised between SOAP ref="tns:arrayAttribut
1.1 and SOAP 1.2. es" />
• You will still see both   <xs:attributeGroup
styles, so I’ll cover both. ref="tns:commonAttri
• The basic array type butes" />
(shown) was intended to </xs:complexType>
hold 0 or 1 Array groups.
 SOAP 1.1 Array Group
• Array elements contain <xs:group name="Array">
zero or more array <xs:sequence>
groups.
  <xs:any
• The array group in turn is namespace="##any"
a sequence of <any> minOccurs="0"
tags. maxOccurs="unbounded"
• So the array group can processContents="lax" />
hold arbitrary XML.   </xs:sequence>
</xs:group>
 SOAP 1.1 Array Attributes
• The array group itself is just for <xs:attributeGroup
holding arbitrary XML. name="arrayAttributes">
• The array attributes are used to   <xs:attribute ref="tns:arrayType" />
further refine our definition.   <xs:attribute ref="tns:offset" />
• The array definition may provide   </xs:attributeGroup>
an arrayType definition and an <xs:attribute name="offset"
offset.
type="tns:arrayCoordinate" />
• Offsets can be used to send
partial arrays.
• According to the SOAP Encoding <xs:attribute name="arrayType"
schema itself, these are only type="xs:string" />
required to be strings.
<xs:simpleType
name="arrayCoordinate">
<xs:restriction base="xs:string" />
</xs:simpleType>
Specifying Array Sizes in SOAP 1.1
• The arrayType specifies only that the it takes a string
value.
• The SOAP specification (part 2) does provide the rules.
• First, it should have the form enc:arraySize.
– Encoding can be an XSD type, but not necessarily.
– Ex: xsd:int[5], xsd:string[2,3], p:Person[5]
– The last is an array of five persons, defined in p.
• Second, use the following notation:
– [] is a 1D array.
– [][] is a array of 1D arrays
– [,] is a 2D array.
– And so on.
 Encoding Arrays in SOAP 1.2
• Array encodings have been <xs:attribute name="arraySize"
revised and simplified in the type="tns:arraySize" />
latest SOAP specifications. <xs:attribute name="itemType"
– http://www.w3.org/2003/05/so type="xs:QName" />
ap-encoding
• ArrayType elements are
derived from a generic <xs:attributeGroup
nodeType element. name="arrayAttributes">
• Now arrays have two attributes   <xs:attribute
– itemType is the the type of the
ref="tns:arraySize" />
array (String, int, XML   <xs:attribute
complex type). ref="tns:itemType" />
– arraySize </xs:attributeGroup>
 SOAP 1.2 Array Sizes
• The arraySize attribute (shown below). The
regular expression means
– I can use a “*” for an unspecified size, OR
– I can specify the size with a range of digits
– I may include multiple groupings of digits for multi-
dimensional arrays, with digit groups separated by
white spaces.

<xs:simpleType name="arraySize">
<xs:restriction base="tns:arraySizeBase">
  <xs:pattern value="(\*|(\d+))(\s+\d+)*" />
  </xs:restriction>
</xs:simpleType>
 Comparison of 1.1 and 1.2 Arrays
<numbers <numbers
enc:arrayType="xs:int[2]"> enc:itemType="xs:int"
enc:arraySize="2">
<number>3
<number>3
</number> <number>4
</number> <number>4
</number>
</number>
</numbers>
</numbers>
 SOAP 1.1 Encoding’s Common
Attributes
• As we have seen, both <xs:attributeGroup
structs and arrays contain name="commonAttributes
a group called ">
commonAttributes.   <xs:attribute name="id"
• The definition is shown at type="xs:ID" />
the right.   <xs:attribute name="href"
• The ID and the HREF type="xs:anyURI" />
attributes are used to   <xs:anyAttribute
make internal references namespace="##other"
within the SOAP processContents="lax" />
message payload. </xs:attributeGroup>
 References and IDs
• As you know, XML provides a simple tree model for
data.
• While you can convert many data models into trees, it
will lead to redundancy.
• The problem is that data models are graphs, which may
be more complicated than simple trees.
• Consider a typical manager/employee data model.
– Managers are an extension of the more general employee class.
– Assume in following example we have defined an appropriate
schema.
 Before/After Referencing
(SOAP 1.1 Encoding)
<manager> <manager id=“GCF”>
<fname>Geoffrey</> <fname>Geoffrey</>
<lname>Fox</> <lname>Fox</>
</manager> </manager>
<employee> <employee>
<fname>Marlon</> <fname>Marlon</>
<lname>Pierce</> <lname>Pierce</>
<manager> <manager href=“#gcf”>
<fname>Geoffrey</> </employee>
<lname>Fox</>
</manager>
</employee>
 References, IDs and Graphs
• References serve two
purposes.
– They save space by employee
avoiding duplication
• A good thing in a
message. href=“#gcf”
– They lower the potential for
errors.
• They also return us to the
graph model.
– Normal nodes and edges manager
get mapped into one
element information item.
– Ref nodes actually split the
edge and node.
 References in SOAP 1.2
• SOAP 1.1 required all <manager id=“GCF”>
references to point to <fname>Geoffrey</>
other top level elements.
<lname>Fox</>
• SOPA 1.2 changed this,
</manager>
so now refs can point to
child elements in a graph <employee>
as well as top level <fname>Marlon</>
elements. <lname>Pierce</>
– See next figure
<manager ref=“gcf”>
• They also changed the
</employee>
tag names and values, so
the encoding looks
slightly different.
 SOAP 1.1 and 1.2 Refs
<e:Books> <e:Books>
<e:Book> <e:Book>
<title>My Life and Work </title> <title>My Life and Work </title>
<author href="#henryford" /> <author id="henryford" >
</e:Book> <name>Henry Ford</name>
<e:Book> </author>
<title>Today and Tomorrow</title> </e:Book>
<author href="#henryford" /> <e:Book>
</e:Book> <title>Today and Tomorrow
</e:Books> </title>
<author ref="henryford" />
<author id="henryford"> </e:Book>
<name>Henry Ford</name> </e:Books>
</author>
Using SOAP for Remote
Procedure Calls
 The Story So Far…
• We have defined a general purpose abstract
data model.
• We have looked at SOAP encoding.
– SOAP does not provide standard encoding rules, but
instead provides a pluggable encoding style attribute.
• We examined a specific set of encoding rules
that may be optionally used.
• We are now ready to look at a special case of
SOAP encodings suitable for remote procedure
calls (RPC).
 Requirements for RPC with SOAP
• RPC is just a way to • Information needed for
invoke a remote RPC:
operation and get some – Location of service
data back. – The method name
– All of your Web Service – The method values
examples use RPC • The values must be
• How do we do this with associated with the
SOAP? We encode method’s argument
carefully to avoid names.
ambiguity.
• But it really is just
common sense.
 Location of the Service
• Obviously the SOAP message needs to get
sent to the right place.
• The location (URL) of the service is not
actually encoded in SOAP.
• Instead, it is part of the transport protocol used
to carry the SOAP message.
• For SOAP over HTTP, this is part of the HTTP
Header:
POST /axis/service/echo HTTP/1.0
Host: www.myservice.com
 RPC Invocation
• Consider the remote invocation of the following
Java method:
– public String echoService(String toEcho);
• RPC invocation conventions are the following:
– The invocation is represented by a single struct.
– The struct is named after the operation (echoService).
– The struct has an outbound edge for each transmitted
parameter.
– Each transmitted parameter is an outbound edge with
a label corresponding to the parameter name.
 SOAP Message by Hand
<env:Envelope xmlns:env=“…” xmlns:xsd=“…”
xmlns:xsi=“…”
env:encodingStyle=“…”>
<env:Body>
<e:echoService xmlns:e=“…”>
<e:toEcho xsi:type=“xsd:string”>Hello
</e:toEcho>
</e:echoService>
</env:Body>
</env:Envelope>
 Notes
• I have omitted the namespace URIs, but you
should know that they are the SOAP, XML, and
XSI schemas.
• I also omitted the encoding style URI, but it is
the SOAP encoding schema.
– Required by RPC convention.
• I assume there is a namespace (e:) that defines
all of the operation and parameter elements.
• The body follows the simple rules:
– One struct, named after the method.
– One child element for each input parameter.
 RPC Responses
• These follow similar rules as requests.
– We need one (and only one) struct for the remote
operation.
– This time, the label of the struct is not important.
– This struct has one child element (edge) for each
argument.
– The child elements are labeled to correspond to the
operational parameters.
• The response may also distinguish the “return”
value.
 RPC Return Values
• Often in RPC we need to distinguish one of the
output values as the “return value”.
– Legacy of C and other programming languages.
• We do this by labeling the return type like this:
<rpc:result>ex:myReturn</rpc:result>
<ex:myReturn xsi:type=“xsd:int”>0</>
• The rpc namespace is
– http://www.w3c.org/2003/05/soap-rpc
 An RPC Response
<env:Envelope xmlns:env=“…” xmlns:xsd=“…”
xmlns:xsi=“…” env:encodingStyle=“…”>
<env:Body>
<e:echoResponse
xmlns:rpc=“…”
xmlns:e=“…”>
<rpc:result>e:echoReturn</rpc:result>
<e:echoReturn xsi:type=“xsd:string”>
Hello
</e:echoReturn>
</e:echoResponse>
</env:Body>
</env:Envelope>
 Going Beyond Simple Types
• Our simple example just communicates in single
strings.
• But it is straightforward to write SOAP encodings
for remote procedures that use
– Single simple type arguments of other types (ints,
floats, and so on).
– Arrays
– Data objects (structs)
– Multiple arguments, both simple and compound.
 Discovering the Descriptions for
RPC
• The RPC encoding rules are based on
some big assumptions:
– You know the location of the service.
– You know the names of the operations.
– You know the parameter names and types of
each operation.
• How you learn this is out of SOAP’s
scope.
• WSDL is one obvious way.
 Relation to WSDL Bindings
• Recall from last WSDL lecture that the
<binding> element binds WSDL portTypes
to SOAP or other message formats.
• Binding to SOAP specified the following:
– RPC or Document Style
– HTTP for transport
– SOAP encoding for the body elements
 The WSDL Binding for Echo
<wsdl:binding name="EchoSoapBinding" type="impl:Echo">
  <wsdlsoap:binding style="rpc"
transport="http://schemas.xmlsoap.org/soap/http" />
<wsdl:operation name="echo">
  <wsdlsoap:operation soapAction="" />
<wsdl:input name="echoRequest">
  <wsdlsoap:body
encodingStyle="http://schemas.xmlsoap.org/so ap/encoding/"
namespace=“…" use="encoded" />
  </wsdl:input>
<wsdl:output name="echoResponse">
  <wsdlsoap:body
encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
namespace=“…" use="encoded" />
  </wsdl:output>
</wsdl:operation>
</wsdl:binding>
 RPC Style for Body Elements
• The body element just contains <soapenv:Body>
XML. <ns1:echo
• Our WSDL specified RPC style soapenv:encodingStyle=""
encoding. xmlns:ns1="">
– So we will structure our body <in0 xsi:type="xsd:string">
element to look like the WSDL
method. Hello World
• First, the body contains an </in0>
element <echo> that corresponds </ns1:echo>
to the remote comnand. </soapenv:Body>
– Using namespace ns1 to connect
<echo> to its WSDL definition
• Then the tag contains the element
<in0> which contains the payload.
 Connection of WSDL Definitions
and SOAP Message for RPC
<wsdl:message <soapenv:Body>
name="echoRequest"> <ns1:echo
<wsdl:part name="in0" soapenv:encodingStyle=""
type="xsd:string" /> xmlns:ns1="">
</wsdl:message> <in0 xsi:type="xsd:string">
Hello World
<wsdl:portType name="Echo"> </in0>
<wsdl:operation name="echo" </ns1:echo>
parameterOrder="in0”>
</soapenv:Body>
<wsdl:input
message="impl:echoRequest
name="echoRequest" />
</wsdl:operation>
</wsdl:portType>
 WSDL-RPC Mappings for
Response
<wsdl:portType name="Echo"> <soapenv:Body>
<wsdl:operation name="echo" <ns1:echoResponse
parameterOrder="in0"> env:encodingStyle=“…”
… xmlns:ns1=“…">
<wsdl:output
message="echoResponse” <echoReturn xsi:type=“String“>
Hello World
name="echoResponse" /> </echoReturn>
</wsdl:operation> </ns1:echoResponse>
</wsdl:portType> </soapenv:Body>

<wsdl:message
name="echoResponse">
  <wsdl:part name="echoReturn"
type="xsd:string" />
</wsdl:message>
Alternative Encoding
Schemes
 Wrap Up
• As we have seen, SOAP itself does not provide
encoding rules for message payloads.
– Instead, it provides a pluggable encoding style
attribute.
• SOAP encoding rules are optional, but likely to
be commonly supported in software like Axis.
• SOAP encoding’s three main parts for RPC:
– Abstract Data Model
– XML Encoding of model
– Further conventions for RPC
• What about other encodings?
 Alternative Encoding Schemes
• SOAP encoding uses graph models for data but,
apart from references, does not explicitly map
the parts of the graph to different XML elements.
• There are other XML data encoding schemes
that make a much more explicit connection
between the graph and the encoding.
• The Resource Description Framework is one
such scheme.
• So we may choose to use RDF instead of SOAP
encoding in a SOAP message.
RDF Encoding Example of Echo
<?xml version=‘1.0’ ?>
<env:Envelope xmlns:env=“…”>
<env:Body
env:encodingStyle=“http://www.w3c.org/1999/02/22-rdf-syntax-
ns#”>
<rdf:RDF>
<rdf:Description about=“echo service uri”>
<e:echoService>
<e:in0>Hello</e:in0>
</e:echoService>
</rdf:Description>
</rdf:RDF>
</env:Body>
</env:Envelope>
 RDF Encoding Notes
• We will look at RDF in detail in next week’s
lectures.
• Basic idea is that <rdf:Description> tags are
envelopes for xml tags from other schemas.
• The <Description>’s about attribute tells you
what is being described.
• Note that standard Web Service engines do not
support RDF or other encodings.
– You would need to extend it yourself.
– But it is possible.