XMLFoundation 2009


Source Download

Support Thread

Nearly every C++
application can benefit from the XMLFoundation (even if it does not use XML).

21. Conclusion                         

XML in the Foundation

As the name suggests it provides a foundation for XML
support in an application, however this is much more than just another XML
parser.  It applies a unique approach to handling XML that allows your
application code to focus on the application rather than traversing DOM or
subscribing to SAX events.  The most unique feature of the
XMLFoundation is the object oriented encapsulation that provides XML support in
the application layer.  XMLFoundation allows you to easily integrate XML
with your GUI, or with your server objects,  and it natively supports COM, DCOM,
and CORBA objects.

XMLFoundation contains a small, fast, and portable XML
tokenizer that has been refined and optimized in many large software projects
for about 10 years.  My involvement with XML
pre-dates the finalization of the XML 1.0 recommendation by W3C.  For years,
the only XML Parser that could match XMLFoundation tokenization performance was
"Xpat" by James Clark – but as you will see the unique ability to
bypass DOM and SAX altogether makes XMLFoundation the fastest solution available for
moving XML to and from application layer objects – and it requires far less
lines of code to do it.  

The performance of the stack based XML parser is at the top
of its class for non-validating parsers.  Parsing and tokenization is only
half the task, the other half is getting the results into the member variables,
lists, and objects that they need to be in to be useful in the application layer
– it is in that task that XMLFoundation is in a class of its own.  The
performance is unparalleled because the memory buffer that contains the source
XML parses directly into your custom class objects without ever being copied or
temporarily stored in a DOM tree.  It parses directly into your lists,
objects, arrays, indexed data structures, and all native C++ data types. 
It even has support for common containers of element data such as MFC CStrings. 
It’s been used in Java too. 
That said, speed of execution is less impressive than the speed of development
and overall reduction in lines of code required to effectively use XML in your


The Solid Foundation 

If you are building an application that does not use XML
and never will….. XMLFoundation is still a very valuable tool available to
solve many very common development tasks.  The data structure classes alone
( List, Hash, Stack, Tree, Array, QSort ) are very useful.  They
all have "Iterator"
objects so that data structures can be
read-referenced by multiple threads at the same time without blocking.  The interface is
standard to all data structures.  If you
find MFC or Rogue Wave Standard C++ library data structures useful, you will likely find XMLFoundation data
structures even more so. 

XMLFoundation also has standard algorithm implementations
( Encryption, Compression, Data hash, Encoding ).  These are based on the
works of other authors.  They have been included into the XMLFoundation in
a simplified build format.  They all compile under C++, so if you are using
them on AS/400, AIX, Solaris, Linux, or other like platforms – you do not even
need to reference a C compiler from the makefile, only your C++
compiler.  They are also organized into single .CPP files for each
implementation – often a consolidation of many individual C source files in the
original authors publications.

XMLFoundation also has a plethora of application utilities
including ( Sorts, Performance Timers, Disk Directory, Exceptions, 
INI Profiles,
Caching, String, Stream
).  XMLFoundation has many utilities that MFC does not. 
They are complete, comment documented with examples, and thoroughly tested on
many software projects.

XMLFoundation is
very portable
.  It builds on all versions of Windows (Win95 through
Windows7 and Windows Mobile).  Portions were initially developed on a RISC machine, and
it was used in Solaris and Linux as early as 2001.  Some of the compilers that have been used to build XMLFoundation
include: CC5.0, Xlc, gcc, IntelC++,  KAIc++,
ForteC++, Visual C++, and eMc++.  However I believe it works with any C++
compiler found
, because it does not use namespaces, iostreams, or STL – all areas that
are prone to porting problems from my experience.  It does have template
classes but their inclusion is optional as part of the implementation rather
than part of the foundation.

build dependencies are meticulously correct.  Smart linkers leave out
everything you don’t use, so don’t expect to see code bloat as a
punishment for using XMLFoundation.  Other development libraries were
not designed as well from a build perspective.  Your application will not
load any DLL’s as a result of using the XMLFoundation.  Consider a utility
built on the XMLFoundation, PackMan
It uses nearly every feature in the XMLFoundation aside from the XML
support.  In the future – Packman will be ready for XML related features
that come quick and simple because it has a good application

I suppose an entire article could be written about each of
the foundational classes, and I’m certain that they will be written.  They
are all well commented and coded with a highly experienced approach.  The
String class uses stack space when possible to avoid heap allocations. 
It’s the best string implementation I’ve ever seen.  The INI Profile class
uses triggers that allows your application to pick up real-time configuration
changes much like RegNotifyChangeKeyValue() in the Windows SDK.  Exceptions
can be configured to unwind the call stack to a memory buffer like Java’s
printStackTrace().  The Tree has an iterator.  The Directory can
delete recursively – on all platforms.  The Stack is entirely inline, with
standard and macro methods – It could not be any faster if it was coded directly
in assembly.  The StringList puts MFC’s CStringList to shame.


History and Future

You cannot build a house on foundation of wet cement that
has not cured yet.  With cement, minimizing stress prior to curing
minimizes cracking in your foundation.  The same is true of software. 
The XMLFoundation is sold and completely cured.  It would be too bold to say that the
XMLFoundation has no bugs in, but it has none that I am aware of and the code
has been heavily used.  It is a complete foundation.  Building an
application on any foundation like Java 1.0 or .NET 1.0 or anything 1.0 means
that if you don’t get slowed down by the bugs, you will be slowed down when you
find all the missing functionality.  This is Version 2009 of XMLFoundation. 
7 years ago – the XMLFoundation was very mature for it’s age – it came from a
good family – it’s mother had already been used on the largest software project
in the world.  Since then I have built several complex applications on
it.  It was completely stress tested with SMP hardware during a recent
Fortune 50 proof of concept implementation. 

The mother of the XMLFoundation was "The XML Object
Framework", born in 1998 & 1999(it was a long labor) for a client of mine.  The XMLFoundation was
born the following year.  The XMLFoundation sported a completely new
implementation of the xml parser based on the custom GString stream class that
was also born in 2000.  XML Journal Magazine reviewed a product built on
the early XMLFoundation object factorization and called it "5 Star / World
Class" (note: they did not review XMLFoundation they reviewed TransactXML). 
XMLFoundation was heavily developed the following two years.  In July of
2002 – a mostly undocumented – and mostly unnoticed release of the XMLFoundation
was made public.  This project is mature.

XML Foundation absolutely IS the future in
certain technology subsets.  It is a gift to the world of engineering, and
it comes with all the source code.  Universities that want to teach
algorithms, applications, or  OO Design will find the XMLFoundation to be a
great source code to base a curriculum on.  Independent authors who want to
write about cutting edge technology will find XMLFoundation a worthy
subject.  The future was written in the past.



The Need for Objects from XML

XML is data.  “Objects from data” is not a new concept. 
Programmers have been doing that for years, even before they were called
objects.  We still need to get data into objects today.  The data
can be XML or a result set, and the object might be a CDialog, a CORBA Object, a
COM object, or your own invention.  You still need to get the same thing done. 
Programmers have been doing this as long as there have been programmers.

If you apply enough force you can make the cube fit into the round
hole.  If you apply enough force you can do anything – even police
California.  The brute force approach is
to parse the XML into a DOM tree, and traverse the tree to gather the data
required by application/object variables.  This approach causes volumes of
“simple” source code to move data from XML into Structured Objects,  a
poor approach with respect to implementation time and long term maintenance.

Alternatively, the OO approach generalizes this process into reusable
functionality that enables objects to serialize to and from XML directly.  Software developers of every language have a similar need.  They must

A.  Write their own Object-XML tools,

B.  Find some production quality framework ready to use or 

C.  Use brute force and budget for maintenance programmers.



Don’t Be Square

From a procedural perspective we put data in square sets just to make
programming simple.  Consider this example data that is a "Customer"
with a list of "Orders" where each order has a list of "LineItems".  That is not a square dataset – but for
the sake of the application layer we have forced it to be square for the last 4

Brian 777 1 July 4, 1777 7 Firecrackers $111
Brian 777 1 July 4, 1777 14 Ariel Shells $222
Brian 777 1 July 4, 1777 21 Party Favors $444
Brian 777 2 July 4, 2009 28 Attorney Fees $222
Brian 777 2 July 4, 2009 35 State Fines $555

The repetition in red filled the hole in to make non-square data be
square.  The data in red is normally a pointer reference to the last sort
break at the DBMS kernel level, but various toolsets often expand it long form
so that 1 instance of a "Row" object does not rely on the data in
another instance.  It’s a terrible situation that has plagued applications
for as long as I can remember.  The problem is that in reality there is no
such thing as a "Row" object – it was more of a temporary/tool-object
to get the data into real objects like Customers, Orders and LineItems.  Countless data access
products in the form of VBX, OCX, ActiveX and various frameworks and libraries
serve up square data sets to applications that MANUALLY code the transfer of
data into their application objects with volumes of code that looked something
like this:

while( Tool.MoreRows() )
         Row = Tool.GetNextRow();
Loop until we find a sort break on OrderID

         if (Row.GetColumn("Order_ID")
!= LastOrderID)
= Row.GetColumn("Order_ID")
// Make an Order object - set the values - then add it to
the customer

              Order = new
              Order.date = Tool.GetColumn("Order_Date")
              Order.id = Tool.GetColumn("Order_ID")
         // Make a
new line item object - set the values - then add it to the order

         LineItem = new
         LineItem.id = Tool.GetColumn("LineItem_ID")
         LineItem.desc = Tool.GetColumn("LineItem_Desc")
         LineItem.price = Tool.GetColumn("Price")
         Order.AddLineItem( LineItem )


Notice all the use of Tool that
represents some sort of data set tool, class or library.  Building software
to accomplish this task of copying data into objects without such a tool would
dramatically increase the lines of code required to move the square dataset into
your application objects.  Knowing what tools to use can be the difference
between the success or failure of an entire project.  One bad tool, or one
missing tool can make all the difference in the world to a software developer.


Learn To Think Outside The Box

This is the same example data from the square result set represented in XML

<Customer id=777>
    <Order id=1>
        <Date>July 4, 1777</Date>
        <LineItem id=7>
        <LineItem id=14>
            <Desc>Ariel Shells</Desc>
        <LineItem id=21>
            <Desc>Party Favors</Desc>
    <Order id=2>
        <Date>July 4, 1777</Date>
        <LineItem id=72>
            <Desc>Attorney Fees</Desc>
        <LineItem id=42>
            <Desc>State Fines</Desc>


There is nothing square about XML.  XML is an N-airy
. That’s why we naturally use DOM (Document Object Model) to traverse
the data.  For 1000’s of years we thought that the world was flat. 
Engineers made it be square because that was easier for them to cope with and
now we live in the days where it begins to take it’s true shape. 
Unfortunately as of 2009, the opportunity of the paradigm data shape shift has
not been harnessed by most programmers that grew up in the square world and are
only familiar with square tools.  They take the most obvious development
path.  If you presented the problem of sorting to someone who has no tools,
they will likely build a "bubble sort" – because that is the most
obvious and immediate solution.

Typically the XML is parsed into a tree structure.  This means that the
linear and contiguous memory buffer of source XML is copied into many
fragmented pieces of memory across the heap – each element and in many cases
each token gets it’s own heap space.  This makes the XML elements
and attributes programmatically accessible with loops and recursion, just like Tool
did for square datasets.  The XML parser puts the Elements and
Attributes into this temporary fragmented memory tree structure so that the
application programmer can get at the information to copy it once more
into a final structure that can be displayed on the GUI or used by the
application.  It is likely going to take as much or more code to get from
the temporary DOM tree into the objects as it did to get from the square result set
into the objects.  In many cases it will require recursion that is
difficult to debug – much more difficult than the old fashioned iterative code
required to copy from square result sets.  Below is a code sample of some
common tasks:

    Due to the large
amount of code to change a DOM node name this article only 
    includes the basic
steps involved.  This is what it takes to
change an element 
    tag name using DOM, it’s much more involved using SAX.  
// 1. create a new node with the desired tag name

// 2. preserve old value to copy later

// 3. move all children of current node to new node

// 4. get parent of current node and replaceChild the current node with the new node

// 5. copy old value or new value to new node


// NOTE: Attributes are handled differently!

// This is how to add a new attribute using

pAttrList = m_pCurNode->Getattributes();
    _bstr_t bstrAttrName = (_bstr_t)(LPCTSTR)m_strName;
pNewAttr = m_pDOMDoc->createAttribute(bstrAttrName);
    _bstr_t bstrAttrValue = (LPCTSTR)m_strValue;
// TODO: The user must first search to be sure the attribute name is unique, or your XML will be invalid!

// One of
the well-formedness constraints that the XML 1.0 specification lists is that no
attribute name 


// may
appear more than once in the same start tag or empty element tag.  DOM
never deals with this issue.

// XML is
case sensitive, so attribute "Name" is different from attribute
"name", some tools built into DOM


// sure
would have been useful – good thing we have the XMLFoundation.

    AddNodeToTree(pNewAttr, m_hCurItem);


New World Order

The square world is becoming part of history like the flat world.  I remember back in the
early 90’s we tried to rid ourselves of the square world with something called
"The Object Database".  It was a great concept and the only
reason square prevailed against it is because nobody could implement an Object
Database that was fast enough.  Who cares how clean the code is if the
application is dysfunctional because it is too slow?  This is why
XMLFoundation is so performance oriented – that’s what it takes to change the
world.  The clean code alone is not enough.

Now I’ll explain how to accomplish the task of loading up your object with
the information in the XML using a fully object oriented
approach to data handling.  Customer, Order, and LineItem are derived from
XMLObject.  They must implement 1 virtual method called MapMembers() that
would look like this:

        MapAttribute(&m_nCustomerID, “id”);
        MapMember(&m_strName, "Name");


        MapAttribute(&m_nOrderID, “id”);
        MapMember(&m_strDate, "Date");   
// can also be mapped to a date/time object




        MapAttribute(&m_nLineItemID, “id”);
        MapMember(&m_strDesc, "Desc");
        MapMember(&m_strPrice, "Price");    

Now all the object assignment and creation code is summed up into this one


This assigns ALL member variables and creates sub-objects

    Customers.FromXML( pzXML )  

If you had a trace
statement in the constructor of the Order, you would see that it was
called  for every appearance of an Order in the XML.

Now suppose you wanted to manipulate some member variables then regenerate
the XML: just assign your member variables normally then regenerate your XML – that’s just as easy

*pzXML = Order.ToXML()

MapXMLTagsToMembers() defines everything needed for your objects to read
or write XML as a base method.  Without the XMLFoundation you
would  have to code all that looping and mapping 2 times if you wanted both
reading and writing XML.  Without the XMLFoundation you will have a larger
maintenance issue if any XML document structure changes because you’ll have to
hunt through your looping and recursion routines to find the Element name to
change.  XMLFoundation provides countless other niceties such as mapping
any number of XML tags to the same member, and conditional inclusion of members
in the output XML based on tag name or the member state such as DIRTY indicating
that the member was updated and you only want ToXML() to generate a delta of the
data rather than the entire set.  You can specify element order or have
them output alphabetically.  Common needs that can all be accomplished in 1
line of code rather than pages of code.


    The following code is
so involved using DOM only the basic steps are described.
    This is what it takes to
change an element tag name using XMLFoundation.
     // This is how backward compatibility is achieved when Element tag
    // names change. For example, if m_nVersion is mapped to an Element
    // named "VersionNumber" but you want all future protocols
to refer to this 
    // Element as to "ProtocolVersion". This is achieved by the following code:
    // This allows either "VersionNumber" or "ProtocolVersion" to set the value of 
    // m_nVersion, but always refers to it as
"ProtocolVersion" while serializing XML.
    SetMemberSerialize("VersionNumber", false );

This is how to add a new attribute using XMLFoundation

    // if nUpdate=1 the attribute list will be searched and updated if there is an existing 
    // attribute called [pzName]if not found or if nUpdate=0 the new attribute will be added.

     void AddAttribute( const char *
pzName, const char * pzValue, int nUpdate=0 );

It’s fun to compare the differences between DOM and XMFoundation, but much of
the functionality in the XMLFoundation cannot be compared to anything in
DOM.  For example, the XMLFoundation maintains a bit flag field for each
member that it manages.  These are the values that can be managed:

    // The value does not sync with the original value set by the Object Factory
    #define DATA_DIRTY 0x01 
    // The member has been set by either the Object or the Object
    #define DATA_NOT_NULL 0x02
    // The member has been assigned a value from the Object
    #define DATA_CACHED 0x04
    // The member has never been assigned a value, it
is uninitialized
    #define DATA_NULL 0x08 
    // The member should be included in the xml serialization stream
    #define DATA_SERIALIZE 0x10 

The following interface uses some of the member state flags:

        // When objects are populated from the XML stream, they have a state of
        // Not dirty and Not null. When the members are assigned by the derived 
        // class through SetMember() or
SetMemberByTag() the state becomes dirty.
        // When using the SetMember() or
SetMemberByTag() there is no need to 
        // setMemberDirty(), but if the derived class does direct assignments
        // to members that should be serialized,
setMemberDirty() should be called.
        // if you ever want to serialize or track the state change only.
        // Setting bDirty to 0 will clear the dirty flag for a member variable.
bool setMemberDirty(void
*pAddressOfMemberToSet, int bDirty = 1);
bool setMemberDirty(char
*pzTagNameOfMemberToSet, int bDirty = 1);

        // true if the memory state does not sync with the 
        // original value set by the Object Factory
bool isMemberDirty(void
bool isMemberDirty(char

        // true if the member has never been assigned a value, it is uninitialized
bool isMemberNull(void
bool isMemberNull(char

        // true if the member has been assigned a value from the Object Factory
bool isMemberCached(void
bool isMemberCached(char



XMLFoundation also has many options available during the creation of
the XML.
DOM has nothing that compares.


    // serialization flags for


    // Otherwise XML tags appear in the order they were mapped


    // deeply recurse without including dirty members

    #define RECURSE_OBJECTS_DEEP    

    // includes any member with a state of DATA_CACHED



    // Do not serialize any data from
MapAttribute() members


    // Do not serialize attributes that came in via XML but
were unmapped with MapAttribute()


    // Adds the DOCTYPE to the beginning of the XML


    // includes any member regardless of it’s state

    #define FULL_SERIALIZE    

    // include OID’s only

    #define USE_OBJECT_MARKERS    

    // more compact, faster, less human readable, The XML will
have no Tabs, Carriage returns or Linefeeds

    #define NO_WHITESPACE    

    // do not include empty strings in XML

    #define NO_EMPTY_STRINGS    

    // do not include the ObjectDataHandler registered with

    #define EXCLUDE_OBJECT_VALUE    


It also has a SAX like (but faster and far
simpler) way to subscribe to notifications

// When a tag is encounterted that does not have a MapMember() entry
// associated with it, this handler is called for the developer
// to supply an “on-the-fly” MemberDescriptor during the Factory process.
// This is useful for dynamic objects. 
virtual MemberDescriptor *HandleUnmappedMember(
const char *pzTag );

// Generic Event Handler for custom object behavior. Abstract to reduce virtual method table
// —————————————————————————————–
// nCase = "XMLAssign", (member tag, XML value, value len, null)
// when the FromXML() contains a value for a member set dirty by
// see MemberDescriptor::Set() in MemberDescriptor.cpp for details.
// nCase = "NonNumeric" (member tag, XML value, value len, null)
// when non-numeric XML data is mapped to an numeric only type.
// nCase = "EmptyString"(member tag, XML value, value len, null)
// when an empty(“”) value is assigned to a string, empty often differs from ‘unknown’ or ‘unassigned’
// nCase = "ObjectUpdate" when OBJECT_UPDATE_NOTIFY is a set behavior flag.
(oid, null, flags, pObjSrc)
// nCase = "MemberUpdate" when MEMBER_UPDATE_NOTIFY is a set behavior flag. (tag, value,
valuelen, null)
virtual void *ObjectMessage( int
nCase, char *pzArg1, char *pzArg2,
unsigned int nArg3, void *pArg4)


Faster than Fast

The approach used by XMLFoundation is faster than SAX.  Since the object
factory and the XML tokenizer were built for each other they did some unusual
tricks for each other.  The tokenizer uses a unique approach to begin
with.  It’s purely pointer based.  Tokens are structures that point
into the source XML, except for entities that get expanded into a special memory
region.  Tokens do not hold copies of any data.  During object
factorization it becomes necessary to have the token data in a null terminated
string format.  The big performance boosting hack is that to obtain null
terminated strings, the tokenizer actually plunks a null down over the first
byte past the end of the token data.  It keeps track of the data it
clobbers and restores it before  parsing out the next token.  There
are no event calls that needlessly push data on the stack just to immediately
pop it back off.  Performance profilers showed that call stack pushes and
pops were the single largest consumer of CPU cycles in the tokenization process. 
eliminates them by "pulling" the data through a call to [void
getToken(token *tok)], rather than the SAX approach that gets the data
"pushed" into the application events with between 2 and 7 arguments
depending on the token type – the SAX specification defines the interface so it
has nothing to do was individual XML parser implementations.  The
XMLFoundation is the only XML parser that uses this approach.  It is
non-standard, and not in compliance with W3C interfaces to an XML Parser – For
our uses, It’s
better than any W3C standard.

I realize that the vast majority of people who use XMLFoundation would never
care about these grungy technical details.  To say that it is very fast is
enough for most people, but I am also writing to the people at the Apache
Foundation, and Microsoft, and IBM, W3C, and the many other people who have
built their own XML Parser implementations.  Fast is an
understatement.  Performance is a prevailing design pattern found throughout the
XMLFoundation.  For example the XMLObject class is carefully designed to
add minimal CPU cycles during construction because it is to the XMLFoundation
what CObject is to MFC.  It has been carefully designed to add minimal
entries to the virtual
method table
.  In many cases virtual calls were consolidated for that


Object Factorization

The Object Factory is the part of the XMLFoundation that instantiates
objects for you based on certain element tags in the source XML.  It is
based on the same principle as DECLARE_DYNCREATE() that allows MFC to instantiate
CView derived classes for you.  In the XMLFoundation it is called
The XMLFoundation uses DECLARE_FACTORY() to instantiate COM and CORBA objects as
well.  If a tag is mapped to an object in a list or tree structure, then
every time that tag is encountered at the level it is mapped it will create a
new instance for you and put it in the data structure you specified with all
it’s member variables already assigned from the source XML as you have them


Member Mapping

XMLFoundation has support for mapping to all native C++
datatypes.  It also has support for mapping into data container
objects.  It has specific support for RWCString, CString, and GString and
it’s very easy to add support for other by deriving from the XMLFoundation class
"StringAbstraction" and supplying the 5 pure virtual methods that will
enable any kind of data container class to interoperate with the Object Factory
for automatic member assignments.  These are the MemberMap Methods in

     // Map an int / long int / or very long int

     void MapMember(int
*pValue,const char
     void MapMember(long
*pValue,const char
     void MapMember(__int64 *pValue, const char

     // Map a string, see StringAbstraction.h for interface and samples

     void MapMember(void
*pValue,const char
*pTag,StringAbstraction *pHandler);

     // Map an object into a hash table, binary tree, or QSort array

     void MapMember(void
          const char
          const char
*pNestedInTag = 0);

     // Map a collection of Strings

     void MapMember(void
*pStringCollection,     const char
          StringCollectionAbstraction *pHandler,
          const char
*pNestedInTag = 0);

     // Map a dynamically growing Integer array

     void MapMember(void
*pIntegerArray,     const char
          IntegerArrayAbstraction *pHandler,
          const char
*pNestedInTag = 0);

     // Map a sub-object using a tag other than defined in the DECLARE_Factory

     void MapMember(XMLObject
*pObj,     const char
*pDefaultTagOverride = 0, 
               const char
*pzWrapper = 0 );
     // Map a collection of Objects

     void MapMember(void
*pList,const char *pObjectTag,
          ListAbstraction *pHandler,const char
*pNestedInTag = 0, 
               ObjectFactory pFactory = 0 );
     // Map a sub-object pointer to an object residing in the ObjectCache

     void MapMember(XMLObject
**pObj,const char *pzTag, 
               const char
*pNestedInTag= 0, ObjectFactory pFactory=0);


Object Model Navigation

By using the XMLFoundation you inherit some powerful
navigation features that can be used to help you debug your application with the
Dump() member.  Because the factory manages all the object relationships, a
new kind of object navigation arises: objects know their creators so an
"Order" can know at runtime if it resides inside a list in a
"Customer", or some other kind of object, or if it is not contained by another
object at all.  This is what a full Dump() output looks like: 

Object Instance name: MyOrder
     string     OID =
     string     UpdateTime =
               References = 1    
     Type :List<XMLObject *>
     Tag :LineItem
     Contains:3 items
     Object Instance name: MyOrderLineItem
          string     OID = 1121.0000
          string     UpdateTime =
                    References = 26         
          Type :string
          Tag :Description
          Value :
          State :(Clean | Null | Uncached)
          Kind :Element
          Type :int
          Tag :ProductID
          Value :11
          State :(Clean | Valid | Cached)
          Kind :Element
          Type :string
          Tag :UnitPrice
          Value :21.0000
          State :(Clean | Valid | Cached)
          Kind :Element
     Object Instance name: MyOrderLineItem
          string     OID = 332.5000
          string     UpdateTime =
                    References = 21         
          Type :string
          Tag :Description
          Value :
          State :(Clean | Null | Uncached)
          Kind :Element
          Type :int
          Tag :ProductID
          Value :33
          State :(Clean | Valid | Cached)
          Kind :Element
          Type :string
          Tag :UnitPrice
          Value :2.5000
          State :(Clean | Valid | Cached)
          Kind :Element
     Object Instance name: MyOrderLineItem
          string     OID = 7234.8000
          string     UpdateTime =
                    References = 23         
          Type :string
          Tag :Description
          Value :
          State :(Clean | Null | Uncached)
          Kind :Element
          Type :int
          Tag :ProductID
          Value :72
          State :(Clean | Valid | Cached)
          Kind :Element
          Type :string
          Tag :UnitPrice
          Value :34.8000
          State :(Clean | Valid | Cached)
          Kind :Element
     Type :string
     Tag :OrderDate
     Value :1997-05-06
     State :(Clean | Valid | Cached)
     Kind :Element
     Type :string
     Tag :ShippedDate
     Value :1997-05-09
     State :(Clean | Valid | Cached)
     Kind :Element


GUI Objects

This is an example of what is involved to get XML to the
GUI.  The XML is somewhat complex to show how simple the code will
be.  The XML is a "Customer" with a list of "Orders"
where each order has a list of "LineItems".  This is the XML:

     <ContactName>New Dude</ContactName>
     <City>Chocolate City</City>


Notice that the XML
foundation will parse directly in to the CStrings that are already DDX bound to
MFC’s UpdateData().  This is accomplished through Multiple
Inheritance.  Our Dialog class derives from both MFC’s CDialog, and
XMLFoundation’s XMLObject.

The sample application reads XML and displays it in the GUI
where it can be changed by the user, then saved back out to XML that reflects
the users changes.

The complete code for this example is in "XMLDialog",
but for the purpose of understanding what it takes to integrate XMLFoundation
with an MFC Dialog this shows you ALL the code of interest.


// Begin XMLDialog.h

#include "xmlObject.h"

#include "GList.h"

class CXMLDialogDlg : public
CDialog, public XMLObject
    //This is XMLFoundation releated code
     GList m_lstOrders;
     virtual void
     virtual void
*ObjectMessage( int nCase, char
*pzArg1, char *pzArg2, unsigned int
nArg3 = 0, void *pArg4 = 0 );
     DECLARE_FACTORY(CXMLDialogDlg, Customer);

    //This is code created by App Wizard

     CString     m_strCity;
     CString     m_strCountry;
     CString     m_strName;
     CString     m_strRichEditXML;


// End XMLDialog.h




// Begin XMLDialog.cpp


void CXMLDialogDlg::MapXMLTagsToMembers()
     MapMember(&m_strCountry,   “Country”,    
MyOrder::GetStaticTag(), &gGListHandler, 0 );

void CXMLDialogDlg::OnBtnMakeXML() 
     UpdateData(TRUE); // pickup the changes
from the GUI into the member variables

     m_strRichEditXML = ToXML();  // create the new XML

     UpdateData(FALSE);           // display the new XML in the edit box


void CXMLDialogDlg::OnBtnLoadGUI() 
     FromXML(m_strRichEditXML);  // parse the XML into ‘this’

     UpdateData(FALSE);// update everything on the GUI that AppWizard has a DDX map for

     // note:UpdateData() does not push the list of
‘Orders'[m_lstOrders] into the 

     // ListCtrl. The simplest way is to iterate
[m_lstOrders] that contains the 

     // ‘Orders’ after the call to
FromXML() is complete.

     // This shows you the complex, SAXish like, way that will add ‘Order’ objects 

     // to the GUI as they are added to
[m_lstOrders] by the Object Factory during 

     // the call to FromXML(). This requires adding
some code to this’s constructor: 


     // This causes the XMLFoundation to call
ObjectMessage(), as each “Order” gets

     // it’s data from the XML.  This Actually adds data to the GUI DURING THE

     // PARSING PROCESS, as opposed to
the ‘simplest’ way that will add the data 

     // into the CListCtrl AFTER the parsing process. 


// The Order
object is yellow
, the
complexity is CListCtrl, not the XMLFoundation.

void *CXMLDialogDlg::ObjectMessage( int
nCase, char *pzArg1, char
*pzArg2, int nArg3, void
*pArg4 )
     if(nCase ==

= (MyOrder
nItemIndex = m_List.InsertItem(LVIF_TEXT|LVIF_PARAM, 0, pO->m_strOrderDate
0, 0, 0, (long)pO);
1, pO->m_strShippedDate);

          // uncomment this to see that we can generate XML subsets very easily

pO->ToXML() ); // call base class method

     return 0;



CORBA Objects

The XMLFoundation was designed and built for CORBA before
it ever added any support for MFC.  If you have a pre-existing CORBA system
that needs some XML tools you have come to the right place.  If you are
building a new CORBA system – this is best tool available for XML support.

If you have read this document all the way to this point
then you will likely understand how the XMLFoundation works for CORBA by showing
you this tiny piece of code:

CustomerImpl : public virtual CustomerBOAImpl, public
virtual XMLObject

along with the IMPLEMENT_ORB_FACTORY() macro defined in
XMLObject.h,  this is how CORBA can natively support the FromXML() and
ToXML()  by using the XMLFoundation.  The Object Factory can instantiate
your interface objects for you based on the XML.

CORBA implementations can be done in Java or C++.  The XMLFoundation
supports both.  CORBA breaks down the language barrier allowing Java
applications to easily, and natively deal with C++ objects.  This example
details the creation of C++ CORBA objects – The Java implementation is nearly
identical further blurring the lines between Java/C++ within the same project.   

The C++ CORBA implementation
will bridge into J2EE Application servers everywhere, it will work for any ORB
but a few of the most popular ones have been tested, and the makefiles are
included with the CORBA sample that ships with the XMLFoundation.  The
three makefiles included are for:



Borland/Enterprise Studio    
–        Visibroker


IONA/iPortal Enterprise     
–        Orbix


BEA/Weblogic Enterprise   
–        ObjectBroker    (works great with Tuxedo implementations)


This example extends the ORB to provide native XML
accessors.  The sample CORBA application is based around 1 very simple object type. 
It has a unique integer we call a CustomerID and a string we call a CustomerName. 
Each customer may contain 0 to n references to another object of the same type
as it’s self, a MyCORBAObject.  This would model something like a list of
Customers that were referred by ‘this’ customer.

The IDL looks like this:

module ExCORBA
     interface MyCORBAObject
          void getXMLState(out
string s);
          void setXMLState(in
string s);
          void setState(in string
s, in long l);
          void addSubObject(in
string s, in long l);
          void delSubObjects();
getSubObjectIOR(in long l);
          void dumpState(out string

Follow this 12 step program

This is a very simple application.  The client application makes 12
calls to the server.  Every even numbered call is exactly the same – it is
a call to getXMLState() to see what’s going on in the server. The client obtains an initial IOR from a server serialized IOR upon server startup.

1. Assign some state in a native CORBA call. This is a typical CORBA data assignment operation. Two values are set in the object. The client assignes two members on the server. The code looks like this on the client:


2. View the state of the object in XML. This uses the XML accessor to return the state of the object. The code looks like this on the client:

     CORBA::String_var s;

and looks like this on the server:

ExCORBAImpl::getXMLState( CORBA::String_out s)
          const char
*p = ToXML();
          s = CORBA::string_dup(p);

and the result XML is this:


The tag names are configured by the ExCORBAImpl object like this:

          MapMember(&_nCustID, "CustomerID");
          MapMember(&_strCustName, "CustomerName",
          MapMember(&m_lstCMyImplObjs, "MyCORBAImpl",

3. Update the state of the object through XML. Step 1 used a typical object accessor to assign the state. Step 3 accomplishes the same through XML. The Code on the client looks like this:


On the server the code looks like this:

ExCORBAImpl::setXMLState( const char*
pzXML ) 
           FromXML( pzXML );

4. View the modified object state in XML. This is the exact same code (client and server) as Step 2. We’re calling
getXMLState() again, and the result is:


5. Add CORBA Sub-Objects through XML. Step 5 is a lot like step 3 where we updated the name “root” to
"SuperUser" through an XML assignment. This time we’ll add an object reference.B The client code looks like this:





Bush Jr.</CustomerName>"


and this is the code on the server:

ExCORBAImpl::setXMLState( const char*
pzXML ) 
           FromXML( pzXML );

6. (exactly like steps 2 & 4) – View the object’s XML stateB

       <CustomerName>Al Gore</CustomerName>
       <CustomerName>George Bush Jr.</CustomerName>

7. Get a CORBA object reference for object instance 456. On the client the code looks like this:

     ExCORBA::MyCORBAObject_var CustObject2;
     CustObject2 = CustObject1->getSubObjectIOR(456);

and on the server we walk the list of objects and return the first one that matches the supplied CustomerID like this:

ExCORBA::MyCORBAObject_ptr ExCORBAImpl::getSubObjectIOR(CORBA::Long CustomerID)
      // create an iterator[it] for the
[m_lstCMyImplObjs] list

      GListIterator it(&m_lstCMyImplObjs);
      while(it()) // while there is more in the list

           XMLObject *pO = (XMLObject
*)it++; // get the next Customer Interface

           ExCORBAImpl*pIO = (ExCORBAImpl*)pO->GetInterfaceObject(); // widen the pointer

            // test for a match – Use a keyed datastructure in the real world

(pIO->GetCustomerID() == CustomerID)  
                // Return the CORBA Interface to the desired object

      return 0;

8. Exactly like steps (2, 4 & 6) EXCEPT we are using the Object ref returned by step 7.

Bush Jr.</CustomerName>

9. Add a Sub-Object without using XML. In the same way we used a traditional member assignment in step 1, we can create a new object reference to demonstrate the two models seamlessly working together. On the client:


and on the server the code looks like this:

ExCORBAImpl::addSubObject( const char*
s, CORBA::Long l )
           ExCORBAImpl *p = new
           p->_nCustID = l;
           p->_strCustName = s;
           m_lstCMyImplObjs.AddLast((XMLObject *)p);

10. Get an object reference to the object created in step 9 and display it’s state in XML. This is the client code:

     CustObject2 = CustObject1->getSubObjectIOR(1475); // exactly like step 7

     CustObject2->getXMLState(s); // like steps (2, 4, 6, and 8) using the new reference.

and the result is:


11. Deleting Sub objects. All objects, no matter how they were created, are destroyed the same. The list contains both Factory created objects and Objects created the tradional way. Once again this shows how seamlessly the ORB fits together with the XMLFoundations’s Object Factory. This the CORBA Implementation/Interface and XMLObject are all one in the same. This cleans up the whole mess.


on the server:

ExCORBAImpl::delSubObjects() IT_THROW_DECL((CORBA::SystemException))
           GListIterator it(&m_lstCMyImplObjs);
       XMLObject *pO = (XMLObject *)it++;

12. To see that step 11 worked, view the XML state like we did in 2,4,6,8, & 10. Now all the contained objects are gone, and
"SuperUser" is alone.



COM Objects

Create a basic ATL COM project with Visual Studio.

Visual Studio will write your
IDL, and implementation header files.  The following code sample is
the standard implementation header file with a few small additions
(highlighted in yellow) required for XML support.


public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CMyATLObj, &CLSID_MyATLObj>,
public IDispatchImpl<IMyATLObj, &IID_IMyATLObj, &LIBID_EXATLCOMLib>,
public XMLObject


void MapXMLTagsToMembers();


In your implementation file
you’ll need to add the macro at a global scope and implement MapXMLTagsToMembers() to define the Object to XML mappings.  This example maps an integer,
a string, and a list of COM objects. 




   MapMember(&m_nInteger, "CustomerID");
   MapMember(&m_strString, "CustomerName", &gGenericStrHandler);
   MapMember(&m_lstCMyATLObj, CMyATLObj::GetStaticTag(),

In the ExATLCOM sample application several additional methods have been added
to the COM Object.  Most notably put_XMLState() that has the ability to
assign members variables and create COM objects when supplied well-formed XML as

CMyATLObj::put_XMLState(BSTR newVal)
     _bstr_t b(newVal);
     FromXML((const char *)b);
     return S_OK;


Object Caching and Instance Management

XMLFoundation has been serving up the XML related needs of
the application layer for nearly a decade.  It has been used to build a
wide variety of application types.  A common recurring need in the
application layer has to do with "data updates".  Any application
that receives XML updates might consider the performance advantages and
reduction in development labor by using the XMLFoundation to solve the problem
for them.

For example, suppose you had some large dump of XML
data.  In your application layer you need to quickly access individual
pieces of that information.  In just a few lines of code, the XML can be
mapped to a keyed data structure for fast indexed reads by your application.  If the initial
XML dataset was 100+million records – you will want to provide updates to your
indexed information rather than rebuilding the entire index.  You could
write the code to search for the data to update, or allow the XMLFoundation to
manage it for you.

Another common example in distributed systems: 
Data is often cached at a middle tier or in the application itself. 
Efficiently designed systems only update an "Address" rather than a
whole "Customer" and all his "Orders" when an
"Address" changes.  XMLFoundation can greatly simplify this task.  

At the core of caching is something XMLFoundation calls the
OID, or Object ID.  It is a unique key to the object, and any object that
participates in XMLFoundation caching must have one.  The definition of the
OID can come from 2 places.  It can be defined in the XML data, or it can
be defined by the object that mapped the data.

This is how a "’MyOrderLineItem" object might
define the OID.  It uses a combination of the "ProductID" and
"UnitPrice" so in this example a price change constitutes a different
object.  Normally an OID has a direct correlation to DBMS indexes in
properly normalized
.  ObjectID’s can be made to work well over poor data models
too.  This example code uses two XML Elements("ProductID" and
"UnitPrice") to build the unique
object ID.   MapObjectID()

also allows you to use Attributes to define the OID.

void MyOrderLineItem::MapXMLTagsToMembers()

Alternatively the OID can be directly defined by the data itself with a special
attribute named "OID" – so that NO CODE needs to be written.

<MyOrderLineItem oid=’777′>

The sample application "ObjectCache" provides over 30 test cases that
detail the usage of object caching. 



5Loaves is included with the XMLFoundation, but not in the foundation. 
It is a tool that uses XMLFoundation.  It needs the XMLFoundation but the
XMLFoundation does not need 5Loaves.  5Loaves is implemented in a single
file called ServerCore.cpp.  It is a unique piece of code and it is very simple to use.  There is no
header file for ServerCore.cpp and it causes no DLL’s to be loaded by your application. 
It is a portable, properly threaded, and well written server core that can be
applied to countless custom server implementations.  It is a Proxy. 
It is an HTTP Server.  It is a ‘connection joiner’.  ServerCore.cpp is
currently used to accept TCP connections for an advanced networking product called Xfer. (note: Xfer is not open
source).  The ServerCore has a good portable threading
model with some unique features that allow you to limit such things as, number
of connections per IP/Subnet and Connections per second.  The 5Loaves HTTP
Server works faster than IIS or Apache in some cases.  It has been designed to be fast and includes unique features such as
‘content caching’ to serve up prebuilt HTTP headers and data from memory rather
than from disk.

The Core of 5Loaves is a ground up POSIX threaded TCP server.  This
server template can be applied to build many types of applications that service
TCP connections.  The sample programs are server applications that use the
5Loaves ServerCore.  One example is the 5Loaves shell console.  It is
a command line interpreter like the DOS prompt or a Unix shell written from
scratch.  It is a useful application starting point if you ever need a
simple shell that has been tested in Linux, Solaris, AIX, HPUX, and
Windows.  There is also a Windows Service application.  It contains
the proper implementation for integrating 5Loaves with the Windows Service
Control Manager.  There is also an example program with all the source to
an ActiveX implementation of 5Loaves so you can see that this server core can be
embed just about anywhere in any development language.

The following code sample shows you how to start the HTTP service from inside
your own process.  You can’t do that with IIS or Apache.  The HTTP
server does support binary plugins like ISAPI, and there is an example program
that creates plugins.  It’s an advanced HTTP server.

This is a complete source example to embed a solid and high performance 

        // HTTP
Server in your application that starts on a variable port with a 

        // variable home


*pzBoundStartupConfig =
// <-- notice the [System] section
// <-- notice the [HTTP] section

CMyClass::StartHTTPServer(const char
*pzHomeDirectory,const char
Fill the two variables into the startup config string

Set the global profile object

GProfile( strCfgData, strCfgData.Length()) );
Start the HTTP service



FiveLoaves Tunneling and

5Loaves provides tunneling with encryption and compression as a base
service.  It can be used to secure internet connections much like SSH. 
It works by running some form of server with the 5Loaves engine on two
machines.  The data passed between those machines can be compressed or
encrypted or simply logged.  Starting the server is done exactly the same
no matter if you are running the HTTP service or the Tunneling service – only
the configuration startup string changes.  The following configuration
example will open a listener on port 1972, anything it receives will be
encrypted and compressed and sent to [www.ExampleServer.com] on port 2009. 
The server will decrypt and decompress the data from port 2009 and forward it to
port 1972, so the data on port 1972 at the server will be as if the client had
directly sent it.  To open a tunnel entry point on the client side use


to exit the tunnel on the server  

RemoteMachine=  (we could also use an internal resource like

Note: you can start any number of tunnels, simply increment the section
[Tunnel2] and [Proxy2].  ServerCore will stop loading tunnels at the first
break of numeric order of the [sectionN]


This type of usage may be of
interest especially to web developers or people who are just curious. 
Sometimes it’s interesting to see the data between the HTTP server and the
browser.  Redirects and HTML frames and Javascript can make that difficult. 
This section sets up a clear text proxy just for the purpose of seeing the
transmission log between the browser and the web-server.  Once this is
running, connect with a web browser to and you will really be
connecting to www.SampleWebSite.com



In the folder HTMLSpy you will see a log file of the communication between
the browser and the web server.  I used Firefox as a browser, and the
5Loaves HTTP server for this example.  I loaded the page twice in the
browser so that you can see the caching mechanism working between the browser
and the web server in this log file:  The header is defined like this:
"tx->s" means "transmit to server" so that is information
that came from the browser, it is followed by the time, then the total bytes
transmitted, 470 in this case.  The reply is "tx->c" or
"transmit to client", you can see that the HTTP server responded with
181 bytes that instructed the browser to use the version it has cached. 
You can also see that the "Server:" name was set to "MyWebServer",
that is a variable in the 5Loaves HTTP server unlike the static names IIS and
Apache use. 


GET / HTTP/1.1
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:
Gecko/2009060215 Firefox/3.0.11
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: en-us,en;q=0.5
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 300
Connection: keep-alive
If-Modified-Since: Thu, 01 Jan 1970 00:00:00 GMT
If-None-Match: 1136664248
Cache-Control: max-age=0

HTTP/1.1 304 Not Modified
Server: MyWebServer
Date: Wed, 01 Jul 2009 00:51:42 GMT
Connection: keep-alive
Keep-Alive: timeout=20, max=149
ETag: 1136664248
Content-Length: 0

By changing the log binary to yes, now we can see what a small gif file looks
like over the wire from the web server.


48 54 54 50 2F 31 2E 31 20 32 30 30 20 4F 4B 0D 0A 44 61 74 65 3A 20 57
65      HTTP/1.1 200 OK..Date: We
64 2C 20 30 31 20 4A 75 6C 20 32 30 30 39 20 30 31 3A 30 35 3A 32 39 20
47      d, 01 Jul 2009 01:05:29 G
4D 54 0D 0A 53 65 72 76 65 72 3A 20 4D 79 57 65 62 53 65 72 76 65 72 0D
0A      MT..Server: MyWebServer..
43 6F 6E 6E 65 63 74 69 6F 6E 3A 20 6B 65 65 70 2D 61 6C 69 76 65 0D 0A
4B      Connection: keep-alive..K
65 65 70 2D 41 6C 69 76 65 3A 20 74 69 6D 65 6F 75 74 3D 32 30 2C 20 6D
61      eep-Alive: timeout=20, ma
78 3D 31 34 39 0D 0A 4C 61 73 74 2D 6D 6F 64 69 66 69 65 64 3A 20 53 75
6E      x=149..Last-modified: Sun
2C 20 31 33 20 4D 61 72 20 32 30 30 35 20 32 32 3A 33 32 3A 33 36 20 47
4D      , 13 Mar 2005 22:32:36 GM
54 0D 0A 45 54 61 67 3A 20 31 31 31 30 37 34 39 35 35 36 0D 0A 43 6F 6E
74      T..ETag: 1110749556..Cont
65 6E 74 2D 74 79 70 65 3A 20 69 6D 61 67 65 2F 67 69 66 0D 0A 43 6F 6E
74      ent-type: image/gif..Cont
65 6E 74 2D 6C 65 6E 67 74 68 3A 20 37 34 0D 0A 0D 0A 47 49 46 38 39 61
10      ent-length: 74….GIF89a.
00 10 00 91 00 00 00 00 00 FF FF FF FF FF FF 00 00 00 21 F9 04 01 00 00
02      ………………!……
00 2C 00 00 00 00 10 00 10 00 00 02 1B 94 8F A9 CB 07 AD C0 83 4E 52 23
2D      .,……………….NR#-
CD BA F1 BE 7C 5B 76 91 E5 54 5E EA CA 1A 05 00



5loaves has a unique network connectivity utility built in.  It allows
machines behind a firewall that cannot "listen" for connections
outside the network to accept connections from anywhere without any firewall
configuration changes.  There is a complete example called "FilePoster"
that puts a file on a machine behind a firewall.  This is a bare bones
‘proof of concept’ implementation that gives you a working model to customize
for your own purposes.  It requires 3 machines to see it work as designed.

Machine 1 – (the switchboard) should be located on the
internet.  You must run the HTTP service along with the "Switchboard
Service", you can see that this example runs it on port 81 just incase you
have IIS or Apache already on port 80.




Machine 2 – (the server) this is the machine behind the
firewall that you want to open up a connection path to.  It will poll the
switchboard server looking for connections.  It should run 5oaves with the
configuration below.  This will accept remote data, and write it to disk in
a file at "c:5LMessagesUBTsAccountForYou".  You could change
the application logic that writes the file – you can do anything with the data
that may contain commands, database queries, or custom logic.



Machine 3 – (the client) runs the FilePoster sample
application.  Machine 3 can reach "the switchboard (machine 1)"
but not "the server (machine 2)".  We will send the data to
Machine 2 and get a response back from that machine.  In this case we are
simply writing the data we send to a file, but the data could just as easily
have been an SQL statement and the return data could be the result set rather
than just a confirmation that the file was written.


How it works:

The "server" polls into the "switchboard" with an HTTP
GET.  The "client" pushes a multipart HTTP POST to the
"switchboard".   The switchboard joins the connections and
proxies the data.  An HTTP GET needs an HTTP  "200 OK" so the
"switchboard" server rips off the POST headers from the data sent up
by the "client" and replaces them with an HTTP 200 followed by the
POST data that gets proxied straight through.  Once this initial message proxy is complete, the client connection that
POSTed it 
waits in the switchboard, for the server to POST back a response. Then the Switchboard goes
through the same process of ripping off the POST HTTP header and replacing it with a 200 OK
before sending the response back to the client. Lastly, the switchboard server
replies with an empty HTTP 200 to the servers response POST to complete the normal HTTP request/response
design for both the client and the server.  This allows it to pass through
HTTP proxy servers and direct support for them is included.  Technically
this is a loophole through most networks that only allow HTTP, because as you
see we invented a new protocol that looks like HTTP, but in fact it is not.

Plugins and Language Drivers

The XMLFoundation supports "Language Drivers".   Language
drivers allow methods written in any language to be invoked programmatically. 
Just about every programming language can have a language driver developed for
it.  Three complete driver implementations come with the XMLFoundation
source code.  Since both the HTTP Server and the Messaging System are built
into ServerCore.cpp, they both allow programmers to easily invoke their own
code, even if they are not C++ programmers.  For years HTTP servers used
CGI to accomplish this task.  They still do.  They do it even better
through ISAPI.  XMLFoundation does it better still by using a language abstraction
to accomplish the same thing.  The programmer’s code is "the plugin"
that is executed by the "Language Driver".  All of the source
code for everything I speak of is included in the XMLFoundation, see [IntegrationBase.h/cpp][IntegrationLanguages.h/cpp],
using DynamicLibrary.h to load the DLL/SO’s on many platforms.

A C++ Plugin is as fast as ISAPI.  The other languages are fast too
because the "interpreter" or "Virtual Machine" or "runtime"
(terminology changes but the concept is the same) is cached in memory.  It
certainly is more functional than ISAPI from an "open standard
interface" point of view.  Conceptually it allows "methods"
to be written in any language or even multiple languages within the same
application.  This technology is stable.  It first appeared in
TransactXML Server, it was also integrated into my XSLT, and DesignerXML and
Xfer which
support Perl and Python language drivers as well.

One sample application titled "PluginExample" is devoted to C++
plugins.  It shows how to write several types of POST handler plugins for
the HTTP server.  You will find the utility class CMultiPartForm handy if
you need to write a handler for a Multipart HTTP POST, other plugin utilities
like PlugInController found in PluginBuilder.h make the job of building a plugin
as easy as possible.

This is how to build a custom dynamic web page from an HTTP server plugin:

extern “C” __declspec(dllexport) void
Test2(void *pHandle, DriverExec Exec,
         const char
*One,     // string

          const char
*Two,     // string

          const char
*Three)     // string

     PlugInController PIC(pHandle, Exec);
     // build the string "5Loaves
Invoked me with [root] and [777] and [superuser]!"

     PIC.AppendResults(“5Loaves HTTP Server Invoked me with [“);
     PIC.AppendResults(“] and [“);
     PIC.AppendResults(“] and [“);

from this URL:

or this HTML:
<!DOCTYPE HTML PUBLIC “-//W3C//DTD HTML 4.0 Transitional//EN”>
<HTML><HEAD><TITLE>United Business
<FORM method=post Action=“/Test2WWWPage”>
<P><INPUT name=One>
<P><INPUT name=Two>
<P><INPUT name=Three>
<INPUT type=submit value=Submit name=B1>
<INPUT type=reset value=Reset

This is the complete source for a C++ Messaging plugin to write bulk data to a
file and return a static message confirming receipt of the data.  While
receiving a file is no big deal, receiving it from outside the network through
the firewall is.


extern “C” __declspec(dllexport) void
CustomMessagingHandler(void *pHandle, DriverExec
Exec,const char *pData)

     PlugInController PIC(pHandle, Exec);

     int nArg1Size =

     FILE *fp = fopen(“copyOfMessage.cpy”,“wb”);
     if (fp)
nBytesIn = nArg1Size;
( 1 )
( PIC.GetExpectedDataSize() == PIC.GetTotalBytesIn() )

               // get the next chunk of the data

nDataSize = 0;
nRslt = PIC.NextDataChunk(&nDataSize);
(nRslt == -1)
     // the connection has been broken

               else if
(nRslt == 1)
     nBytesIn += nDataSize;
     fwrite(PIC.GetBuffer1(),1, nDataSize ,fp);
               else if
(nRslt == 0)
     // there is no more data here yet


     PIC << “Your message was received in full and processed by a custom message handler.”;



XMLFoundation for Java

XMLFoundation supports Java too, it supports all Java data
types like (string byte bool double long short) as well as Java data structures
such as (Vector Stack ArrayList LinkedList TreeSet HashSet ).  The XMLFoundation for Java is
binary.  If you want to build it yourself,
you can because the source code to the entire JavaXMLFoundation is public and
included in this release. 
There is no need to build it, but it’s nice to be able to.  I can hear some
uneducated Java programmer already saying "I only want a ‘Pure’ Java
solution".  This is Pure.  It’s as pure as the JVM, because if
you look close you’ll see that it is actually an enhancement to the JVM.

The JVM (Java Virtual Machine) is written in C and
C++.  Your Java code runs anywhere the JVM can compile, and the
XMLFoundation works the same way.  The good news is that Java programmers
don’t have to deal with C++ just because their JVM is written in C++.  The
same is true of the XMLFoundation for Java.

The XMLFoundation uses JNI (Java Native Interface). 
It parses the XML in the native binary (that was created by a C++ compiler just
like the JVM) and instantiates ‘pure’ Java Objects through JNI, then it assigns
all the member variables just like the C++ XMLFoundation does.  Can you
think of a faster way to get the job done?

This is some sample code that uses the JavaXMLFoundation. 
A much more detailed example is included in the source code:


// Java source code that shows how to use the XMLFoundation using

// inheritance or containment. It is nearly identical to C++

import java.util.Iterator;
import java.util.Vector;

// MyLineItem has an “ObjectId”, that is a value created from

// members and/or attributes of this object that uniquely identify

// it among all instances of it’s own type. “ObjectId”‘s are

// optional but allow you to perform complex “object updates”

// easily through XML. In main() below, this functionality is

// used when getXML2() is applied. ObjectId’s are like a database

// primary key in that they are generally not modified.

class MyLineItem extends XMLObject
     private String item;
     private String quantity;
     private int
          // call the base class constructor with the XML-tag for ‘this’

     MyLineItem(int nID,
String itm, String qty)
          item = itm;
          quantity = qty;
          ItemID = nID;

     void MapXMLTagsToMembers()
member      member        
xml-tag           (optional)wrapper


          MapMember(quantity,     “quantity”,   
"ItemID",      “SKU”);


          // MapObjectId() is optional.


          MapObjectId(this, “ItemID”); // takes 1 to 5 mapped ‘Key parts’


// Not “derived from” but “contains” XML support.

class Customer2
String          name;
     private int    
     private XMLObject     
     public MyOrder    
Vector         vecStrings;

     long l;
     short s;
     double d;
     byte b;
     boolean z;

     public void
ContainedXMLObj.toXML() );

     // indirect inheritance manages calls to FromXML() and ToXML()

     // manually, exposing them is optional. Member<-->XML exchange is

     // also done manually, generally following calls to FromXML() or ToXML()

     // this affords the developer full control for atypical implementations.

     void MyExchange(String
inOut, b,     “b”,“byte”,“”,
inOut, z,     “z”,“bool”,“”,
inOut, l,     “l”,“long”,“”,
inOut, d,     “d”,“double”,“”,
inOut, s,     “s”,“short”,“”,
inOut, name,     “name”,“FirstName”,“”,
inOut, CustID,     “CustID”,“CustomerID”,“”,
inOut, objOrder,     “objOrder”,     “Order”,“MyOrder”,“”);
inOut, vecStrings,“vecStrings”,“StringItem”,“String”“StringList Level2Wrapper”);

     public Customer2( String
strXML ) 
          // create a new ’empty’ object that maps to “Customer”.

          ContainedXMLObj = new
XMLObject(“Customer”, false
     // FromXML is not inheriterd, so we can expose the functionality through a

     // controlled accessor.

     public void
ApplyXML( String strXML )
// End of sample Java source code 

This is some of the code I developed inside the JavaXMLFoundation that
interacts with the JVM.  (Don’t worry you’ll never have to work with this
code )

jobject MakeObjectInstance(JNIEnv *env,const char
        DynamicXMLObject *pDO,DynamicXMLObject
     if (env->ExceptionOccurred())


     jclass clazzA = env->FindClass(pzObjectType);
     jobject objReturnValue = 0;
     GString strType("Ljava/lang/String;");

     // if this class type exposes a ‘ctor that takes a single XMLObject

     // we are using containment.

     jmethodID midctor = env->GetMethodID(clazzA, , "(LXMLObject;)V");
     if (env->ExceptionOccurred())

     if (midctor)
          // create a new java XMLObject instance

          jclass clazzX = env->FindClass("XMLObject");
          jmethodID midX =
     env->GetMethodID(clazzX, , "(Ljava/lang/String;)V");
tagX = env->NewStringUTF(pDO->GetObjectTag());
          jobject objX = env->NewObject(clazzA,
midX, tagX );

          // assign the object handle into the instance just created.

          jclass clazz = env->GetObjectClass(objX);
          jfieldID fid = env->GetFieldID(clazz, "oH", “I”);
          env->SetIntField(objX, fid,

          // create a new instance of some user defined java class that is

          // not extending
XMLObject. Pass the XMLObject to the ‘ctor.

          objReturnValue = env->NewObject(clazzA,
midctor, objX );

          // create an instance of a java object derived from the java XMLObject

          objReturnValue = env->AllocObject(clazzA);

          // assign the base class object handle directly.

          jclass clazz = env->GetObjectClass(objReturnValue);

          // any object that extends XMLObject will have the oH (Object Handle)

          // If the Object created is a String the fid will be 0.

          jfieldID fid = env->GetFieldID(clazz, "oH", “I”);
fid, CastDXMLO(pDO));

               // create the jobject to DXO index

     jobject   mbrObj;
     void *   

               Member.mbrObj =

cacheManager.addAlternate( pDO );
          else if
(strType.CompareNoCase(pzObjectType) != 0)
            GString Err;
            Err.Format(“Object type [%s] must either be derived from

            “or supply a constructor
%s(XMLObject o)”
     return objReturnValue;

Java programmers will derive from this
class – then the use is nearly identical to the C++ XMLFoundation:


import java.util.Vector;
import java.util.Stack;
import java.util.ArrayList;
import java.util.LinkedList;
import java.util.TreeSet;
import java.util.HashSet;

public class XMLObject {
     static { System.loadLibrary("JavaXMLFoundation"); }
     private static int InstanceId = 0;

     private native void JavaMap(int oH, int DataType,String strName,String xmlTag,String strWrapper, String ObjType, String strContainerType, int nSource);
     private native void JavaExchange(Object o, int oH, String inout,int nType,String b,String c,String d,String strObjectType, String strContainerType, int nSource);
     private native int JavaConstruct(int n, String strXMLTag, int bAutoDataSync);
     private native void JavaDestruct(int oH);
     private native void JavaMapCacheDisable(int oH);
     private native void JavaMapOID(int oH, Object o, String Key1, String Key2, String Key3, String Key4, String Key5);
     private native XMLObject JavaGetSubObj(int oH);
     private native void JavaFromXML(int oH, String strXML);
     private native String JavaToXML(int oH);
     private native void JavaRemoveAll(int oH);

     private int          oH; // XML-Object Handle

     public void MemberRemoveAll()

     protected void finalize()
     protected void freedom()

     public XMLObject(String strXMLTag)
          oH = ++InstanceId;
          oH = JavaConstruct(oH, strXMLTag, 1);
     public XMLObject(String strXMLTag, boolean bAutoDataSync)
          oH = ++InstanceId;
          if (bAutoDataSync)
               oH = JavaConstruct(oH, strXMLTag, 1);
               oH = JavaConstruct(oH, strXMLTag, 0);

     // Override this ‘virtual’ method to map member to XML tags
     // through calls to MapMember();
     private void MapXMLTagsToMembers()
     void DontCacheMemberMaps()
     void fromXML(String strXML)
          JavaFromXML(oH, strXML);
     String toXML()
          return JavaToXML(oH);

     private String MakeObjectName(String strIn)
          if (strIn.compareToIgnoreCase("String") == 0)
               strIn = "Ljava/lang/String;";
               String sTemp = "L" + strIn + ";";
               strIn = sTemp;
          return strIn;

     // MapAttrib for each native data type (1st without the optional ‘nested’ argument)
     public void MapAttrib(long z, String pzName, String pzXMLTag)
          JavaMap(oH,0,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(double z, String pzName, String pzXMLTag )
          JavaMap(oH,1,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(short z, String pzName, String pzXMLTag )
          JavaMap(oH,2,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(byte z, String pzName, String pzXMLTag)
          JavaMap(oH,3,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(String z, String pzName, String pzXMLTag)
          JavaMap(oH,4,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(int z, String pzName, String pzXMLTag )
          JavaMap(oH,5,pzName,pzXMLTag,"","", "", 2);
     public void MapAttrib(boolean z, String pzName, String pzXMLTag )
          JavaMap(oH,6,pzName,pzXMLTag,"","", "", 2);

     // MapAttrib for each native data type (now with the optional ‘nested’ argument)
     public void MapAttrib(long z, String pzName, String pzXMLTag, String pzNestedInTag)
          JavaMap(oH,0,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(double z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,1,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(short z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,2,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(byte z, String pzName, String pzXMLTag, String pzNestedInTag)
          JavaMap(oH,3,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(String z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,4,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(int z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,5,pzName,pzXMLTag,pzNestedInTag,"", "", 2);
     public void MapAttrib(boolean z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,6,pzName,pzXMLTag,pzNestedInTag,"", "", 2);

     // MapMember for each native data type (1st without optional nested-in-tag)
     public void MapMember(long z, String pzName, String pzXMLTag)
          JavaMap(oH,0,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(double z, String pzName, String pzXMLTag )
          JavaMap(oH,1,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(short z, String pzName, String pzXMLTag )
          JavaMap(oH,2,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(byte z, String pzName, String pzXMLTag)
          JavaMap(oH,3,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(String z, String pzName, String pzXMLTag )
          JavaMap(oH,4,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(int z, String pzName, String pzXMLTag )
          JavaMap(oH,5,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(boolean z, String pzName, String pzXMLTag )
          JavaMap(oH,6,pzName,pzXMLTag,"","", "", 1);
     public void MapMember(Object ob, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,7,pzName,pzXMLTag,"",MakeObjectName(strObjectType), "", 1);
     public void MapMember(Vector a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/Vector;", 1);
     public void MapMember(Stack a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/Stack;", 1);
     public void MapMember(ArrayList a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/ArrayList;", 1);
     public void MapMember(LinkedList a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/LinkedList;", 1);
     public void MapMember(TreeSet a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/TreeSet;", 1);
     public void MapMember(HashSet a, String pzName, String pzXMLTag, String strObjectType )
          JavaMap(oH,8,pzName,pzXMLTag,"",MakeObjectName(strObjectType),"Ljava/util/HashSet;", 1);

     // MapMember for each native data type (now with optional nested-in-tag)
     public void MapMember(long z, String pzName, String pzXMLTag, String pzNestedInTag)
          JavaMap(oH,0,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(double z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,1,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(short z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,2,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(byte z, String pzName, String pzXMLTag, String pzNestedInTag)
          JavaMap(oH,3,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(String z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,4,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(int z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,5,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(boolean z, String pzName, String pzXMLTag, String pzNestedInTag )
          JavaMap(oH,6,pzName,pzXMLTag,pzNestedInTag,"", "", 1);
     public void MapMember(Object ob, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,7,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType), "", 1);
     public void MapMember(Vector a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/Vector;", 1);
     public void MapMember(Stack a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/Stack;", 1);
     public void MapMember(ArrayList a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/ArrayList;", 1);
     public void MapMember(LinkedList a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/LinkedList;", 1);
     public void MapMember(TreeSet a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/TreeSet;", 1);
     public void MapMember(HashSet a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
          JavaMap(oH,8,pzName,pzXMLTag,pzNestedInTag,MakeObjectName(strObjectType),"Ljava/util/HashSet;", 1);

     // Map ObjectID’s – the JavaXMLFoundation works just the same with respect to object caching
    public void MapObjectId(Object o, String Key1)
          JavaMapOID(oH, o, Key1, "", "", "", "");
     public void MapObjectId(Object o, String Key1, String Key2)
          JavaMapOID(oH, o, Key1, Key2, "", "", "");
     public void MapObjectId(Object o, String Key1, String Key2, String Key3)
          JavaMapOID(oH, o, Key1, Key2, Key3, "", "");
     public void MapObjectId(Object o, String Key1, String Key2, String Key3, String Key4)
          JavaMapOID(oH, o, Key1, Key2, Key3, Key4, "");
     public void MapObjectId(Object o, String Key1, String Key2, String Key3, String Key4, String Key5)
          JavaMapOID(oH, o, Key1, Key2, Key3, Key4, Key5);

     // Get/Set Member for each native data type
     public void Member(Object o, String xfer, long z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource )
          JavaExchange(o,oH,xfer,0,pzName,pzXMLTag,pzNestedInTag,"","", nSource);
     public void Member(Object o, String xfer, double z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource )
     public void Member(Object o, String xfer, short z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource)
     public void Member(Object o, String xfer, byte z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource )
     public void Member(Object o, String xfer, String z, String pzName, String pzXMLTag, String pzNestedInTag , int nSource)
     public void Member(Object o, String xfer, int z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource )
     public void Member(Object o, String xfer, boolean z, String pzName, String pzXMLTag, String pzNestedInTag, int nSource )
     public void Member(Object o, String xfer, Object o2, String pzName, String pzXMLTag, String pzObjectType, String pzNestedInTag )
     public void Member(Object o, String xfer, Vector a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
     public void Member(Object o, String xfer, Stack a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag)
     public void Member(Object o, String xfer, ArrayList a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
     public void Member(Object o, String xfer, LinkedList a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
     public void Member(Object o, String xfer, TreeSet a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )
     public void Member(Object o, String xfer, HashSet a, String pzName, String pzXMLTag, String strObjectType, String pzNestedInTag )

About the
author(s) of XMLFoundation

The XMLFoundation was designed and developed mostly by myself,
however I had help from
co-designers on the NCIS (National Clinical Information System) Project, and
mostly only 1 other co-developer.  Special thanks to the SAIC management who
believed in me and gave me the opportunity to develop the mother product. 
Thanks to everyone who stood beside me during the long and expensive labor of
putting this work together.  "I told you so" to everyone
else.  In the end technology prevails over politics and ignorance.  The XMLFoundation contains
several independent algorithms that were developed by other authors – and their
supporting communities – their source code contains more information about
them.   Historically it has been the solo engineers that set
the direction for the masses.  Bjorne Stroustrup incremented the work of
Brian Kernighan and  Dennis Ritchie.  There are countless contributors
from the purest perspective of technology advancement and they could not
possibly all be named here.


In conclusion, this started a long time ago and it’s not going away anytime
soon.  It’s a masterpiece conspiracy to end times of high crimes. 

More by Author

Must Read