Method and apparatus for efficient management of XML documents

Abstract

A storage manager represents XML-compliant documents as a collection of objects, each associated with an element of the XML document. The objects are created with a hierarchy that conforms to the hierarchy of elements in the XML document. Access to the document is then through methods of the objects exposed to application programs through a programming interface. The methods provide a consistent interface to documents regardless of how the data associated with the documents is stored in memory. Consequently, the underlying content of XML documents can be stored using a distributed virtual object system that provides a mechanism for multiple processes to access each object. As an example of the benefits of such a storage manager, the storage manager may be used to maintain a telespace, such as may store information shared by multiple users participating in a collaboration session.

Description

FIELD OF THE INVENTION

This invention relates to storage and retrieval of information and, in particular, to storage and retrieval of information encoded in Extended Markup Language (XML).

BACKGROUND OF THE INVENTION

Modern computing systems are capable of storing, retrieving and managing large amounts of data. However, while computers are fast and efficient at handling numeric data they are less efficient at manipulating text data and are especially poor at interpreting human-readable text data. Generally, present day computers are unable to understand subtle context information that is necessary to understand and recognize pieces of information that comprise a human-readable text document. Consequently, although they can detect predefined text orderings or pieces, such as words in an undifferentiated text document, they cannot easily locate a particular piece of information where the word or words defining the information have specific meanings. For example, human readers have no difficulty in differentiating the word “will” in the sentence “The attorney will read the text of Mark's will.”, but a computer may have great difficulty in distinguishing the two uses and locating only the second such use.

Therefore, schemes have been developed in order to assist a computer in interpreting text documents by appropriately coding the document. Many of these schemes identify selected portions of a text document by adding into the document information, called “markup tags”, which differentiates different document parts in such a way that a computer can reliably recognize the information. Such schemes are generally called “markup” languages.

One of these languages is called SGML (Standard Generalized Markup Language) and is an internationally agreed upon standard for information representation. This language standard grew out of development work on generic coding and mark-up languages, which was carried out in the early 1970s. Various lines of research merged into a subcommittee of the International Standards Organization called the subcommittee on Text Description and Processing Languages. This subcommittee produced the SGML standard in 1986.

SGML itself is not a mark-up language in that it does not define mark-up tags nor does it provide a markup template for a particular type of document. Instead, SGML denotes a way of describing and developing generalized descriptive markup schemes. These schemes are generalized because the markup is not oriented towards a specific application and descriptive because the markup describes what the text represents, instead of how it should be displayed. SGML is very flexible in that markup schemes written in conformance with the standard allow users to define their own formats for documents, and to handle large and complex documents, and to manage large information repositories.

Recently, another development has changed the general situation. The extraordinary growth of the Internet, and particularly, the World Wide Web, has been driven by the ability it gives authors, or content providers, to easily and cheaply distribute electronic documents to an international audience. SGML contains many optional features that are not needed for Web-based applications and has proven to have a cost/benefit ratio unattractive to current vendors of Web browsers. Consequently, it is not generally used. Instead, most documents on the Web are stored and transmitted in a markup language called the Hypertext Markup Language or HTML.

HTML is a simple markup language based on SGML and it is well suited for hypertext, multimedia, and the display of small and reasonably simple documents that are commonly transmitted on the Web. It uses a small, fixed set of markup tags to describe document portions. The small number of fixed tags simplifies document construction and makes it much easier to build applications. However, since the tags are fixed, HTML is not extensible and has very limited structure and validation capabilities. As electronic Web documents have become larger and more complex, it has become increasingly clear that HTML does not have the capabilities needed for large-scale commercial publishing.

In order to address the requirements of such large-scale commercial publishing and to enable the newly emerging technology of distributed document processing, an industry group called the World Wide Web Consortium has developed another markup language called the Extensible Markup Language (XML) for applications that require capabilities beyond those provided by HTML. Like HTML, XML is a simplified subset of SGML specially designed for Web applications and is easier to learn, use, and implement than full SGML. Unlike HTML, XML retains SGML advantages of extensibility, structure, and validation, but XML restricts the use of SGML constructs to ensure that defaults are available when access to certain components of the document is not currently possible over the Internet. XML also defines how Internet Uniform Resource Locators can be used to identify component parts of XML documents.

An XML document is composed of a series of entities or objects. Each entity can contain one or more logical elements and each element can have certain attributes or properties that describe the way in which it is to be processed. XML provides a formal syntax for describing the relationships between the entities, elements and attributes that make up an XML document. This syntax tells the computer how to recognize the component parts of each document.

XML uses paired markup tags to identify document components. In particular, the start and end of each logical element is clearly identified by entry of a start-tag before the element and an end-tag after the element. For example, the tags <to> and </to> could be used to identify the “recipient” element of a document in the following manner:

• • document text . . . <to>Recipient</to> . . . document text.

The form and composition of markup tags can be defined by users, but are often defined by a trade association or similar body in order to provide interoperability between users. In order to operate with a predefined set of tags, users need to know how the markup tags are delimited from normal text and the relationship between the various elements. For example, in XML systems, elements and their attributes are entered between matched pairs of angle brackets (< . . . >), while entity references start with an ampersand and end with a semicolon (& . . . ;). Because XML tag sets are based on the logical structure of the document, they are easy to read and understand.

Since different documents have different parts or components, it is not practical to predefine tags for all elements of all documents. Instead, documents can be classified into “types” which have certain elements. A document type definition (DTD) indicates which elements to expect in a document type and indicates whether each element found in the document is not allowed, allowed and required or allowed, but not required. By defining the role of each document element in a DTD, it is possible to check that each element occurs in a valid place within the document. For example, an XML DTD allows a check to be made that a third-level heading is not entered without the existence of a second-level heading. Such a hierarchical check cannot be made with HTML. The DTD for a document is typically inserted into the document header and each element is marked with an identifier such as <!ELEMENT>.

However, unlike SGML, XML does not require the presence of a DTD. If no DTD is available for a document, either because all or part of the DTD is not accessible over the Internet or because the document author failed to create the DTD, an XML system can assign a default definition for undeclared elements in the document.

XML provides a coding scheme that is flexible enough to describe nearly any logical text structure, such as letters, reports, memos, databases or dictionaries. However, XML does not specify how an XML-compliant data structure is to be stored and displayed, much less efficiently stored and displayed. Consequently, there is a need for a storage mechanism that can efficiently manipulate and store XML-compliant documents.

SUMMARY OF THE INVENTION

In accordance with one embodiment of the invention, an in-memory storage manager represents XML-compliant documents as a collection of objects in memory. The collection of objects allows the storage manager to manipulate the document, or parts of the document with a consistent interface and to provide for features that are not available in conventional XML documents, such as element attributes with types other than text and documents that contain binary, rather than text, information. In addition, in the storage manager, the XML-compliant document is associated with a schema document (which is also an XML document) that defines the arrangement of the document elements and attributes. The storage manager can operate with conventional storage services to persist the XML-compliant document. Storage containers contain pieces of the document that can be quickly located by the storage manager.

In accordance with another embodiment, the storage manager also has predefined methods that allow it to access and manipulate elements and attributes of the document content in a consistent manner. For example, the schema data can be accessed and manipulated with the same methods used to access and manipulate the document content.

In accordance with yet another embodiment, the schema data associated with a document can contain a mapping between document elements and program code to be associated with each element. The storage manager further has methods for retrieving the code from the element tag. The retrieved code can then be invoked using attributes and content from the associated element and the element then acts like a conventional object.

In all embodiments, the storage manager provides dynamic, real-time data access to clients by multiple processes in multiple contexts. Synchronization among multiple processes accessing the same document is coordinated with event-driven queues and locks. The objects that are used to represent the document are constructed from common code found locally in each process. In addition, the data in the objects is also stored in memory local to each process. The local memories are synchronized by means of a distributed memory system that continually equates the data copies of the same element in different processes.

In still another embodiment, client-specified collections are managed by a separate collection manager. The collection manager maintains a data structure called a “waffle” that represents the XML data structures in tabular form. A record set engine that is driven by user commands propagates a set of updates for a collection to the collection manager. Based on those updates, the collection manager updates index structures and may notify waffle users via the notification system. The waffle user may also navigate within the collection using cursors.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and further advantages of the invention may be better understood by referring to the following description in conjunction with the accompanying drawings in which:

FIG. 1 is a schematic diagram of a computer system on which the inventive storage manager system can run.

FIG. 2 is a block schematic diagram illustrating the relationship of the in-memory storage manager and persistent storage.

FIG. 3 is a block schematic diagram illustrating the representation of an XML document on the storage manager memory as a collection of objects.

FIG. 4A is a block schematic diagram illustrating the components involved in binding code to XML elements.

FIG. 4B is a flowchart showing the steps involved in retrieving program code bound to an element.

FIG. 5 illustrates the relationship of XML text documents and binary sub-documents.

FIG. 6 is a block schematic diagram illustrating the major internal parts of the storage manager in different processes.

FIG. 7 illustrates the mechanism for synchronizing objects across processes.

FIG. 8 is an illustration that shows the major control paths from the storage manager APIs through the major internal parts of the storage manager.

FIG. 9 is an illustration of the storage manager interface constructed in accordance with an object-oriented implementation of the invention.

FIG. 10 is an illustration of the interfaces constructed in accordance with an object-oriented implementation of the invention, that are defined by the storage manager and may be called during the processing of links or element RPCs.

FIG. 11 is an illustration of the database and transaction interfaces constructed in accordance with an object-oriented implementation of the invention.

FIG. 12 is an illustration of the document and element interfaces constructed in accordance with an object-oriented implementation of the invention.

FIG. 13 is an illustration of the element communication and synchronization interfaces constructed in accordance with an object-oriented implementation of the invention.

FIG. 14 is an illustration that shows the major control paths from the collection manager APIs through the major internal parts of the collection and storage managers.

FIG. 15 is an illustration of the collection manager interfaces constructed in accordance with an object-oriented implementation of the invention.

DETAILED DESCRIPTION

FIG. 1 illustrates the system architecture for an exemplary client computer 100, such as an IBM THINKPAD 600®, on which the disclosed document management system can be implemented. The exemplary computer system of FIG. 1 is discussed only for descriptive purposes, however, and should not be considered a limitation of the invention. Although the description below may refer to terms commonly used in describing particular computer systems, the described concepts apply equally to other computer systems, including systems having architectures that are dissimilar to that shown in FIG. 1 and also to devices with computers in them, such as game consoles or cable TV set-top boxes, which may not traditionally be thought of as computers.

The client computer 100 includes a central processing unit (CPU) 105, which may include a conventional microprocessor, random access memory (RAM) 110 for temporary storage of information, and read only memory (ROM) 115 for permanent storage of information. A memory controller 120 is provided for controlling system RAM 110. A bus controller 125 is provided for controlling bus 130, and an interrupt controller 135 is used for receiving and processing various interrupt signals from the other system components.

Mass storage may be provided by diskette 142, CD-ROM 147, or hard disk 152. Data and software may be exchanged with client computer 100 via removable media, such as diskette 142 and CD-ROM 147. Diskette 142 is insertable into diskette drive 141, which is connected to bus 130 by controller 140. Similarly, CD-ROM 147 can be inserted into CD-ROM drive 146, which is connected to bus 130 by controller 145. Finally, the hard disk 152 is part of a fixed disk drive 151, which is connected to bus 130 by controller 150.

User input to the client computer 100 may be provided by a number of devices. For example, a keyboard 156 and a mouse 157 may be connected to bus 130 by keyboard and mouse controller 155. An audio transducer 196, which may act as both a microphone and a speaker, is connected to bus 130 by audio controller 197. It should be obvious to those reasonably skilled in the art that other input devices, such as a pen and/or tablet and a microphone for voice input, may be connected to client computer 100 through bus 130 and an appropriate controller. DMA controller 160 is provided for performing direct memory access to system RAM 110. A visual display is generated by a video controller 165, which controls video display 170.

Client computer 100 also includes a network adapter 190 that allows the client computer 100 to be interconnected to a network 195 via a bus 191. The network 195, which may be a local area network (LAN), a wide area network (WAN), or the Internet, may utilize general-purpose communication lines that interconnect multiple network devices.

Client computer system 100 generally is controlled and coordinated by operating system software, such as the WINDOWS NT™ operating system (available from Microsoft Corp., Redmond, Wash.). Among other computer system control functions, the operating system controls allocation of system resources and performs tasks such as process scheduling, memory management, networking and I/O services.

As illustrated in more detail in FIG. 2, the storage manager 206 resides in RAM 200 (equivalent to RAM 110 in FIG. 1) and provides an interface between an application program 202 which uses XML documents 228 and 230 and the persistent storage 208 in which the documents 228 and 230 are stored. The application 202 can interact with storage manager 206 by means of a consistent application programming interface 204 irregardless of the type of persistent storage 208 used to store the objects. Internally, the storage manager 206 represents each document 210, 218, as a hierarchical series of objects 212-216 and 220-224, respectively. The storage manager 206 can store the documents 210 and 218 in persistent storage 208 as schematically illustrated by arrow 226 using a variety of file systems, such as directory-based file services, object stores and relational file systems.

The inventive system operates with conventional XML files. A complete XML file normally consists of three components that are defined by specific markup tags. The first two components are optional, the last component is required, and the components are defined as follows:

• • 1. An XML processing statement which identifies the version of XML being used, the way in which it is encoded, and whether it references other files or not. Such a statement takes the form:
  • <?xml version=“1.0” encoding=“UTF-8” standalone=“yes”?>
  • 2. A document type definition (DTD) that defines the elements present in the file and their relationship. The DTD either contains formal markup tag declarations describing the type and content of the markup tags in the file in an internal subset (between square brackets) or references a file containing the relevant markup declarations (an external subset). This declaration has the form:
  • <!DOCTYPE Appl SYSTEM “app.dat”>
  • 3. A tagged document instance which consists of a root element, whose element type name must match the document type name in the document type declaration. All other markup elements are nested in the root element.

If all three components are present, and the document instance conforms to the document model defined in the DTD, the document is said to be “valid.” If only the last component is present, and no formal document model is present, but each element is properly nested within its parent elements, and each attribute is specified as an attribute name followed by a value indicator (=) and a quoted string, document instance is said to be “well-formed.” The inventive system can work with and generate well-formed XML documents.

Within the storage manager 206, XML documents are represented by means of data storage partitions which are collectively referred to by the name “Groove Document” to distinguish the representation from conventional XML documents. Each Groove document can be described by a DTD that formally identifies the relationships between the various elements that form the document. These DTDs follow the standard XML format. In addition, each Groove document has a definition, or schema, that describes the pattern of elements and attributes in the body of the document. XML version 1.0 does not support schemas. Therefore, in order to associate a Groove schema document with an XML data document, a special XML processing instruction containing a URI reference to the schema is inserted in the data document. This processing instruction has the form:

• • <?schema URI=“groovedocument:///GrooveXSS/$PersistRoot/sample.xml”?>

Some elements do not have, or require, content and act as placeholders that indicate where a certain process is to take place. A special form of tag is used in XML to indicate empty elements that do not have any contents, and therefore, have no end-tag. For example, a <ThumbnailBox> element is typically an empty element that acts as a placeholder for an image embedded in a line of text and would have the following declaration within a DTD:

• • <!ELEMENT ThumbnailBox EMPTY>

Where elements can have variable forms, or need to be linked together, they can be given suitable attributes to specify the properties to be applied to them. These attributes are specified in a list. For example, it might be decided that the <ThumbnailBox> element could include a Location and Size attributes. A suitable attribute list declaration for such an attribute would be as follows:

	<!ATTLIST ThumbnailBox
	Location	ENTITY	#REQUIRED
	Size	CDATA	#IMPLIED
	>

This tells the computer that the <ThumbnailBox> element includes a required Location entity and may include a Size attribute. The keyword #IMPLIED indicates that it is permissible to omit the attribute in some instances of the <ThumbnailBox> element.

XML also permits custom definition statements similar to the #DEFINE statements used with some compilers. Commonly used definitions can be declared within the DTD as “entities.” A typical entity definition could take the form:

• • <!ENTITY BinDoc3487 SYSTEM “./3487.gif” NDATA>which defines a file location for the binary document “BinDoc3487.” Once such a declaration has been made in the DTD, users can use a reference in place of the full value. For example, the <ThumbnailBox> element described previously could be specified as <ThumbnailBox Location=BinDoc3487 Size=“Autosize”/>. An advantage of using this technique is that, should the defined value change at a later time, only the entity declaration in the DTD will need to be updated as the entity reference will automatically use the contents of the current declaration.

Within the storage manager, each document part is identified by a Uniform Resource Identifier (URI) which conforms to a standard format such as specified in RFC 2396. URIs can be absolute or relative, but relative URIs must be used only within the context of a base, absolute URI. When the document is stored in persistent storage, its parts may be identified by a different STORAGEURI that is assigned and managed by the particular file system in use.

In accordance with the principles of the invention, within each document part, in the storage manager internal memory is represented by a collection of objects. For example, separate elements in the XML document are represented as element objects in the storage manager. This results in a structure that is illustrated in FIG. 3. In FIG. 3, an illustrative XML document 300 is represented as a collection of objects in storage manager 302. In particular, the XML document 300 contains the conventional XML processing statement 304 which identifies the XML version, encoding and file references as discussed above. Document 300 also contains an XML processing statement 306 which identifies a schema document 320 in storage manager 302 which is associated with the document 300. The illustrative XML document also contains a set of hierarchical elements, including ElementA 308 which contains some text 318, ElementA contains ElementB 310 which has no text associated with it. ElementB also contains ElementC 312, which, in turn, contains two elements. Specifically, ElementC contains ElementD 314 that has an attribute (ID, with a value “foo”) and ElementE 316.

In the storage manager 302, the elements, ElementA-ElementE, are represented as element objects arranged in a hierarchy. In particular, ElementA is represented by ElementA object 322. Each element object contains the text and attributes included in the corresponding XML element. Therefore, element object 322 contains the text 318. Similarly, ElementB 310 is represented by element object 324 and elements ElementC, ElementD and ElementE are represented by objects 326, 328 and 330, respectively. Element object 328, which represents element ElementD, also includes the attribute ID that is included in the corresponding element. Each element object references its child element objects by means of database pointers (indicated by arrows between the objects) into order to arrange the element objects into a hierarchy. There may also be attribute indices, such as index 332 that indexes the ID attribute in element object 328.

The representation of the XML document 300 by means of an object collection allows the storage manager 302 to manipulate its internal representation of the document 300 with a consistent interface that is discussed in detail below. The storage manager 302 can also provide features that are not available in conventional XML documents, such as collection services that are available via a collection manager that is also discussed in detail below.

As described above, Groove documents that contain XML data may have a definition, or schema document, that describes the pattern of elements and attributes in the body of the document. The schema document is stored in a distinct XML document identified by a URI. The schema document has a standard XML DTD definition, called the meta-schema, which is shown below:

<!-- The Document element is the root element in the schema -->
<!ELEMENT Document (Registry*, AttrGroup*, ElementDecl*)>
<!ATTLIST Document
URL	CDATA	#REQUIRED
>
<!ELEMENT Registry TagToProgID*>
<!ELEMENT TagToProgID EMPTY>
<!ATTLIST TagToProgID
Tag	CDATA	#REQUIRED
ProgID	CDATA	#REQUIRED
>
<!ELEMENT AttrGroup AttrDef*>
<!ELEMENT AttrDef EMPTY>
<!ATTLIST AttrDef
Name	CDATA	#REQUIRED
Type	CDATA	#REQUIRED
Index	CDATA	#IMPLIED
DefaultValue	CDATA	#IMPLIED
>
<!ELEMENT ElementDecl (ElementDecl* | AttrGroup | ElementRef*)>
<!ATTLIST ElementDecl
Name	CDATA	#REQUIRED
>
<!ELEMENT ElementRef EMPTY>
<!ATTLIST ElementRef
Ref	CDATA	#REQUIRED
>

Each of the elements in the schema defines information used by the storage manager while processing the document. The “Registry” section forms an XML representation of a two-column table that maps XML element tags to Windows ProgIDs. (In the Common Object Model (COM) developed by Microsoft Corporation, a ProgID is a text name for an object that, in the COM system, is “bound” to, or associated with, a section of program code. The mapping between a given ProgID and the program code, which is stored in a library, is specified in a definition area such as the Windows™ registry.)

This arrangement is shown in FIG. 4A that illustrates an XML document 400 and its related schema document 402. Both of these documents are resident in the storage manager 406 and would actually be represented by objects as shown in FIG. 3. However, in FIG. 4, the documents have been represented in conventional XML format for clarity. FIG. 4 shows the storage manager operational in a Windows™ environment that uses objects constructed in accordance with the Common Object Model (COM) developed by the Microsoft Corporation, Redmond, Wash., however, the same principles apply in other operating system environments.

XML document 400 includes the normal XML processing statement 414 that identifies the XML version, encoding and file references. A schema XML processing statement 416 references the schema document 402 which schema document is associated with document 400 and has the name “urn:groove.net:sample.xml” defined by name statement 426. It also includes a root element 418 which defines a name “doc.xml” and the “g” XML namespace which is defined as “urn:groove.net”

Document 400 has three other elements, including element 420 defined by tag “urn:groove.net:AAA”, element 422 defined by tag “urn:groove.net:BBB” and element 424 defined by tag “urn:groove.net:NoCode”. Element 424 is a simple element that has no corresponding bound code and no corresponding tag-to-ProgID mapping in the schema document 402.

Within the “registry” section defined by tag 428, the schema document 402 has two element-to-COM ProgID mappings defined. One mapping is defined for elements with the tag “urn:groove.net:AAA” and one for elements with the tag “urn:groove.net:BBB.” The bound code is accessed when the client application 404 invokes a method “OpenBoundCode( ).” The syntax for this invocation is given in Table 15 below and the steps involved are illustrated in FIG. 4B. Invoking the OpenBoundCode( ) method on a simple element, such as element 424 generates an exception. The process of retrieving the bound code starts in step 434 and proceeds to step 436 in which the OpenBoundCode( ) is invoked. Invoking the OpenBoundCode( ) method on an element with the element tag “urn:groove.net:AAA” causes the storage manager 406 to consult the registry element 428 in the schema document 602 with the element tag as set forth in step 438. From section 430, the storage manager retrieves the ProgID “Groove.Command” as indicated in step 440. In step 442, the storage manager calls the COM manager 408 in instructs it to create an object with this ProgID. In a conventional, well-known manner, in step 444, the COM manager translates the ProgID to a CSLID using a key in the Windows Registry 410. In step 446, the COM manager uses the CSLID to find a dynamically loadable library (DLL) file in the code database 412 that has the code for the object. Finally, in step 448, the COM manager creates the object and returns an interface pointer for the object to the storage manager 406 which, in turn, returns the pointer to the client application 404. The routine then finishes in step 450. The client application 404 can then use the pointer to invoke methods in the code that use attributes and content in the associated element. The element then behaves like any other COM object. A similar process occurs if the OpenBoundCode( ) method is invoked on elements with the tag “urn:groove.net:BBB.”

The “AttrGroup” section defines non-XML characteristics for attributes. An attribute's data type can be defined as some type other than text and the attribute may be indexed to facilitate fast retrieval of the elements that containing it.

The “ElementDecl” section provides a form of element definition similar to the DTD <!ELEMENT> declaration, but allows for extended attribute characteristics and the definition of non-containment element references.

The following example shows the sample portions of a schema document for an XML document that defines a “telespace” that is previously described.

<groove:Document URL=“TelespaceSchema.xml”
xmlns:groove=“urn:groove.net:schema.1”>
<groove:Registry>
<groove:TagToProgID groove:Tag=“g:Command”
groove:ProgID=“Groove.Command”/>
<groove:TagToProgID groove:Tag=“groove:PropertySetChanged”
groove:ProgID=“Groove.PropSetChangeAdvise”/>
</groove:Registry>
<groove:AttrGroup>
<groove:AttrDef Name=“ID” Index=“true”/>
<!-- KEY EXCHANGE ATTRIBUTES -->
<groove:AttrDef Name=“NKey” Type=“Binary”/>
<groove:AttrDef Name=“ReKeyId” Type=“String”/>
<groove:AttrDef Name=“T” Type=“String”/>
<!-- AUTHENTICATION ATTRIBUTES -->
<groove:AttrDef Name=“MAC” Type=“Binary”/>
<groove:AttrDef Name=“Sig” Type=“Binary”/>
<!-- ENCRYPTION ATTRIBUTES -->
<groove:AttrDef Name=“IV” Type=“Binary”/>
<groove:AttrDef Name=“EC” Type=“Binary”/>
<!-- XML Wrapper Attributes -->
<groove:AttrDef Name=“Rows” Type=“Long”/>
<groove:AttrDef Name=“Cols” Type=“Long”/>
<groove:AttrDef Name=“Items” Type=“Long”/>
<groove:AttrDef Name=“ItemID” Type=“Bool” Index=“true”/>
</groove:AttrGroup>
<groove:ElementDecl Name=“groove:Telespace”>
<AttrGroup>
<AttrDef Name=“Persist” DefaultValue=“True” Type=“Bool”/>
<AttrDef Name=“Access” DefaultValue=“Identity”
Type=“String”/>
</AttrGroup>
<ElementRef Element=“Dynamics”/>
<ElementRef Element=“Members”/>
</groove:ElementDecl>
</groove:Document>

In this example, there are two entries in the Tag to ProgID mapping table. The first maps the tag “g:Command” (which, using XML namespace expansion, is “urn:groove.net.schema.1:Command”) to the ProgID “Groove.Command.” In the section defining attributes, the “ID” attribute is indexed, the data type of the NKey attribute is binary, and so on.

This schema data is represented by element objects and can be accessed and manipulated by the same storage manager element and attribute interface methods used to manipulate documents as described in detail below. In particular, the information that describes a document can be manipulated using the same interfaces that are used for manipulating the document content.

In accordance with another aspect of the invention, sub-documents can be associated with a primary document. Any document may be a sub-document of a given document. If a document contains a sub-document reference to another document, then the referenced document is a sub-document. If two documents contain sub-document references to each other, then each document is a sub-document of the other document. Each sub-document is referenced from the primary document with conventional XML XLink language, which is described in detail at website www.w3.org/TR/xlink. Links may also establish a relationship between an all-text XML document and a binary sub-document. Binary documents do not have links to any kind of sub-document. If the link is to a document fragment, a subdocument relationship is established with the document that contains the fragment. The relationship of documents and sub-documents is illustrated in FIG. 5.

For example, main document 500 contains links 502 which include a link, represented by arrow 510, to document 504 and a link, represented by arrow 508, to a binary document 506. Documents 504 and 506 are thus sub-documents of document 500. Document 504, in turn, contains links 512 which include a link, represented by arrow 514 to document 516 with content 518. Document 516 is a sub-document of document 500. Document 506 contains binary content 520 and, therefore, cannot have links to sub-documents.

Sub-document links follow the standard definition for simple links. An exemplary element definition of a link is as follows:

	<!ELEMENT GrooveLink ANY>
	<!ATTLIST GrooveLink
	xml:link	CDATA	#FIXED “simple”
	href	CDATA	#REQUIRED
	role	CDATA	#IMPLIED “sub-document”
	title	CDATA	#IMPLIED
	show	(parsed|replace|new) #IMPLIED
	actuate	(auto|user)	#IMPLIED
	serialize	(byvalue|byreference|ignored) #IMPLIED
	behavior	CDATA	#IMPLIED
	content-role	CDATA	#IMPLIED
	content-title	CDATA	#IMPLIED
	inline	(true|false)	#IMPLIED “true”
	>

It is also possible to establish a sub-document relationship without using the above definition by adding to a document an XML link which has an xml:link attribute with a value “simple”, and a href attribute. Such a link will establish a sub-document relationship to the document identified by a URI value in the href attribute.

Given the relationships from a document to its sub-documents, it is possible to make a copy of an arbitrary set of documents and sub-documents. Within a single storage service, it may be possible to directly perform such a copy. To cross storage services or to send multiple documents to another machine, the entire hierarchy of such documents must be “describable” in a serialized fashion. The inventive Storage Manager serializes multiple documents to a text representation conforming to the specification of MIME Encapsulation of Aggregate documents, such as HTML (MHTML) which is described in detail at website ftp.isi.edu/in-notes/rfc2557.txt.

The following data stream fragment is an example of a document and a referenced sub-document as they would appear in an MHTML character stream. In the example, “SP” means one space is present and “CRLF” represents a carriage return-line feed ASCII character pair. All other characters are transmitted literally. The MIME version header has the normal MIME version and the Groove protocol version is in a RFC822 comment. The comment is just the word “Groove” followed by an integer. The boundary separator string is unique, so a system that parsed the MIME, and then each body part, will work correctly. The serialized XML text is illustrated in UTF-8 format, but it could also be transmitted in WBXML format. The XML document has a XML prefix, which includes the version and character encoding. The binary document is encoded in base64.

MIME-Version: SP 1.0 SP (Groove SP 2) CRLF
Content-Type: SP multipart/related; SP boundary=“<<[[&&&]]>>” CRLF
CRLF
--<<[[&&&]]>>Content-Type: SP text/XML; SP charset=“UTF-8”
<?xml version=“1.0” encoding=‘utf-8’?>
<rootelement>
.
.
.
</rootelement> CRLF
CRLF
--<<[[&&&]]>>
Content-ID: SP <URI> CRLF
Content-Type: SP application/octet-stream CRLF
Content-Transfer-Encoding: base64 CRLF
CRLF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==
--<<[[&&&]]>>--

Unlike most XML processors, such as document editors or Internet browsers, the storage manager provides for concurrent document operations. Documents may be concurrently searched, elements may be concurrently created, deleted, updated, or moved. Copies of element hierarchies may be moved from one document to another. In most XML processors, all of the updates to a document are driven by a single user, who is usually controlling a single thread within a single process on a single computer.

The storage manager maintains XML document integrity among many users updating the same document, using multiple threads in multiple processes. In a preferred embodiment, all of the updates occur on a single computer, but, using other different, conventional inter-processor communication mechanisms, other operational embodiments are possible. FIG. 6 shows the basic structure of the storage manager and illustrates how it isolates application programs from cross-process communication issues. For example, two separate processes 600 and 602 may be operating concurrently in the same computer or in different computers. Process 600 is a “home” process as described below, while process 602 is another process designated as Process N. Within process 600, a multi-threaded client application program 606 is operating and within process 602, a multi-threaded client application program 616 is operating.

Each application program 606 and 616 interfaces with a storage manager designated as 605 and 615, respectively. In process 600, the storage manager comprises a storage manager interface layer 608 which is used by application program 608 to control and interface with the storage manager. It comprises the database, document, element and schema objects that are actually manipulated by the application. The API exported by this layer is discussed in detail below. The storage manager 605 also includes distributed virtual object (DVO) database methods 610, DVO methods for fundamental data types 612, DVO common system methods 609 and distributed shared memory 614. Similarly, the storage manager operating in process 602 includes transaction layer 618, DVO database methods 620, DVO methods for fundamental data types 622, DVO common system methods 617 and distributed shared memory 624.

The two processes 600 and 602 communicate via a conventional message passing protocol or inter-process communication (IPC) system 604. For processes that run in a single computer, such a system can be implemented in the Windows® operating system by means of shared memory buffers. If the processes are running in separate computers, another message passing protocol, such as TCP/IP, can be used. Other conventional messaging or communications systems can also be used without modifying the operation of the invention. However, as is shown in FIG. 6, application programs 606 and 616 do not directly interact with the message passing system 604. Instead, the application programs 606 and 616 interact with storage managers 605 and 615, respectively, and storage managers 605 and 615 interact with the message passing system 604 via a distributed shared memory (DSM) system of which DSM systems 614 and 624 are a part.

A number of well-known DSM systems exist and are suitable for use with the invention. In accordance with a preferred embodiment, the DSM system used with the storage manager is called a C Region Library (CRL) system. The CRL system is an all-software distributed shared memory system intended for use on message-passing multi-computers and distributed systems. A CRL system and code for implementing such as system is described in detail in an article entitled “CRL: High-Performance All-Software Distributed Memory System”, K. L. Johnson, M. F. Kaashoek and D. A. Wallach, Proceedings of the Fifteenth Symposium on Operating Systems Principles, ACM, December 1995; and “CRL version 1.0 User Documentation”, K. L. Johnson, J. Adler and S. K. Gupta, , MIT Laboratory for Computer Science, Cambridge, Mass. 02139, August 1995. Both articles are available at web address www.pdos.lcs.mit.edu/crl.

Parallel applications built on top of the CRL, such as the storage manager, share data through memory “regions.” Each region is an arbitrarily sized, contiguous area of memory. Regions of shared memory are created, mapped in other processes, unmapped, and destroyed by various functions of the DSM system. The DSM system used in the present invention provides a super-set of the functions that are used in the CRL DSM system. Users of memory regions synchronize their access by declaring to the DSM when they need to read from, or write to, a region, and then, after using a region, declaring the read or write complete. The effects of write operations are not propagated to other processes sharing the region until those processes declare their need for it. In addition to the basic shared memory and synchronization operations, DSM provides error handling and reliability with transactions. The full interface to inventive DSM is shown in Table 1.

TABLE 1
DSM Method                       Description
AddNotification(DSMRgn* i_pRgn, const Adds a local event that will be signaled
IgrooveManualResetEvent * i_pEvent); with the data in the region changes.
Close( );                        Shuts down the DSM. There must be no
                                 mapped regions at this client.
Create(UINT4 i_Size, INT4        Creates a new region. It also atomically
i_CallbackParam, INCAddress      maps the new region and initiates a
i_InitialOwner, DSMRId & io_RId, StartWrite on the new region if Size is
DSMRgn * & o_pRgn, void * & o_pData); non-zero. Size is the initial size of the
                                 data in the new region. RId is identifier of
                                 the new region. pRgn is the new region if
                                 Size is non-zero.
AddDatabase(UINT2 i_DatabaseNumber); Adds a new database to the region
                                 mapping tables.
DatabaseFlushNotify(UINT2        Cleans up unused region resources.
i_DatabaseNumber, TimeMillis
i_StartTime);
Destroy(DSMRId& i_RId);          Destroys an existing region entirely. RId
                                 is a valid identifier of the region to be
                                 destroyed.
EndRead(DSMRgn* i_pRgn);         Closes a read operation on the region's
                                 data. pRgn is the valid region.
EndWrite(DSMRgn* i_pRgn);        Closes a write operation on the region's
                                 data. pRgn is the valid region.
Flush(DSMRgn* i_pRgn);           Flushes the region from this client's local
                                 cache to the region's home client. pRgn is
                                 the valid region.
GetSize(DSMRgn* i_pRgn);         Returns the size(number of bytes) of the
                                 given valid region. pRgn is the valid
                                 region.
Init(CBSTR i_BroadcastGroup,     Initializes the DSM. BroadcastGroup is
DSMRgnMapCallback * i_pCallback = the name of the group in which this DSM
NULL, void * i_pCallbackParam = NULL, client belongs. URCSize is the size of the
BOOL * o_pMasterClient = NULL, UINT4 Unmapped Regions Cache. PAddress is
i_WaitTimeOut = 1000,            the Inter-node Communication Address of
UINT4 i_URCSize = 1<<10,         this DSM client. pMasterClient specifies
INCAddress * o_pAddress = NULL); whether this DSM client is the
                                 Master(First) client.
Map(const DSMRId& i_RId, INT4    Maps the region to this client's memory
i_CallbackParam, BOOL i_InitialOwner); space. RId is a valid identifier of the
                                 region to be mapped.
RemoveDatabase(UINT2             Removes the specified database from the
i_DatabaseNumber);               region mapping tables.
RemoveNotification(DSMRgn* i_pRgn, Removes interest in changes to data in a
const IGrooveManualResetEvent *  region.
i_pEvent);
Resize(DSMRgn* i_pRgn, UINT4 i_Size); Resizes the given valid region while
                                 maintaining the original data(which may
                                 be truncated if the size is decreased).
                                 pRgn is the valid region. Size is the new
                                 size.
GetRId(const DSMRgn* i_pRgn);    Returns the identifier for the given valid
                                 region. pRgn is the valid region.
SignalNotification(DSMRgn* i_pRgn); Sets the signal that notification has
                                 occurred.
StartRead(DSMRgn* i_pRgn, INT4   Initiates a read operation on the region's
i_CallbackParam, void * & o_pData); data. RgnStartRead (or RgnStartWrite)
                                 must be called before the data can be
                                 read. pRgn is the valid region.
StartTransactionRead(DSMRgn* i_pRgn, Initiates a transactional read operation on
INT4 i_CallbackParam, void * & o_pData); the region's data. RgnStartRead (or
                                 RgnStartWrite) must be called before the
                                 data can be read. pRgn is the valid
                                 region.
StartTransactionWrite(DSMRgn* i_pRgn, Initiates a transactional write operation on
INT4 i_CallbackParam, void * & o_pData); the region's data. RgnStartWrite must be
                                 called before the data can be modified.
                                 pRgn is the valid region.
StartWrite(DSMRgn* i_pRgn, INT4  Initiates a write operation on the region's
i_CallbackParam, void * & o_pData); data. RgnStartWrite must be called before
                                 the data can be modified. pRgn is the
                                 valid region.
Unmap(DSMRgn* & io_pRgn);        Unmaps the region from this client's
                                 memory space. pRgn is the valid region
                                 to be unmapped.

Each storage manager 605 and 615 comprises a DSM node that uses one or more DSM regions (not shown in FIG. 6) located in the address space of the corresponding process 600, 602. These regions contain DVO objects and classes that can be used to represent documents, elements and schema of the XML data that is managed by the storage manager. Portions of documents, usually elements and index sections, are wholly contained within a region. Although the DSM system provides a conceptually uniform node space for sharing regions, there are issues that result in the need to single out a specific node or process to perform special tasks.

Consequently, within the DSM synchronization protocol, a single node is identified as a “home node” for each region. Within the many processes running the storage manager on a single computer, one process, called the “home process”, is the process that performs all disk I/O operations. To reduce the amount of data movement between processes, the home process is the home node for all regions. Other implementations are possible, in which any node may be the home for any region and any process may perform disk I/O. However, for personal computers with a single disk drive, allowing multiple processes to perform disk I/O introduces the need for I/O synchronization while not alleviating the main performance bottleneck, which is the single disk.

In accordance with the DSM operation, if a process has the most recent copy of a region, then it can read and write into the region. Otherwise, the process must request the most-recent copy from the home process before it can read and write in the region. Each DSM system 614, 624 interfaces with the message passing system 604 via an interface layer called an internode communication layer (615, 625 ) which isolates the DVM system from the underlying transport mechanism. It contains methods that send messages to a broadcast group, and manipulate addresses for the corresponding process and the home process.

The inventive storage manager uses shared objects as the basis for XML objects. Many systems exist for sharing objects across processes and computers. One such object-sharing model is based on the use of the shared memory facilities provided by an operating system. One of the biggest drawbacks of such a shared memory model is unreliability due to memory write failures that impact the integrity of other processes. For example, if one process is in the process of updating the state of an object and the process fails before setting the object to a known good state, other processes will either see the object in an invalid state or may blocked indefinitely waiting for the failed process to release its synchronization locks. The shared memory model also suffers from the locality constraints of shared memory in a tightly coupled multi-computer—it provides no way to share objects over a network.

Another model that provides distributed object sharing and remote method invocation is the basis for the distributed object management facilities in Java or the Object Management Group's CORBA system. Although providing the ability to share objects over a computer network, clients of such systems need to be aware of whether an object is local or remote—objects are not location independent. Performance is another drawback of this approach. All operations on an object need to be transmitted to the object server, since the server contains the only copy of the object state and serves as the synchronization point for that data.

In order to overcome these drawbacks, the inventive storage manager uses a distributed virtual object (DVO) system to provide the primitive data types that XML object types are built upon. The DVO system also provides its callers with the illusion that all data is reliably contained in one process on a single computer node, even though the data may be in multiple processes on many computers or may truly be just in one process on a single computer node.

The DVO object-sharing model is shown in FIG. 7. All processes, on all computers, that are sharing an object have the same method code. For example, process 700 and process 702 in FIG. 7 have copies of the same object. Thus, each of processes 700 and 702 has a copy of the same method code 704 and 706 in the respective process address space. The volatile data state for an object is stored in DSM regions. Thus, the object data 708 for the object copy in process 700 is stored in region 710 in the address space of process 700. Similarly, the object data 712 for the object copy in process 702 is stored in region 714 in the address space of process 702. Object methods synchronize their access to the object's data by using the DSM synchronization functions that synchronize the regions as illustrated by arrow 716. In this manner, DVO objects are location independent, failures are contained within a single process, and multiple changes to a local object do not require data movement across the inter-node transport.

The DVO system provides basic objects that may be used as building blocks to manage XML documents for the storage manager and is divided into three functional pieces. The DVO database 610 contains objects that handle the DVO local context in each process and the shared tables that contain information about open databases and documents contained within those databases. In DVO, “databases” are conceptual storage containers and may channel objects that are ultimately stored in any kind of storage service 609. DVO documents are associated with XML or binary documents, which are visible to a client of the storage manager. DVO documents are also used to contain the indices and metadata associated with a collection.

DVO types 612 is a set of object classes that can be used within DVO documents to implement higher-level data model constructs. DVO types range from simple data containment objects through complex, scalable index structures. Each DVO type is implemented with two classes—one is a “non-shared class” that uses memory pointers in object references and the other is a “shared class” that uses logical addresses, called database pointers, for object references. The “shared class” has two sub-forms—one is the representation of the object in a shared DSM region and the other is the representation of the object stored on-disk in an object store database. The DVO system 607 provides methods to transfer objects between their shared and non-shared implementations.

The different DVO types are shown in Table 2.

TABLE 2
DVO Type               Description
Binary Document        A kind of document that handles binary data.
B-tree Index           The type of the root of a b-tree index. It contains a
                       description of the index, as well as the address of the root
                       index node.
Btree Node             A piece of a Btree index which can contain variable numbers
                       of records, sorted by one or more keys.
Collection Document    A kind of document that handles Collection documents. In
                       addition to the Document methods, it has methods to handle
                       the collection descriptor, indices within the collection, and
                       read marks.
Document               The base type from which the other document types inherit
                       common methods, such as Open, Close, Create, and Write.
Extendible Hashing     A type implementation of extendible hashing, as defined in
                       “Extendible Hashing - A Fast Access Method for Dynamic
                       Files”, Ronald Fagin, Jürg Nievergelt, Nicholas Pippenger, H.
                       Raymond Strong. ACM Transactions on Database Systems
                       4(3), pages 315-344, 1979.
FlatCollectionDocument A specific kind of CollectionDocument used in shared
                       regions.
FlatDocument           A specific kind of XMLDocument used in shared regions.
FlatNode               A specific kind of Node used in shared regions.
Node                   The type used to store XML elements. It has methods to
                       manage the element name, the element's parent, element
                       content, element attributes, links to other elements, and
                       change notifications.
Ordered Bucket         A kind of index which supports key ordered sorting (integer,
                       double, string)
Ordered Index          A type that provides a collated data vector. It has methods
                       for adding, removing, and changing key/data pairs,
                       managing index cursors, and managing parent and sub-
                       indicies.
Ordered Index Types    Data types, called records and fields, that can be stored in
                       ordered indices.
Ordinal Ordered Index  A kind of index that support ordinal addressing. It is
                       conceptually similar to vector that allows any entry to be
                       addressed by position (e.g., vec[14]). In addition to the index
                       methods, it has methods to move entries to specific positions
                       within the index.
Red-Black Index        A kind of ordered index that implements balancing using the
                       red-black binary tree algorithm.
W32BinaryDocument      A specific kind of binary document for 32-bit Windows
                       platforms.
XML Document           A kind of document that handles XML documents. In
                       addition to the Document methods, it has methods to handle
                       schemas and indexes.

The DVO system 607 objects isolate the upper levels of DVO from physical storage and process locality issues. The DVO system objects use DSM for invoking and handling requests to and from the home process. Requests include operations such as opening, closing, and deleting a database, finding documents in a database, and opening, closing, deleting, and writing database documents. The DVO system 607 in the master process 600 can also retrieve DVO objects from a storage service 609. A storage service, such as service 609, is a utility program that stores and retrieves information from a persistent medium and is responsible for the physical integrity of a container, database or file. It ensures that all updates are durable and that all internal data structures (e.g., redirection tables, space allocation maps) are always consistent on disk. Other processes, such as process 602 cannot access the storage service 609 directly, but can access the system indirectly via its DSM regions 624.

The storage manager 605 can operate with different types of physical storage systems, including container or object stores, stream file systems and ZIP files. In order to achieve atomic commits, the object store storage service can be implemented using page-oriented input/output operations and a ping-pong shadow page table.

Individual storage manager methods are atomic. Multiple storage manager operations, even operations on different documents, may be grouped into “transactions.” Transactions not only protect XML data integrity, but they also improve performance because they enable the storage manager to reduce the number of region lock operations and reduce the amount of data movement over the message passing system.

The storage manager supports both read-write and read-only transactions built on DSM synchronization primitives described in the DSM documentation referenced above, which primitives insure consistency in multiple processes or computers. Read-write transactions provide for the atomicity and consistency of a set of database read and write operations. Each region that is changed as part of a transaction will be kept in a “locked” state until the transaction is committed or aborted. This prevents operations that are not part of the transaction from seeing the changes. Further, each transaction stores a “before image” of the regions it modifies so that, if the transaction is aborted (as a result of an explicit API call or an exception), the effects of the transaction can be undone. Depending on the performance requirements, an alternative implementation would write undo information rather than storing the full “before image.” A read-only transaction uses the same interface as a read-write transaction. A read-only transaction ensures that multiple read operations are consistent. Like other transactions, it uses DSM functions to keep all read regions in a “read state” until it is finished.

In addition, checkpoints can be used to ensure that changes are persistent and provide durability for storage manager operations. A checkpoint may be performed at any time. Checkpoints are used in conjunction with data recovery logging. All operations write “redo” information to a sequential recovery log file when they are committed. When the checkpoint is committed, the recovery log file will be flushed to persistent storage and will ensure that the operations can be recovered. Since transactions do not write “redo” information until they are committed, if a checkpoint operation is commenced in the middle of a transaction, the transaction operations will not be flushed.

Transactions are scoped to a thread and a database. Once a transaction is started on a thread for a particular database, that transaction will be automatically used for all subsequent storage manager operations on that database and thread. An extension of conventional operating system threads is used, so that transactions correctly handle calls that need to be marshaled to other threads, for example, a user interface thread, using the Groove system's simple marshaler. Storage manager calls made on a thread and database that doesn't have a transaction started will cause the storage manager to create a “default transaction” that will be committed just before the call ends. Alternatively, starting a new transaction on a thread and database that already has an existing transaction in progress will cause the new transaction to automatically “nest” in the existing transaction. Nested transactions provide the ability to roll back the system within the outer transaction. In particular, inner, nested transactions are not finally committed until the outermost transaction is committed. For example, if a nested transaction is committed, but the containing transaction is later aborted, the nested transaction will be aborted.

In a preferred embodiment of the invention, the storage manager is implemented in an object-oriented environment. Accordingly, both the storage manager itself and all of the document components, such as documents, elements, entities, etc. are implemented as objects. These objects, their interface, the underlying structure and the API used to interface with the storage manager are illustrated in FIG. 8. The API is described in more detail in connection with FIGS. 9-11. Referring to FIG. 8, the storage manager provides shared access to documents, via the document manipulation API 802, but, in order to enable a full programming model for client applications, additional communication and synchronization operations are provided, within the context of a document. For example, the storage manager provides queued element operations, which enable one process to send an element to another process via the Queue API 804. Elements can be sent by value (a copy of the whole element) or by reference to the element. Synchronization operations are also provided to allow one or more threads to wait for an element to be enqueued to a given queue. The storage manager also provides RPC-style element communication and synchronization, via the RPC API 804.

Other client components may need to be aware of when documents are created in or deleted from storage manager. Accordingly, the storage manager provides an interface to an interest-based notification system for those client components via notification API 800. The notification system 806 provides notifications to client components that have registered an interest when a document is created or deleted.

Document data is represented by a collection of objects including database objects, document objects, element objects and schema objects 808. The objects can be directly manipulated by means of the document manipulation API 802.

The document related objects 808 are actually implemented by the distributed virtual object system 810 that was discussed in,detail above. The distributed virtual object system 810 can also be manipulated by element queue and RPC objects 812 under control of the queue and RPC API 804.

The distributed virtual object system 810 communicates with the distributed shared memory via interface 814 and communicates with the logging operations via interface 816. Similarly, the distributed virtual object system can interact with the storage services via interface 818.

The following is a description of the interfaces for each of the objects used to implement a preferred embodiment of the inventive storage manager. These object are designed in accordance with the Common Object Model (COM) promulgated by Microsoft Corporation, Redmond, Wash., and can be manipulated in memory as COM objects. However, COM is just one object model and one set of interface methodologies. The invention could also be implemented using other styles of interface and object models, including but not limited to the Java and CORBA object models.

FIG. 9 illustrates object interfaces for a storage manager object. An interface 900 (IGrooveStorageManager) encapsulates the basic framework for the storage manager. This interface is a subclass of an IDispatch interface which is a common class defined by the COM model. Table 3 defines the methods included in the storage manager interface.

TABLE 3
Interface IGrooveStorageManager : IDispatch
CreateDatabase (BSTR             Creates a database. A database can be
i_DatabaseURI, VARIANT_BOOL      either temporary or permanent, and single or
i_Temporary, VARIANT_BOOL        multi-process. The DatabaseURI specifies
i_SingleProcess, IUnknown *      the location of the database.
i_pSecurityContext, VARIANT_BOOL
i_CreateOnCheckpoint,
IgrooveDatabase ** o_ppDatabase);
CreateOrOpenDatabase (BSTR       Creates a new database or opens an existing
i_DatabaseURI, VARIANT_BOOL      database.
i_Temporary, VARIANT_BOOL
i_SingleProcess, IUnknown *
i_pSecurityContext, VARIANT_BOOL
i_CreateOnCheckpoint,
VARIANT_BOOL * o_pCreated,
IgrooveDatabase ** o_ppDatabase);
CreateTemporaryElement (BSTR     Creates a temporary element.
i_Name, Iunknown * i_pParent,
IgrooveElement ** o_ppElement);
CreateTemporaryXMLDocument       Creates an empty temporary document with a
(BSTR i_NamePrefix, BSTR         unique URI
i_SchemaURI, IUnknown*
i_pAdditionalSchemaURIs,
IgrooveXMLDocument **
o_ppXMLDocument);
CreateTransform (BSTR            Creates a transformation interface.
i_CollectionDescriptorURI, BSTR
i_SecondaryDescriptorURI, BSTR
i_CollectionDescriptorName,
IgrooveTransform ** o_ppTransfom);
DeleteDatabase (BSTR             Deletes a database.
i_DatabaseURI);
IsHomeProcess (VARIANT_BOOL *    Determine whether we are the home process
o_pHomeProcess);
OpenCrossProcessSemaphore (BSTR  Creates a semaphore object that can be used
i_Name, VARIANT_BOOL i_Reentrant, to synchronize activity in different processes.
IgrooveCrossProcessSemaphore **  If the semaphore is not Reentrant, repeated
o_ppSemaphore);                  attempts to lock the semaphore within the
                                 same thread and process will block.
OpenDatabase (BSTR i_DatabaseURI, Open an existing database.
VARIANT_BOOL i_SingleProcess,
Iunknown * i_pSecurityContext,
IgrooveDatabase ** o_ppDatabase);
OpenDatabaseURIEnum(IGrooveBST   Returns an Enumeration of the databases that
REnum ** o_ppDatabaseURI);       are currently open.

Another interface 902 (IGrooveStorageURISyntax) is used by a client of a storage manager that needs to perform operations on parts of standard names, which are in the form of Uniform Resource Identifiers (URIs). Table 4 includes the methods for the IGrooveStorageURISyntax interface.

TABLE 4
Interface IGrooveStorageURISyntax : IDispatch
BuildDatabaseURI (BSTR           Builds a database URI from its pieces.
i_ServiceName, BSTR
i_DatabasePath, VARIANT_BOOL
i_Relative, BSTR *o_pURI);
BuildDocumentURI (BSTR           Builds a document URI from its pieces.
i_ServiceName, BSTR
i_DatabasePath, BSTR
i_DocumentName, VARIANT_BOOL
i_Relative, BSTR * o_pURI);
MakeAbsolute (BSTR i_RelativeURI, Given a relative URI within the scope of this
BSTR * o_pAbsoluteURI);          database, return an absolute URI.
MakeRelative (BSTR i_AbsoluteURI, Given an absolute URI within this database,
BSTR * o_pRelativeURI);          return a relative URI within the scope of this
                                 database.
OpenDatabasePath (BSTR I_URI,    Returns the directory path portion of a URI.
BSTR * o_pDatabasePath);
OpenDocumentName (BSTR i_URI,    Returns the document name portion of a URI.
BSTR * o_pDocumentName);
OpenPersistRootPath (BSTR *      Returns the directory path to the root of the
o_pPath);                        Groove persistent data directories.
OpenServiceName (BSTR i_URI,     Returns the storage service portion of a URI.
BSTR * o_pServiceName);
Parse (BSTR i_URI BSTR *         Parses the pieces of the given URI.
o_pServiceName, BSTR *
o_pDatabasePath, BSTR *
o_pDocumentName);

FIG. 10 illustrates the notification system interfaces. Interface 1000 (IGrooveLinkCallback) is an interface for use by a client of a storage manager that needs to be notified during the input processing of XML document or element when a definition for a link is found. The interface includes the methods defined in Table 5.

TABLE 5
Interface IGrooveLinkCallback : IDispatch
HandleLink (IGrooveElement *     Called when the specified
i_pLinkElement, IGrooveByteInputStream * element contains a
i_pLinkData);                    link attribute definition.

Another interface 1002 (IGrooveRPCServerCallback) is used by a client of a storage manager that needs to handle remote procedure calls (RPCs) on elements within XML documents. RPC server callbacks are a sub-class of the “util” base class (described below), that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveRPCServerCallback. Table 6 defines the methods used in the storage manager RPC server callback interface.

TABLE 6
Interface IGrooveElementRPCServerCallback : IDispatch
HandleCall (IGrooveElement * i_pInput, Handle a RPC, receiving
IgrooveElement ** o_ppOutput);   input parameters in the
                                 Input element and returning
                                 output parameters in the
                                 Output element.

FIGS. 11, 12 and 13 illustrate the document manipulation interfaces and the queue and RPC interfaces. In particular, FIG. 11 shows the interfaces used to manipulate databases. An interface 1100 (IGrooveDatabase) is used by a client of a storage manager that needs to manage the databases in which documents are stored. It includes the methods in Table 7.

TABLE 7
Interface IGrooveDatabase : IDispatch
Checkpoint ( );                  Creates a durable point of state for the
                                 database.
ClearDataLost ( );               Clears the database flag that indicates data
                                 may have been lost since the database was
                                 opened or the last transaction was
                                 committed.
CreateBinaryDocumentFromStream   Creates a binary document with the specified
(IgrooveByteInputStream *i_pStream, name in the database.
BSTR I_DocumentName,
IgrooveBinaryDocument **
o_ppDocument);
CreateOrOpenXMLDocument (BSTR    Opens the specified XML document; creates
i_DocumentName, BSTR             an empty document with the specified name
i_RootElementName, BSTR          and schema it if it doesn't already exist.
i_SchemaURI, IUnknown *
i_pAdditionalSchemaURIs,
VARIANT_BOOL * o_pCreated,
IGrooveXMLDocument **
o_ppDocument);
CreateXMLDocument (BSTR          Creates an empty XML document with the
i_DocumentName, BSTR             specified name and schema in the database.
i_RootElementName, BSTR
i_SchemaURI, IUnknown *
i_pAdditionalSchemaURIs,
IGrooveXMLDocument **
o_ppDocument);
CreateXMLDocumentFromStream      Given a stream of bytes, representing one of
(IGrooveByteInputStream * i_pStream, the supported character set encodings of a
GrooveParseOptions i_ParseOptions, XML document, creates an XML document in
BSTR i_DocumentName, BSTR        the database.
i_SchemaURI, IUnknown *
i_pAdditionalSchemaURIs, IUnknown *
i_pLinkCallback,
IGrooveXMLDocument **
o_ppDocument);
DeleteDocument (BSTR             Deletes the named document.
i_DocumentName);
DocumentExists (BSTR             Given the specified document name, checks
i_DocumentName, VARIANT_BOOL *   for the existence of the document in the
o_pDocumentExists);              database.
IsTransactionInProgress          Returns TRUE if a transaction is in progress.
(VARIANT_BOOL *
o_pTransactionInProgress);
OpenBinaryDocument (BSTR         Opens the specified binary document.
i_DocumentName,
IGrooveBinaryDocument **
o_ppDocument);
OpenCrossProcessSemaphore (BSTR  Creates a new cross process synchronization
i_Name, VARIANT_BOOL             object. If Name is not specified, the default
i_Reentrant,                     name for the database is used. If the
IGrooveCrossProcessSemaphore **  semaphore is not Reentrant, repeated
o_ppSemaphore);                  attempts to lock the semaphore within the
                                 same thread and process will block.
OpenDocumentNameEnum             Returns an enumeration of the documents
(VARIANT_BOOL i_OpenOnly,        currently in a database.
IGrooveBSTREnum **
o_ppDocumentNames);
OpenTransaction (VARIANT_BOOL    Creates a new transaction on the database.
i_BeginLock, VARIANT_BOOL        BeginLock specifies whether the database
i_ReadOnly, VARIANT_BOOL         cross process semaphore should be locked.
i_BeginTransaction, VARIANT_BOOL BeginTransaction specifies whether the
i_Reentrant, BSTR i_LockName,    transaction should start now. If LockName is
IGrooveTransaction **            not specified, the default name for the
o_ppTransaction);                database is used. If the semaphore is not
                                 Reentrant, repeated attempts to lock the
                                 semaphore within the same thread and
                                 process will block.
OpenURI (BSTR * o_pDatabaseURI); Returns the URI for this database.
OpenXMLDocument (BSTR            Opens the specified XML document.
i_DocumentName,
IGrooveXMLDocument **
o_ppDocument);
WasDataLost (VARIANT_BOOL *      Returns the value of a flag indicating whether
o_pDataLost);                    data may have been lost since the database
                                 was opened or the last transaction was
                                 committed.

Table 8 illustrates the methods for an interface 1102 (IGrooveCrossProcessSemaphore) for a client of a storage manager that needs to synchronize access among processes.

TABLE 8
Interface IGrooveCrossProcessSemaphore : IDispatch
DoLock (VARIANT_BOOL Locks the semaphore. If ReadOnly
i_ReadOnly);         is TRUE, only retrieval operations
                     may be performed on
                     the database, otherwise,
                     any operation may
                     be performed.
DoUnlock ( );        Unlocks the semaphore.

Table 9 illustrates an interface 1104 (IGrooveTransaction) for a client of a storage manager that needs to group operations within a database. Transactions are a sub-class of cross-process semaphores, that is, all of the methods for IGrooveCrossProcessSemaphore also apply to IGrooveTransaction. The storage manager transaction interface includes the following methods:

TABLE 9
Interface IGrooveTransaction : IGrooveCrossProcessSemaphore
Abort ( );                     Ends the transaction.
                               All work done to the
                               database since the start
                               of the transaction is
                               discarded.
Begin (VARIANT_BOOL            Starts a transaction.
i_ReadOnly);                   If ReadOnly is false, the
                               database may be updated.
BeginIndependent (VARIANT_BOOL Starts another transaction
i_ReadOnly);                   for this thread. Only one
                               independent transaction
                               is allowed per thread.
Commit ( );                    Ends the transaction. All work
                               done to the database since the
                               start of the transaction is
                               reliably stored in the database.

FIG. 12 shows interfaces which allows clients of the storage manager to manipulate documents and elements within those documents. Table 10 illustrates an interface 1200 (IGrooveDocument) for a client of a storage manager that needs to manage documents within a database. The storage manager document interface includes the following methods:

TABLE 10
Interface IGrooveDocument : IDispatch
OpenCrossProcessSemaphore (BSTR  Creates a new cross
i_Name, VARIANT_BOOL             process synchronization object.
i_Reentrant,                     If Name is not specified, the
IgrooveCrossProcessSemaphore **  URI for the document is used.
o_ppSemaphore);                  If the semaphore is not Reentrant,
                                 repeated attempts to lock the
                                 semaphore within the same
                                 thread and process will block.
OpenDatabase (IGrooveDatabase ** Returns an interface to
o_ppDatabase);                   the database object that
                                 contains this document.
OpenName (BSTR *                 Returns the document name.
o_pDocumentName);
OpenURI (BSTR * o_pURI);         Returns the URI that
                                 identifies this document.

Table 11 illustrates an interface 1202 (IGrooveXMLDocument) for a client of a storage manager that needs to manage XML documents within a database. XML documents are a sub-class of documents, that is, all of the methods for IGrooveDocument also apply to IGrooveXMLDocument. The storage manager XML document interface includes the following methods:

TABLE 11
interface IGrooveXMLDocument : IGrooveDocument
GenerateGrooveID (BSTR           Generates an 8 byte identifier from the string
i_GrooveIDBase, double *         identifier I_GrooveIDBase.
o_pGrooveID);
ConvertGrooveIDToSerializedGrooveID Converts an 8 byte identifier to the string
(double i_GrooveID, BSTR *       i_GrooveID.
o_pGrooveIDString);
ConvertSerializedGrooveIDToGrooveID Converts a string version of a Groove
(BSTR i_GrooveIDString, double * identifier to an 8 byte version.
o_pGrooveID);
CreateElement (BSTR i_Name,      Creates a new element with the supplied Tag;
IUnknown * i_pParent, IGrooveElement the tag cannot be altered once created. If a
** o_ppElement);                 Parent reference is supplied, the new element
                                 is created as a child of that parent.
CreateElementCopy (IGrooveElement * Does a deep/shallow copy of the specified
i_pSource, IGrooveElement *      element and all of its children (recursively for
i_pParent, VARIANT_BOOL          deep; just the one level for shallow), putting
i_ShallowCopy, IGrooveElement ** the new element(s) in under the Parent
o_ppElement);                    element.
CreateElementFromSchema (BSTR    Creates an element that conforms to the
i_Name, IGrooveElement * i_pParent, element's definition in the schema. Creates
IGrooveElement ** o_ppElement);  the element, its attributes, and any child
                                 elements.
CreateElementFromStream          Using a parser, creates an element, reads
(IGrooveByteInputStream * i_pStream, from a byte input stream and creates
GrooveParseOptions i_ParseOptions, elements and attributes from the text stream
IUnknown * i_pParent, IUnknown * as necessary, inserting them into the element,
i_pLinkCallback, IGrooveElement ** which is then returned to the caller. If a
o_ppElement);                    Parent reference is supplied, the new element
                                 is created as a child of that parent.
CreateLocator (IGrooveLocator ** Returns the interface to a new locator object.
o_ppLocator);
FindElementByID (BSTR i_ID,      Looks for an element of the specified ID and
IGrooveElement ** o_ppElement,   returns a boolean value if found.
VARIANT_BOOL * o_pFound);
OpenElementByID (BSTR i_ID,      Looks for an element of the specified ID.
IGrooveElement **o_ppElement);
OpenElementEnumByAttributeValue  Returns an enumeration of all of the elements
(BSTR i_ElementName, BSTR        within the document that have the named
i_AttributeName, BSTR            attribute with the specified value.
i_AttributeValue, IGrooveElementEnum
**o_ppElementEnum);
OpenElementEnumByAttributeValueAs Returns an enumeration of all of the elements
Bool (BSTR i_ElementName, BSTR   within the document that have the named
i_AttributeName, VARIANT_BOOL    attribute with the specified boolean type
i_AttributeValue, IGrooveElementEnum value.
**o_ppElementEnum);
OpenElementEnumByAttributeValueAs Returns an enumeration of all of the elements
Double (BSTR i_ElementName, BSTR within the document that have the named
i_AttributeName, double          attribute with the specified double floating
i_AttributeValue, IGrooveElementEnum type value.
**o_ppElementEnum);
OpenElementEnumByAttributeValueAs Returns an enumeration of all of the elements
Long (BSTR i_AttributeName, long within the document that have the named
i_AttributeValue, IGrooveElementEnum attribute with the specified long integer type
**o_ppElementEnum);              value.
OpenElementEnumByLocator (BSTR   Returns an element enumerator with
i_LocatorText, IGrooveElementEnum ** references to all elements satisfying the
o_ppElementEnum);                specified element locator expression. If there
                                 are no matching elements, the element
                                 enumerator will be created with no contents.
OpenElementEnumByName (BSTR      Returns an enumeration of all of the elements
i_Name, IGrooveElementEnum **    within the document that have the specified
o_ppElementEnum);                tag name.
OpenMetaElement (IGrooveElement ** Returns the interface to the meta element that
o_ppElement);                    defines this XML document.
OpenRootElement (IGrooveElement ** Opens the root element for the XML
o_ppRootElement);                document.

Table 12 illustrates the methods for an interface 1204 (IGrooveBinaryDocument) for a client of a storage manager that needs to manage binary documents within a database. Binary documents are a sub-class of documents, that is, all of the methods for IGrooveDocument also apply to IGrooveBinaryDocument.

TABLE 12
interface IGrooveBinaryDocument : IGrooveDocument
OpenByteInputStream        Returns the interface to a byte stream
(IGrooveByteInputStream ** object that can be used to read bytes
o_ppByteInputStream);      within the binary document.

Table 13 illustrates an interface 1206 (IGrooveLocator) for a client of a storage manager that needs to search for elements using locator queries as defined in a specification called XSLT. Details of the XSLT specification can be found at web address www.w3.org/TR/xslt. The storage manager locator interface includes the following methods:

TABLE 13
interface IGrooveLocator : IDispatch
FindElement (BSTR i_LocatorStr,  Returns an interface to the element object
IGrooveElement * i_pContextElement, that satisfies the search specified by the
IGrooveElement ** o_ppElement,   Locator string within the scope of the context
VARIANT_BOOL * o_pFound);        element.
Invalidate (VARIANT_BOOL         Clears the state information in the interface
i_AssignNewIDs);                 instance.
OpenElementEnum (BSTR            Returns an enumerator of all elements that
i_LocatorStr, IGrooveElement *   match the Locator string, collated according to
i_pContextElement, VARIANT_BOOL  the specified sorting criteria.
i_Sort, BSTR i_SortConstraint, BSTR
i_SortKey, GrooveSortOrder
i_SortOrder, IGrooveElementEnum **
o_ppElements);
OpenElementEnumWithTumblers      Perform the search specified by the Locator
(BSTR i_LocatorStr, IGrooveElement string on the elements pointed to by the
*i_pContextElement, VARIANT_BOOL context element, returning the tumbler values
i_RelativeTumblers,              for each match as well as the matching
IGrooveBSTREnum ** o_ppTumblers, elements, collated according to the specified
VARIANT_BOOL i_Sort, BSTR        sorting criteria.
i_SortConstraint, BSTR i_SortKey,
GrooveSortOrder i_SortOrder,
IGrooveElementEnum **
o_ppElements);
OpenText (BSTR i_LocatorStr,     Returns the text from element or attribute that
IGrooveElement * i_pContextElement, satisfies the search specified by the Locator
BSTR * o_pValue);                string within the scope of the context element.

Table 14 illustrates an interface 1208 (IGrooveTransform) for a client of a storage manager that needs to perform XML document transformations as defined in XSLT. The storage manager transform interface includes the following methods:

TABLE 14
Interface IGrooveTransform : IDispatch
TransformXMLDocument             Transforms the input XML document,
(IGrooveXMLDocument *            returning the result of the transformation in
i_pXMLDocument, IGrooveElement * ResultDocument.
i_pStartElement, BSTR i_SortRule,
long i_StartElementNum, long
i_NumElements,
IGrooveXMLDocument *
io_pResultDocument, VARIANT_BOOL
i_AlwaysOutputHeader, long *
o_pElementsProcessed);
TransformElement (IGrooveElement * Transforms the input ContextElement,
i_pContextElement, BSTR          returning the result of the transformation in
i_TansformationTemplate,         ResultDocument.
IGrooveXMLDocument **
o_ppResultDocument);

Table 15 illustrates an interface 1210 (IGrooveElement) which allows a client of a storage manager to manipulate elements within XML documents. The storage manager element interface includes the following methods:

TABLE 15
Interface IGrooveElement : IDispatch
AppendContent (BSTR i_Text,      Inserts the kind of content as the last of its
GrooveContentType i_Type);       type within this element.
AppendContentElement             Inserts the element as the last content
(IGrooveElement * i_pElement);   element.
AppendContentProcessingInstruction Inserts a processing instruction, with target
(BSTR i_Target, BSTR i_Text);    Target, as the last processing instruction.
CreateElement (BSTR i_Name,      Create a new element in the same
IGrooveElement * i_pParent,      document.
IGrooveElement ** o_ppElement);
CreateElementCopy (IGrooveElement * Does a deep/shallow copy of the specified
i_pSource, IGrooveElement * i_pParent, element and all of its children (recursively for
VARIANT_BOOL i_ShallowCopy,      deep; just the one level for shallow), putting
IGrooveElement ** o_ppElement);  the new element(s) in the destination
                                 document. The returned element must be
                                 attached into the document's element tree.
CreateElementFromSchema (BSTR    Creates an element that conforms to the
i_Name, IGrooveElement * i_pParent, element's definition in the schema. Creates
IGrooveElement ** o_ppElement);  the element, its attributes, and any child
                                 elements.
CreateElementRPCClient           Creates and returns the interface to the
(IGrooveElementRPCClient         element RPC client.
**o_ppRPCClient);
CreateElementRPCServer           Creates and returns the interface to the
(IGrooveElementRPCServer **      element RPC server.
o_ppRPCServer);
CreateElementRPCServerThread     Creates and returns the interface to the
(IGrooveElementRPCServerCallback * element RPC server thread.
i_pCallback,
IGrooveElementRPCServerThread **
o_ppRPCServerThread);
CreateLink (IGrooveDocument *    Creates a link to another document, using
i_pDocument, BSTR i_Title, BSTR  the specified XLink parameters.
i_Role, GrooveXLinkShow i_Show,
GrooveXLinkActuate i_Actuate,
GrooveXLinkSerialize i_Serialize);
DecrementAttributeAsLong (BSTR   Subtracts 1 from the value of a long integer
i_Name, long * o_pOldValue);     type attribute.
Delete ( );                      Permanently removes the element from the
                                 document. No further operations may be
                                 performed on a deleted element
DeleteAllAttributes ( );         Removes all attributes from the element.
DeleteAllContent ( );            Removes all child content elements and text
                                 from the element and deletes them from the
                                 document.
DeleteAttribute (BSTR i_Name);   Removes the named attribute from the
                                 element.
DeleteContent (long i_Ordinal);  Removes the content at the specified
                                 position from the element.
DeleteLinkAttributes ( );        Removes all attributes that are links from the
                                 element.
DetachFromParent ( );            Removes this element from the content of its
                                 parent. The element is still part of the
                                 document and must be reattached or
                                 destroyed before it is released.
DoesAttributeExist (BSTR i_Name, Returns whether the attribute is set on the
VARIANT_BOOL * o_pFound);        element.
Duplicate (IGrooveElement *      Make the specified target element a
i_pTargetElement, VARIANT_BOOL   duplicate of this element, overriding
i_ShallowDuplicate);             attributes and, if ShallowDuplicate is FALSE,
                                 all descendent elements.
FindAttribute (BSTR i_Name, BSTR * Gets any arbitrary attribute as text. If the
o_pValue, VARIANT_BOOL *         attribute is not in the element, Found is
o_pFound);                       FALSE and no value is returned.
FindAttributeAsBinary (BSTR i_Name, Gets any arbitrary attribute as Binary. The
IGrooveByteInputStream ** o_ppValue, attribute must have been set as the given
VARIANT_BOOL *o_pFound);         type or be specified as that type in the
                                 document schema. If the attribute is not in
                                 the element, Found is FALSE and no value
                                 is returned.
FindAttributeAsBinaryArray (BSTR Gets any arbitrary attribute as Binary and
i_Name, SAFEARRAY(BYTE) *        return the value in an array. The attribute
o_ppValue, VARIANT_BOOL *        must have been set as the given type or be
o_pFound);                       specified as that type in the document
                                 schema. If the attribute is not in the
                                 element, Found is FALSE and no value is
                                 returned.
FindAttributeAsBinaryToStream (BSTR Gets any arbitrary attribute as Binary and
i_Name, IGrooveByteOutputStream * returns the value in a stream. The attribute
i_pStream, VARIANT_BOOL          must have been set as the given type or be
*o_pFound);                      specified as that type in the document
                                 schema. If the attribute is not in the
                                 element, Found is FALSE and no value is
                                 returned.
FindAttributeAsBool (BSTR i_Name, Gets any arbitrary attribute as Boolean. The
VARIANT_BOOL * o_pValue,         attribute must have been set as the given
VARIANT_BOOL * o_pFound);        type or be specified as that type in the
                                 document schema. If the attribute is not in
                                 the element, Found is FALSE and no value
                                 is returned.
FindAttributeAsDouble (BSTR i_Name, Gets any arbitrary attribute as Double. The
double * o_pValue, VARIANT_BOOL * attribute must have been set as the given
o_pFound);                       type or be specified as that type in the
                                 document schema. If the attribute is not in
                                 the element, Found is FALSE and no value
                                 is returned.
FindAttributeAsGrooveID (BSTR    Gets any arbitrary attribute as a Groove
i_Name, double * o_pValue,       identifier. The attribute must have been set
VARIANT_BOOL * o_pFound);        as the given type or be specified as that type
                                 in the document schema. If the attribute is
                                 not in the element, Found is FALSE and no
                                 value is returned.
FindAttributeAsLong (BSTR i_Name, Gets any arbitrary attribute as Long. The
long * o_pValue, VARIANT_BOOL *  attribute must have been set as the given
o_pFound);                       type or be specified as that type in the
                                 document schema. If the attribute is not in
                                 the element, Found is FALSE and no value
                                 is returned.
FindAttributeAsVARIANT (BSTR     Gets any arbitrary attribute as a variant
i_Name, VARIANT * o_pValue,      value. If the attribute is not in the element,
VARIANT_BOOL * o_pFound);        Found is FALSE and no value is returned.
FindContentElementByName (BSTR   Within the context of this element, find an
i_Name, IGrooveElement **        element with the specified tag name. If the
o_ppElement, VARIANT_BOOL *      element is not found, Found is FALSE and
o_pFound);                       no element reference is returned.
FindContentElementByNameAndAttribute Within the context of this element, find an
(BSTR i_Name, BSTR               element with the specified tag name and
i_AttributeName, BSTR i_AttributeValue, attribute name with the specified attribute
IGrooveElement ** o_ppElement,   value. If the element is not found, Found is
VARIANT_BOOL * o_pFound);        FALSE and no element reference is
                                 returned
FindParent (IGrooveElement **    Gets an object's parent element. An
o_ppParent, VARIANT_BOOL *       element can have only a single parent and
o_pFound);                       may only be referenced from a single
                                 content entry of a single element. If the
                                 element does not have a parent, Found is
                                 FALSE and no value is returned.
GetActuate (GrooveXLinkActuate * Returns the value of the Actuate parameter
o_pActuate);                     in this element's link attribute.
GetAttributeCount (long * o_pCount); Returns the number of attributes an element
                                 has.
GetContentCount (long * o_pCount); Returns the number of content and text
                                 entries in this element.
GetContentType (long i_Ordinal   Returns the type of content at the specified
GrooveContentType * o_pType);    ordinal position.
GetOrdinal (long * o_pOrdinal);  Gets the ordinal position within the parent's
                                 content of this element.
GetSerialize (GrooveXLinkSerialize * Returns the value of the Serialize parameter
o_pSerialize);                   in this element's link attribute.
GetShow (GrooveXLinkShow *       Returns the value of the Show parameter in
o_pShow);                        this element's link attribute.
IncrementAttributeAsLong (BSTR   Adds 1 to the value of a long integer type
i_Name, long * o_pOldValue);     attribute.
InsertContent (long i_Ordinal, BSTR Inserts the text entry at the specified ordinal
i_Text, GrooveContentType i_Type); location
InsertContentElement (long i_Ordinal, Inserts the element at the specified ordinal
IGrooveElement * i_pElement);    location
InsertContentProcessingInstruction (long Inserts a Text processing instruction, with
i_Ordinal, BSTR i_Target, BSTR i_Text); target Target, at the specified ordinal
                                 position.
IsLinkElement (VARIANT_BOOL *    Determines whether or not the element
o_plsLink);                      contains XLink markup.
IsReferenced (VARIANT_BOOL *     Returns TRUE if this element is referenced.
o_plsReferenced);
IsSame (IGrooveElement * i_pElement, Returns TRUE if the specified element
VARIANT_BOOL * o_plsSame);       object is this element or equal to this
                                 element.
OpenAttribute (BSTR i_Name, BSTR Gets any arbitrary attribute as text.
*o_pValue);
OpenAttributeAsBinary (BSTR i_Name, Gets any arbitrary attribute as Binary. The
IGrooveByteInputStream ** o_ppValue); attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
OpenAttributeAsBinaryArray (BSTR Gets any arbitrary attribute as Binary and
i_Name, SAFEARRAY(BYTE) *        return the value in an array. The attribute
o_ppValue);                      must have been set as the given type or be
                                 specified as that type in the document
                                 schema.
OpenAttributeAsBinaryToStream (BSTR Gets any arbitrary attribute as Binary and
i_Name, IGrooveByteOutputStream * returns the value in a stream. The attribute
i_pStream);                      must have been set as the given type or be
                                 specified as that type in the document
                                 schema.
OpenAttributeAsBool (BSTR i_Name, Gets any arbitrary attribute as Boolean. The
VARIANT_BOOL * o_pValue);        attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
OpenAttributeAsDouble (BSTR i_Name, Gets any arbitrary attribute as Double. The
double * o_pValue);              attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
OpenAttributeAsGrooveID (BSTR    Gets any arbitrary attribute as a Groove
i_Name, double * o_pValue);      identifier. The attribute must have been set
                                 as the given type or be specified as that type
                                 in the document schema.
OpenAttributeAsLong (BSTR i_Name, Gets any arbitrary attribute as Long. The
long * o_pValue);                attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
OpenAttributeAsVARlANT (BSTR     Gets any arbitrary attribute as a variant
i_Name, VARIANT * o_pValue);     value.
OpenAttributeEnum                Enumerates all of the element's attributes as
(IGrooveStringStringEnum **      text.
o_ppAttributes);
OpenAttributeVariantEnum         Enumerates all of the element's attributes as
(IGrooveNameValueEnum **         variant data types.
o_ppEnum);
OpenBoundCode (IGrooveBoundCode  Returns an instance of the object bound to
** o_ppBoundCode);               the element.
OpenContentComment (long i_Ordinal, Returns the text of the comment that is a
BSTR * o_pComment);              contained in this element at the specified
                                 Ordinal position.
OpenContentElement (long i_Ordinal, Returns the child element interface that is a
IGrooveElement ** o_ppElement);  contained in this element at the specified
                                 Ordinal position.
OpenContentElementByName (BSTR   Within the context of this element, find an
i_Name, IGrooveElement **        element with the specified tag name and
o_ppElement);                    return its interface.
OpenContentElementByNameAndAttribute Within the context of this element, find an
(BSTR i_Name, BSTR               element with the specified tag name and
i_AttributeName, BSTR i_AttributeValue, attribute name with the specified attribute
IGrooveElement ** o_ppElement);  value.
OpenContentElementEnum           Returns an enumeration of all child content
(IGrooveElementEnum **           elements (non-recursively).
o_ppElements);
OpenContentElementEnumByName     Returns an enumeration of all child content
(BSTR i_Name, IGrooveElementEnum ** elements (non-recursively). Only elements
o_ppElements);                   with the given name will be returned.
OpenContentElementEnumByNameAnd  Returns an enumeration of all content
Attribute (BSTR i_Name, BSTR     elements within the scope of this element
i_AttributeName, BSTR i_AttributeValue, that have the specified tag name and
IGrooveElementEnum ** o_ppElements); attribute name with the specified attribute
                                 value.
OpenContentProcessingInstruction (long Returns the XML processing instruction at
i_Ordinal, BSTR * o_pTarget, BSTR * the specified ordinal position.
o_pText);
OpenContentProcessingInstructionTarget Returns the target of the XML processing
(long i_Ordinal, BSTR * o_pTarget); instruction at the specified ordinal position.
OpenContentProcessingInstructionText Returns the PI text of the XML processing
(long i_Ordinal, BSTR * o_pText); instruction at the specified ordinal position.
OpenContentText (long i_Ordinal, BSTR Returns the context text at the specified
* o_pText);                      ordinal position.
OpenContentTextEnum              Enumerates the text entries
(IGrooveBSTREnum ** o_ppText);   (non-recursively).
OpenElementQueue                 Create an element queue on the element.
(IGrooveElementQueue ** o_ppQueue); The element queue does not affect the
                                 element's structure.
OpenElementReferenceQueue        Returns the interface to reference queue
(IGrooveElementReferenceQueue ** object.
o_ppQueue);
OpenHRef (BSTR * o_pHref);       Returns the value of the HREF parameter in
                                 this element's link attribute.
OpenLinkAttributes (BSTR * o_pHref, Retrieves all the standard link elements.
BSTR * o_pTitle, BSTR * o_pRole, Note : not all the attributes are mandatory
GrooveXLinkShow * o_pShow,
GrooveXLinkActuate * o_pActuate,
GrooveXLinkSerialize * o_pSerialize);
OpenLinkedBinaryDocument         Returns the interface to the binary document
(VARIANT_BOOL i_SingleProcess,   that is referenced in the HREF parameter in
IUnknown * i_pSecurityContext,   this element's link attribute.
IGrooveBinaryDocument **
o_ppDocument);
OpenLinkedXMLDocument            Returns the interface to the XML document
(VARIANT_BOOL i_SingleProcess,   that is referenced in the HREF parameter in
IUnknown * i_pSecurityContext,   this element's link attribute.
IGrooveXMLDocument **
o_ppDocument);
OpenMultiReaderElementQueueReader Create an element multi-reader queue on
(IGrooveMultiReaderElementQueueReader the element and add a reader. This could
** o_ppQueue);                   change the structure of the element.
OpenMultiReaderElementQueueWriter Create an element multi-writer queue on the
(GrooveMultiReaderQueueOptions   element and add a writer. This could
i_Options,                       change the structure of the element.
IGrooveMultiReaderElementQueueWriter
** o_ppQueue);
OpenMultiReaderElementReferenceQueue Returns the interface to the multi-reader
Reader                           element reference queue reader object.
(IGrooveMultiReaderElementQueueReader
** o_ppQueue);
OpenMultiReaderElementReferenceQueue Returns the interface to the multi-reader
Writer                           element reference queue writer object.
(GrooveMultiReaderQueueOptions
i_Options,
IGrooveMultiReaderElementQueueWriter
** o_ppQueue);
OpenName (BSTR * o_pName);       Returns the element's tag name.
OpenParent (IGrooveElement **    Gets an object's parent element. An
o_ppParent);                     element can have only a single parent and
                                 may only be referenced from a single
                                 content entry of a single element.
OpenReadOnlyElement              Return the read-only element interface to
(VARIANT_BOOL i_AllowOpenParent, this element.
IGrooveReadOnlyElement **
o_ppReadOnlyElement);
OpenReference                    Returns the element reference interface to
(IGrooveElementReference **      this element.
o_ppElementReference);
OpenRole (BSTR * o_pRole);       Returns the value of the Role parameter in
                                 this element's link attribute.
OpenTitle (BSTR * o_pTitle);     Returns the value of the Title parameter in
                                 this element's link attribute.
OpenURI (BSTR * o_pName);        Returns the URI to this element.
OpenXMLDocument                  Returns the interface pointer to the XML
(IGrooveXMLDocument **           document containing this element.
o_ppDocument);
Serialize (GrooveSerializeType i_Type, Serializes the element to a stream with the
enum GrooveCharEncoding i_Encoding, specified encoding and options.
GrooveSerializeOptions i_Options,
IGrooveByteInputStream **
o_ppStream);
SerializeReturnAdditionalLinkedDocuments Serializes the element to a stream with the
(GrooveSerializeType i_Type, enum specified encoding and options. Returns an
GrooveCharEncoding i_Encoding,   enumeration of interfaces to documents
GrooveSerializeOptions i_Options, referenced by links in this element and all
IGrooveDocumentEnum **           descendents.
o_ppAdditionalLinkedDocuments,
IGrooveByteInputStream **
o_ppStream);
SerializeToStream                Serializes the element to a stream with the
(IGrooveByteOutputStream * i_pStream, specified encoding and options.
GrooveSerializeType i_Type, enum
GrooveCharEncoding i_Encoding,
GrooveSerializeOptions i_Options);
SerializeToStreamReturnAdditionalLinked Serializes the element to a stream with the
Documents (IGrooveByteOutputStream specified encoding and options. Returns an
* i_pStream, GrooveSerializeType enumeration of interfaces to documents
i_Type, enum GrooveCharEncoding  referenced by links in this element and all
i_Encoding, GrooveSerializeOptions descendents.
i_Options, IGrooveDocumentEnum **
o_ppAdditionalLinkedDocuments);
SetAttribute (BSTR i_Name, BSTR  Sets any arbitrary attribute as text.
i_Value);
SetAttributeAsBinary (BSTR i_Name, Sets any arbitrary attribute as Binary. The
IGrooveByteInputStream * i_pValue); attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
SetAttributeAsBinaryArray (BSTR  Sets any arbitrary attribute as Binary and
i_Name, SAFEARRAY(BYTE) *        returns the value in an array. The attribute
i_pValue);                       must have been set as the given type or be
                                 specified as that type in the document
                                 schema.
SetAttributeAsBool (BSTR i_Name, Sets any arbitrary attribute as Boolean. The
VARIANT_BOOL i_Value);           attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
SetAttributeAsDouble (BSTR i_Name, Sets any arbitrary attribute as Double. The
double i_Value);                 attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
SetAttributeAsGrooveID (BSTR i_Name, Sets any arbitrary attribute as a Groove
double i_pValue);                identifier. The attribute must have been set
                                 as the given type or be specified as that type
                                 in the document schema.
SetAttributeAsLong (BSTR i_Name, long Sets any arbitrary attribute as Long. The
i_Value);                        attribute must have been set as the given
                                 type or be specified as that type in the
                                 document schema.
SetAttributeAsVARIANT (BSTR i_Name, Sets any arbitrary attribute using a Variant,
VARIANT * i_pValue);             which may be any variant type.
SetContent (long i_Ordinal, BSTR Sets the content as the type's ordinal
i_Text, GrooveContentType i_Type); position to the specified text. Note that
                                 content of different types have independent
                                 ordinal positions.
SetContentElement (long i_Ordinal, Set the content element at the specified
IGrooveElement * i_pElement);    ordinal position.
SetContentProcessingInstruction (long Set the content processing instruction at the
i_Ordinal, BSTR i_Target, BSTR i_Text); specified ordinal position.
SetContentTextEnum               Creates text entries, separated by <BR>
(IGrooveBSTREnum * i_pEnum);     elements, for each text string in the
                                 enumerator.
SetLinkAttributes (BSTR i_Href, BSTR Sets the link attributes needed to make the
i_Title, BSTR i_Role, GrooveXLinkShow element a link element, including the
i_Show, GrooveXLinkActuate i_Actuate, ‘xml:link’ attribute, which is implicitly set to
GrooveXLinkSerialize i_Serialize); ‘simple’.
SetName (BSTR i_Name);           Sets the name of the element.
SetTempAttribute (BSTR i_Name, BSTR Sets an attribute with a temporary value,
i_Value);                        which will not be committed in a transaction.

Table 16 illustrates the methods for an interface 1212 (IGrooveReadOnlyElement) for a client of a storage manager that needs to manipulate read-only elements within XML documents. Read-only elements are a sub-class of elements, that is, all of the methods for IGrooveElement also apply to IGrooveReadOnlyElement.

TABLE 16
interface IGrooveReadOnlyElement : IGrooveElement
OpenReadOnlyParent               Returns a read-only element interface to the
(IGrooveReadOnlyElement **       parent of this element.
o_ppParent);
OpenContentReadOnlyElement (long Returns a read-only element interface to the
i_Ordinal, IGrooveReadOnlyElement ** content element at the specified Ordinal
o_ppElement);                    position.
OpenContentReadOnlyElementByName Within the context of this element, find an
(BSTR i_Name,                    element with the specified tag name and
IGrooveReadOnlyElement **        return its read-only interface.
o_ppElement);
FindContentReadOnlyElementByName Within the context of this element, find an
(BSTR i_Name,                    element with the specified tag name and
IGrooveReadOnlyElement **        return its read-only interface. If the element is
o_ppElement, VARIANT_BOOL *      not found, Found is FALSE and no element
o_pFound);                       reference is returned.
OpenContentReadOnlyElementEnum   Returns an enumeration of all child content
(IGrooveReadOnlyElementEnum **   elements read-only interfaces
o_ppElements);                   (non-recursively).
OpenContentReadOnlyElementEnumByName Returns an enumeration of all child content
(BSTR i_Name,                    elements read-only interfaces
IGrooveReadOnlyElementEnum **    (non-recursively). Only elements with the
o_ppElements);                   given name will be returned.

Table 17 illustrates an interface 1214 (IGrooveElementReference) for a client of a storage manager that needs to manipulate element references within XML documents. The storage manager element reference interface includes the following methods:

TABLE 17
Interface IGrooveElementReference : IDispatch
OpenElement                Returns a read-only element interface to
(IgrooveReadOnlyElement ** the referenced element.
o_ppElement);

An interface 1216 (IGrooveElementUtilBase) for use within the storage manager's other interfaces is shown in Table 18. The IGrooveElementUtilBase is not an interface for commonly-used objects, but is intended to serve as the base class for other sub-classes (shown in FIG. 13) that do have commonly-used objects. All of the “util” interfaces are associated with an element. The storage manager element util base interface includes the following methods:

TABLE 18
Interface IGrooveElementUtilBase : IDispatch
OpenDocument                   Returns the interface of the
(IgrooveXMLDocument **         containing XML document.
o_ppDocument);
OpenElement (IGrooveElement ** Returns the element's interface.
o_ppElement);

Table 19 illustrates an interface 1218 (IGrooveBoundCode) for a client of a storage manager that needs to handle executable code associated with elements within XML documents. The storage manager bound code interface includes the following methods:

TABLE 19
interface IGrooveBoundCode : IDispatch
SetElement (IGrooveElement *   Sets the element interface pointer
i_pElement);                   associated with this element tag.
OpenElement (IGrooveElement ** Retrieves the element interface
o_ppElement);                  pointer associated with this element
                               tag.

FIG. 13 illustrates interfaces which are sub-classes of the IGrooveElementUtilBase base class 1300, discussed above. Table 20 illustrates an interface 1302 (IGrooveElementQueue) for a client of a storage manager that needs to manipulate queues on elements within XML documents. Element queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveElementQueue. The storage manager element queue interface includes the following methods:

TABLE 20
interface IGrooveElementQueue : IGrooveElementUtilBase
Enqueue (IGrooveElement *        Enqueues the element. Note that the element
i_pElement);                     must already be contained in the queue's
                                 document.
Dequeue (long i_TimeoutMilliseconds, Dequeues the next available element in the
IGrooveElement ** o_ppElement);  queue. Returns only when an element is
                                 available or after the timeout period. The
                                 returned IGrooveElement pointer will be NULL
                                 if the timeout period expires.
DequeueEnum (long                Dequeues all available elements in the queue.
i_TimeoutMilliseconds,           Returns only when an element is available or
IGrooveElementEnum **            after the timeout period. The returned
o_ppElements);                   IGrooveElement pointer will be NULL if the
                                 timeout period expires.
OpenEvent (IGrooveEvent **       Returns an event that can be used to ‘Wait’
o_ppEvent);                      for an element to be enqueued

Table 21 illustrates an interface 1306 (IGrooveElementReferenceQueue) for a client of a storage manager that needs to manipulate queues on element references within XML documents. Element reference queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveElementReferenceQueue. The storage manager element reference queue interface includes the following methods:

TABLE 21
interface IGrooveElementReferenceQueue : IGrooveElementUtilBase
Enqueue (IGrooveElement *        Enqueues the element. Note that the element
i_pElement);                     must already be contained in the queue's
                                 document.
EnqueueReference (IGrooveElement * Enqueues a reference to the element. Note
i_pElement);                     that the element must already be contained in
                                 the queue's document.
Dequeue (long i_TimeoutMilliseconds, Dequeues the next available element in the
IGrooveElementReference **       queue. Returns only when an element is
o_ppElementReference);           available or after the timeout period. The
                                 returned IGrooveElementReference pointer
                                 will be NULL if the timeout period expires.
DequeueEnum (long                Dequeues all available elements in the queue.
i_TimeoutMilliseconds,           Returns only when an element is available or
IGrooveElementReferenceEnum **   after the timeout period. The returned
o_ppElementReferences);          IGrooveElementReferenceEnum pointer will
                                 be NULL if the timeout period expires.
OpenEvent (IGrooveEvent **       Returns an event that can be used to ‘Wait’
o_ppEvent);                      for an element to be enqueued

Table 22 illustrates an interface 1310 (IGrooveMultiReaderElementQueueReader) for a client of a storage manager that needs to remove elements from multi-reader queues on elements within XML documents. Multi-reader element queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveMultiReaderElementQueueReader. The storage manager multi-reader element queue reader interface includes the following methods:

TABLE 22
interface IGrooveMultiReaderElementQueueReader : IGrooveElementUtilBase
Dequeue (long i_TimeoutMilliseconds, Dequeues the next available element in the
IGrooveElement ** o_ppElement);  queue. Returns only when an element is
                                 available or after the timeout period. The
                                 returned IGrooveElement pointer will be NULL
                                 if the timeout period expires.
DequeueEnum (long                Dequeues all available elements in the queue.
i_TimeoutMilliseconds,           Returns only when an element is available or
IGrooveElementEnum **            after the timeout period. The returned
o_ppElements);                   IGrooveElement pointer will be NULL if the
                                 timeout period expires.
OpenEvent (IGrooveEvent **       Returns an event that can be used to ‘Wait’
o_ppEvent);                      for an element to be enqueued

Table 23 illustrates an interface 1314 (IGrooveMultiReaderElementQueueWriter) for a client of a storage manager that needs to add elements to multi-reader queues on elements within XML documents. Multi-reader element queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveMultiReaderElementQueueWriter. The storage manager multi-reader element queue writer interface includes the following methods:

TABLE 23
interface IGrooveMultiReaderElementQueueWriter :
IGrooveElementUtilBase
Enqueue (IGrooveElement Enqueues the element and returns the
*i_pElement, long *     number already enqueued. Note that the
o_pNumEnqueued);        element must already be contained in the
                        queue's document.
GetNumReaders (long *   Get the number of readers on the queue.
o_pNumReaders);

Table 24 illustrates an interface 1318 (IGrooveMultiReaderElementReferenceQueueWriter) for a client of a storage manager that needs to add element references to multi-reader queues on elements within XML documents. Multi-reader element reference queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveMultiReaderElementReferenceQueueWriter. The storage manager multi-reader element reference queue writer interface includes the following methods:

TABLE 24
interface IGrooveMultiReaderElementReferenceQueueWriter :
IGrooveElementUtilBase
Enqueue (IGrooveElement * i_pElement, Enqueues the element and returns the
long * o_pNumEnqueued);          number already enqueued. Note that the
                                 element must already be contained in the
                                 queue's document.
EnqueueReference (IGrooveElement * Enqueues the element reference and
i_pElement, long * o_pNumEnqueued); returns the number already enqueued.
                                 Note that the element must already be
                                 contained in the queue's document.
GetNumReaders (long * o_pNumReaders); Get the number of readers on the queue.

Table 25 illustrates an interface 1316 (IGrooveMultiReaderElementReferenceQueueReader) for a client of a storage manager that needs to remove element references from multi-reader queues on elements within XML documents. Multi-reader element reference queues are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveMultiReaderElementReferenceQueueReader. The storage manager multi-reader element reference queue reader interface includes the following methods:

TABLE 25
interface IGrooveMultiReaderElementReferenceQueueReader :
IGrooveElementUtilBase
Dequeue (long i_TimeoutMilliseconds, Dequeues the next available element
IGrooveElementReference **       reference in the queue. Returns only
o_ppElementReference);           when an element is available or after the
                                 timeout period. The returned
                                 IGrooveElementReference pointer will be
                                 NULL if the timeout period expires.
DequeueEnum (long                Dequeues all available element references
i_TimeoutMilliseconds,           in the queue. Returns only when an
IGrooveElementReferenceEnum **   element is available or after the timeout
o_ppElementReferences);          period. The returned
                                 IGrooveElementReference pointer will be
                                 NULL if the timeout period expires.
OpenEvent (IGrooveEvent ** o_ppEvent); Returns an event that can be used to
                                 ‘Wait’ for an element to be enqueued

Table 26 illustrates an interface 1304 (IGrooveRPCClient) for a client of a storage manager that needs to perform remote procedure calls (RPCS) on elements within XML documents. RPC clients are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveRPCClient. The storage manager RPC client interface includes the following methods:

TABLE 26
interface IGrooveElementRPCClient : IGrooveElementUtilBase
DoCall (IGrooveElement * Make a RPC, using the Input element as the
i_pInput,                input parameters and receiving output
IGrooveElement **        parameters in the Output element.
o_ppOutput);
SendCall                 Make an asynchronous RPC, using the Input
(IGrooveElement *        element as the input parameters.
i_pInput);
OpenResponseQueue        Returns the queue where responses are
(IGrooveElementQueue **  received.
o_ppQueue);

An interface 1308 (IGrooveRPCServerThread) for a client of a storage manager that needs to handle remote procedure calls (RPCs) on elements within XML documents is shown in Table 27. RPC server threads are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveRPCServerThread. The storage manager RPC server callback interface has no methods of its own, only those inherited from IGrooveElementUtilBase. It is provided as a distinct interface for type checking.

TABLE 27
interface IGrooveElementRPCServerThread : IGrooveElementUtilBase
(none)

Table 28 illustrates an interface 1312 (IGrooveRPCServer) for a client of a storage manager that needs to handle remote procedure calls (RPCs) on elements within XML documents. RPC servers are a sub-class of the “util” base class, that is, all of the methods for IGrooveElementUtilBase also apply to IGrooveRPCServer. The storage manager RPC server interface includes the following methods:

TABLE 28
interface IGrooveElementRPCServer : IGrooveElementUtilBase
OpenCallQueue                  Returns the queue where calls are
(IGrooveElementQueue **        received.
o_ppQueue);
SendResponse (IGrooveElement * Sends a response to the caller,
i_pInput, IGrooveElement *     returning output parameters in the
i_pOutput,                     Output element.
VARIANT_BOOL * o_bResult);

The following tables illustrate allowed values for the enumerated data types listed in the above interfaces. In particular, Table 29, illustrates allowed values for the GrooveSerializeType enumerated data type.

TABLE 29
GrooveSerializeType
GrooveSerializeAuto  On input, Groove will determine the correct
                     format by examining the first few bytes of the
                     input stream. On output, Groove will select a
                     format based on the kind of document or
                     element data.
GrooveSerializeMIME  Format is MHTML, as defined in RFC 2557.
GrooveSerializeXML   Format is XML. Note that binary documents
                     are not supported with this format, but it may
                     be a body type in MHTML.
GrooveSerializeWBXML Format is WBXML. Note that binary
                     documents are not supported with this format,
                     but it may be a body type in MHTML.

Table 30 illustrates the allowed values for the GrooveSerializeOptions enumerated data type.

TABLE 30
GrooveSerializeOptions
GrooveSerializeDefault           Use default serialization options.
GrooveSerializeWithFormatting    Indent, with blanks, each level of child content
                                 elements beneath the parent element.
GrooveSerializeSortedAttrs       Output the attributes for each element in order
                                 of ascending attribute name.
GrooveSerializeNoFragmentWrapper Output without the fragment wrapper for
                                 document fragments (elements).
GrooveSerializeNoNamespaceContraction Output with fully expanded element and
                                 attribute names.
GrooveSerializeNoProlog          Output without the XML document prolog.
GrooveSerializeNoLinks           Output without linked documents.
GrooveSerializeNotMinimum        Don't spend as much local processor time as
                                 needed to ensure the resulting output is the
                                 minimum size.

Table 31 illustrates the allowed values for the GrooveParseOptions enumerated data type.

TABLE 31
GrooveParseOptions
GrooveParseDefault               Use default parse options.
GrooveParseStripContentWhitespace Remove all extraneous
                                 whitespace from
                                 element content.
GrooveParseNoFragment            Parse a fragment that doesn't
                                 have a fragment wrapper.
GrooveParseNoNamespaceExpansion  Parse the document,
                                 but don't expand
                                 namespaces to their
                                 fully qualified form.
GrooveParseNoLinks               Parse a document
                                 and skip the links.

Table 32 illustrates the allowed values for the GrooveContentType enumerated data type.

TABLE 32
GrooveContentType
GrooveContentElement             Content is a child element.
GrooveContentText                Content is body text.
GrooveContentCDATASection        Content is a CDATA section.
GrooveContentProcessingInstruction Content is a processing instruction.
GrooveContentComment             Content is a comment.

Table 33 illustrates the allowed values for the GrooveXLinkShow enumerated data type.

TABLE 33
GrooveXLinkShow
GrooveXLinkShowNew     New.
GrooveXLinkShowParsed  Parsed.
GrooveXLinkShowReplace Replace

Table 34 illustrates the allowed values for the GrooveXLinkActuate enumerated data type:

TABLE 34
GrooveXLinkActuate
GrooveXLinkActuateUser User.
GrooveXLinkActuateAuto Auto.

Table 35 illustrates the allowed values for the GrooveXLinkSerialize enumerated data type.

TABLE 35
GrooveXLinkSerialize
GrooveXLinkSerializeByValue     By value.
GrooveXLinkSerializeByReference By reference.
GrooveXLinkSerializeIgnore      Ignore.

Table 36 illustrates the allowed values for the GrooveMultiReaderQueueOptions enumerated data type.

TABLE 36
GrooveMultiReaderQueueOptions
GrooveMRQDefault            Use default options.
GrooveMRQAllReceive         All readers receive each
                            event notification.
GrooveMRQEnqueueIfNoReaders Enqueue even if no
                            reader is currently
                            queued to receive the element.

The fundamental data model of the storage manager is XML. XML is a semi-structured, hierarchical, hyper-linked data model. Many real world problems are not well represented with such complex structures and are better represented in tabular form. For example, spreadsheets and relational databases provide simple, tabular interfaces. In accordance with one aspect of the invention, in order to simplify the representation, XML structures are mapped to a tabular display, generally called a “waffle”. The waffle represents a collection of data. This mapping is performed by the collection manager, a component of the storage manager.

Collections are defined by a collection descriptor, which is an XML document type description. Like a document schema, the collection descriptor is a special kind of document that is stored apart from the collection data itself. There are many sources of collection data, but the primary source of collection data is a software routine called a record set engine. Driven by user commands, the record set engine propagates a set of updates for a collection to the collection manager. Based on those updates, the collection manager updates index structures and may notify waffle users via the notification system. When a waffle user needs updated or new collection data, the waffle user will call the collection manager to return a new result array containing the updated data. The waffle user may also navigate within the collection using cursors.

The following list shows the XML DTD contents for a collection descriptor document:

<!ELEMENT Collection ANY>
<!ATTLIST Collection
Name	CDATA	#REQUIRED
Start	(record|index)	“record”   #REQUIRED
Version	CDATA	#REQUIRED
Location	CDATA	#IMPLIED
>
<!ELEMENT Level (Column|Sorting|Level)*>
<!ATTLIST Level
Mapping	(Flatten|Direct)
Links	(Embed|Traverse)	“Traverse”
>
<!ELEMENT Column EMPTY>
<!ATTLIST Column
Source	CDATA	#REQUIRED
Output	CDATA	#REQUIRED
MultiValue	(OnlyFirst|MultiLine|Concatenate)“   OnlyFirst”
MultiValueSeparator	CDATA	#IMPLIED “,”
>
<!ELEMENT Sorting SortDescription+>
<!ELEMENT SortDescription Group?|SortColumn+|Interval?>
<!ATTLIST SortDescription
Name	CDATA	#REQUIRED
>
<!ELEMENT SortColumn EMPTY>
<!ATTLIST SortColumn
Source	CDATA	#REQUIRED
Order	(Ascending|Descending)	#REQUIRED
DataType	CDATA	#REQUIRED
Strength	(Primary|Secondary|Tertiary|Identical) “Identical”
Decomposition (None|Canonical|Full) “None”
>
<!ELEMENT Group Group?|GroupColumn+>
<!ATTLIST Group
Grouping (Unique|Units) #REQUIRED
GroupUnits (Years|Months|Days|Hours)
AtGroupBreak (None|Count|Total) “None”
Order (Ascending|Descending)	#REQUIRED
Strength	(Primary|Secondary|Tertiary|Identical) “Identical”
Decomposition (None|Canonical|Full) “None”
>
<!ELEMENT GroupColumn EMPTY>
<!ATTLIST GroupColumn
Source	CDATA	#REQUIRED
>
<!ELEMENT Interval EMPTY>
<!ATTLIST Interval
Start	CDATA	#REQUIRED
End	CDATA	#REQUIRED
>

Every Collection has a name that is used to reference the collection. The Start attribute specifies how to find the “root” of the collection. A collection with a record root is just a set of records, whereas a collection that starts with an index is navigated through the index and then the set of records. An index may be a concordance or full-text. The optional Location attribute is a relative URL that identifies where in the root to actually begin.

A Level defines the contents of part of the output hierarchy. A level consists of the columns in the level, the ordering or grouping of records in the level, and definitions of sub-levels. A level is associated with records in the source record stream through the Mapping attribute. If the mapping is Direct, a level represents a single source record type. If the mapping is Flatten, the level contains a source record type and all descendants of that record. The Flatten mapping may only be specified on the only or lowest level in the collection. The Links attribute specifies how records with link attributes should handled. If links are Traversed, the record will be output as a distinct level. If links are Embedded, the child record of the source record will appear as though it is part of the source record.

A Column defines the mapping between a source field and the output array column. The Source attribute is a XSLT path expression in the source records. The Result attribute is a name of the field in the result array. The MultiValue and MultiValueSeparator attributes define how multi-valued source values are returned in the result.

Every collection must have at least one defined order. The order can be sorted collation or multi-level grouping with aggregate functions.

The SortColumn element defines the collation characteristics within a SortDescription. The Source attribute defines the name of the output column to be sorted. The Order must be either Ascending or Descending. The Strength and Decomposition values are input parameters that have the same meaning as defined in Unicode.

The two kinds of grouping are by unique values and by units. When a collection is grouped by unique values, all records with the same GroupColumn values will be together in the same group—breaks between groups will occur at the change of GroupColumn values. When a collection is grouped by units, all records with the same GroupColumn values, resolved to the value of GroupUnits, will be together in the same group. For example, if GroupUnits is “Days”, all records for a given day will be in the same group. If AtGroupBreak is specified, a synthetic row will be returned that contains the result of the aggregate function at each value or unit break value.

The GroupColumn identifies the result column to be grouped.

The Interval identifies the two fields in each record that define a range. The datatypes of the Start and End columns must be either numeric or datetime.

The following example shows a collection descriptor document for a simple document discussion record view with six collation orders:

<Collection Name=“Main” Start=“Record” Version=“0,1,0,0”>
<Level Mapping=“Flatten”>
<Column Source=“Title” Output=“Title”/>
<Column Source=“_Modified” Output=“_Modified”/>
<Column Source=“_CreatedBy” Output=“_CreatedBy”/>
<Sorting>
<SortDescription Name=“ByAscModified”>
<SortColumn Source=“_Modified” Order=“Ascending”
DataType=“DateTime”/>
</SortDescription>
<SortDescription Name=“ByDescModified”>
<SortColumn Source=“_Modified”
Order=“Descending” DataType=“DateTime”/>
</SortDescription>
<SortDescription Name=“ByAscAuthor”>
<SortColumn Source=“_CreatedBy”
Order=“Ascending” DataType=“String”/>
</SortDescription>
<SortDescription Name=“ByDescAuthor”>
<SortColumn Source=“_CreatedBy”
Order=“Descending” DataType=“String”/>
</SortDescription>
<SortDescription Name=“ByAscTitle”>
<SortColumn Source=“Title” Order=“Ascending”
DataType=“String”/>
</SortDescription>
<SortDescription Name=“ByOrdinal”>
<SortColumn Source=“” Order=“Ordinal”
DataType=“Long”/>
</SortDescription>
</Sorting>
</Level>
</Collection>

The following example shows a collection descriptor for a calendar view. Note the similarity to the prior example, but with a small change to the sort description, the collection is ordered by ranges of date intervals.

<Collection Name=“Main” Start=“Record” Version=“0,1,0,0”>
<Level Mapping=“Flatten”>
<Column Source=“from-attributes(Subject)”
Output=“Subject”/>
<Column Source=“from-attributes(Start)”
Output=“Start”/>
<Column Source=“from-attributes(End)”
Output=“End”/>
<Column Source=“from-attributes(RecurrenceEnd)”
Output=“RecurrenceEnd”/>
<Column Source=“from-attributes(IsAllDay)”
Output=“IsAllDay”/>
<Column Source=“from-attributes(IsRecurrent)”
Output=“IsRecurrent”/>
<Sorting>
<SortDescription Name=“DateRanges”>
<Interval Start=“Start” End=“End”/>
</SortDescription>
</Sorting>
</Level>
</Collection>

As is the basic storage manager, the collection manager is implemented in an object-oriented environment. Accordingly, both the collection manager itself and all of the collection components including collections, waffles, cursors, result arrays and the record set engine are implemented as objects. These objects, their interface, the underlying structure and the API used to interface with the collection manager are illustrated in FIG. 14. The API is described in more detail in connection with FIG. 15. Referring to FIG. 14, the collection manager provides shared access to collections, via the collection manipulation API 1402, but, in order to enable a full programming model for client applications, additional communication and synchronization operations are provided, within the context of a collection. For example, a user can control a record set engine 1412 by means of the engine API 1404. Under control of commands in the engine API 1404, the record set engine 1412 propagates a set of updates for a collection to the distributed virtual object system 1410 that is discussed above. Based on those updates, the distributed virtual object system 1410 updates index and other structures.

Other client components may need to be aware of changes within components, such as waffles, managed by the collection manager. Accordingly, the collection manager provides an interface 1400 to an interest-based notification system 1406 for those client components. The notification system 1406 provides notifications to client component listeners who have registered an interest when values within objects 1408 that represent a collection change;

Collection data is represented by a set of objects including collection objects, record objects, waffle objects, cursor objects and result array objects 1408. The objects can be directly manipulated by means of the collection manipulation API 1402. The collection related objects 1408 are actually implemented by the distributed virtual object system 1410 that was discussed in detail above.

FIG. 15 and the following tables comprise a description of the interfaces for each of the objects used to implement a preferred embodiment of the inventive collection manager. As with the storage manager implementation, these objects are designed in accordance with the Common Object Model (COM), but could also be implemented using other styles of interface and object model.

Table 37 illustrates an interface 1500 (IGrooveCollectionManager) for a collection manager that encapsulates the basic framework for the major operations performed on a collection. The collection manager interface includes the following methods:

TABLE 37
Interface IGrooveCollectionManager : IGrooveDispatch
CreateCollection(IGrooveElement  Creates a new collection object. The
*i_pCollectionDescriptor, BSTR   CollectionDescriptor should contain a
i_CollectionURL, BSTR i_EngineID, collection descriptor in XML according to the
IGrooveCollection **o_ppCollection); GrooveCollection XML DTD.
DeleteCollection(IGrooveXMLDocument Deletes the specified collection from the
*i_pSourceDocument, BSTR         SourceDocument.
i_CollectionURL);
OpenCollection(IGrooveElement    Opens an existing collection object.
*i_pCollectionDescriptor, BSTR
i_CollectionURL, BSTR i_EngineID,
IGrooveCollection **o_ppCollection);
OpenCollectionEnum(IGrooveXMLDocument * Return an enumeration of all collections within
i_pSourceDocument,               a document.
IGrooveBSTREnum
**o_ppCollectionNames);
ParseCollectionDescriptor(IGrooveElement * Creates a collection document according to
i_pCollectionElement, void *     the specified collection descriptor.
m_Levels);
UpdateCollection(void *i_Updates, Perform the requested sequence of
BSTR i_EngineID, IGrooveElement ** operations (of kind
o_ppUpdateContext);              GrooveCollectionUpdateOp) on the collection
                                 for EngineID.

Table 38 illustrates an interface 1502 (IGrooveCollection) for a collection that encapsulates the basic framework for the major operations performed on a collection. The collection interface includes the following methods:

TABLE 38
Interface IGrooveCollection : IGrooveDispatch
AdviseListeners(IGrooveElement   Notifies subscribing listeners of changes to this
*i_UpdateContext);               element.
CloseWaffle(IGrooveWaffle        Removes an IGrooveWaffle instance from the list
*i_pWaffle);                     of the collection's listeners.
Delete(void);                    Deletes the collection from the database.
DisableListeners (void);         Disables event notifications for all subscribing
                                 listeners.
EnableListeners (void);          Enables event notifications for all subscribing
                                 listeners. Event notifications are enabled by
                                 default, so this is only necessary if
                                 DisableListeners was previously called.
Find(BSTR i_pQuery,              Using the specified XSLT query expression,
IGrooveCollection **             evaluate it on the collection and return a new
o_ppQueryResult);                collection as the result.
                                 XSLT locators have the form:
                                 AxisIdentifier(Node Test Predicate)
                                 where AxisIdentifier is one of:
                                 from-ancestors
                                 from-ancestors-or-self
                                 from-attributes
                                 from-children
                                 from-descendants
                                 from-descendants-or-self
                                 from-following
                                 from-following-siblings
                                 from-parent
                                 from-preceding
                                 from-preceding-siblings
                                 from-self
                                 from-source-link
                                 NodeTest is of the form QName and tests
                                 whether the
                                 node is an element or attribute with the
                                 specified name.
                                 A Predicate is of the form [ PredicateExpr ]
                                 PredicateExpr is a Expr
                                 Expr is one of:
                                 VariableReference
                                 ( Expr )
                                 Literal
                                 Number
                                 FunctionCall
                                 Multiple predicates are separated by “/”
                                 For example:
                                 from-children(ElementName[from-attributes(Attribute
                                 Name)])
GetCursor(IGrooveCollectionCursor Returns a copy of the cursor currently used by the
**o_ppCursor);                   collection.
GetCursorPosition(double *       Returns the relative position of the cursor as a
o_pRelativePosition);            number between 0.0 (first row) and 100.0 (last
                                 row).
GetEngineMappingTable(void       Returns the engine mapping table.
**o_ppEngineURLs);
GetExpansionMask(long            Gets the current value of the expansion mask.
*o_pMask);
GetRecordCount(long *            Returns the number of records in the collection.
o_pRecordCount);
HasOrdinalSort(BSTR *            If the collection has an ordinal index, returns the
o_pSortName, VARIANT_BOOL        sort name and the value TRUE, otherwise it
*o_pHaveSort);                   returns FALSE.
HasSort(BSTR i_ColumnName,       Returns a bool indicating whether or not a sort
GrooveCollationOrder             exists in the collection for the column specified by
i_CollationOrder, long i_Level,  i_ColumnName on level i_Level in collation order
BSTR *o_pSortName,               i_AscendingSort. If a sort exists the sort name is
VARIANT_BOOL *o_pHaveSort);      returned in o_pSortName.
IsEmpty(VARIANT_BOOL             Returns a bool indicating whether or not the
*o_pIsEmpty);                    collection is empty.
MarkAll(VARIANT_BOOL i_Read);    Sets the record read/unread indicator for all
                                 records in the collection to be the value of Read.
MarkRead(double i_RecordID);     Sets a specific record to be marked as read.
MarkUnread(double i_RecordID);   Sets a specific record to be marked as unread.
MoveCursor(GrooveCollectionCursor Every collection has a cursor. The cursor
Position i_AbsolutePosition,     establishes the starting position in the source
GrooveCollectionNavigationOp     document, which will then be used to build the
i_Navigator, long i_Distance, long result document.
*o_pDistanceMoved);              AbsolutePosition may have the values First, Last,
                                 or Current.
                                 Navigator may have the following values:
                                 Value
                                 Description
                                 NextAny, PriorAny
                                 Move the cursor to the next/previous source row,
                                 traversing down through child rows and up
                                 through parent rows.
                                 NextPeer, PriorPeer
                                 Move the cursor to the next/previous source row
                                 at the same level, stopping if a row at a higher
                                 level is reached.
                                 NextParent, PriorParent
                                 Move the cursor to the next/previous parent
                                 source row, traversing until the root row is
                                 reached.
                                 NextData, PriorData
                                 Move the cursor to the next/previous row that
                                 contains a data record.
                                 NextUnread, PriorUnread
                                 Move the cursor to the next/previous unread row.
                                 Distance sets the numbers of iterations to move
                                 the cursor, starting at AbsolutePosition and
                                 moving through Distance iterations of Navigator
                                 movement.
                                 MoveCursor returns the number of iterations the
                                 cursor was actually moved.
MoveCursorToRecord(double        Sets the collection's cursor to point to the
i_RecordID);                     specified record.
MoveCursorToValue(BSTR           Using the current sort order, positions the cursor
i_pQuery, double * o_pRecordID); to the row that meets the criteria of matching the
                                 relop to the input query values. The relop
                                 (relational operator) may be EQ, LT, LE, GT, or
                                 GE. The query values must match, in order, the
                                 datatypes of the columns of the current sort order
                                 or must be able to be converted in a loss-less
                                 manner to those datatypes. Fewer query values
                                 may be specified than are defined in the sort
                                 order, which will result in a partial match. For
                                 collections ordered on an interval, the first query
                                 value is the interval's starting value and the
                                 second is the ending value.
MoveToCursor(IGrooveCollection   Moves the collection to the position specified by
Cursor *i_pCursor);              i_pCursor.
Open(BSTR i_CollectionURL,       Creates or opens the collection specified by
IGrooveElement                   I_CollectionURL within the Groove storage service
*i_pCollectionDescriptorElement, i_ServiceType. Returns a bool indicating whether
VARIANT_BOOL i_Temp,             or not the collection was created for the first time.
VARIANT_BOOL i_Shared,
VARIANT_BOOL * o_pCreated);
OpenRecord(double i_RecordID,    Returns an interface pointer to a specific record in
IGrooveRecord ** o_ppRecord);    the collection.
OpenRecordID(double              Starting from the position of the SourceRecordID,
i_SourceRecordID, enum           perform the specified collection navigation
GrooveCollectionNavigationOp     operation and return the resulting record ID.
i_Relation, double *
o_pTargetRecordID);
OpenResultArray(long             Given the collection's expansion mask, current
i_NumReturnRows, void            cursor position and current sort order, return at
*io_pResultArray);               most NumReturnRows into a result array
                                 conforming to the description below. Note that
                                 NumReturnRows is a quota only on the data rows -
                                 other synthesized header and footer rows may
                                 be returned as necessary.
                                 Column Name
                                 Data Type
                                 Description
                                 RowType
                                 UINT1
                                 ==WAFFLE_ROW_DATA if the row is a data
                                 record returned from an engine,
                                 ==WAFFLE_ROW_HEADER false if the row is a
                                 synthesized header (e.g., category),
                                 ==WAFFLE_ROW_FOOTER if the row is a
                                 synthesized footer (e.g., aggregate result).
                                 SynthKind
                                 UINT1
                                 If the row is a data row, this value is 0. If the row
                                 is a synthesized row, this value will be one of:
                                 BreakUnique: Indicates a change in value of
                                 categorized or sorted column. One of the
                                 ColumnName(i) columns will have the new
                                 value.
                                 BreakUnitDay
                                 BreakUnitWeek
                                 BreakUnitMonth
                                 BreakUnitYear
                                 FuncTotal
                                 FuncCount
                                 EngineID
                                 UINT4
                                 If the row is a data row: Index into the EngineID
                                 table, which is a vector of URLs stored as BSTRs.
                                 If the row is a synthesized row, EngineID is 0.
                                 RecordID
                                 UINT4
                                 If the row is a data row: RecordID returned from
                                 the engine identified by EngineID. RecordIDs are
                                 unique within EngineIDs.
                                 If the row is a synthesized row: RecordID is a
                                 unique number within the collection.
                                 Level
                                 UINT1
                                 Number of levels to indent this row. Level 0 is the
                                 top or outermost level.
                                 RelativePosition
                                 UINT2
                                 A number between 0 and 10000 indicating the
                                 relative offset of this row from the beginning of the
                                 collection. [It may be an approximation.] For
                                 example, 6823 is the value for a row that is
                                 68.23% of the way through the collection.
                                 Read
                                 BOOL
                                 If the row is a data row: True if the [account??]
                                 has read the record. If the row is a synthesized
                                 row, Read is always true (even if it is collapsed).
                                 ColumnName(i)
                                 Defined by the collection descriptor.
                                 Data value for this row/column. There will be as
                                 many columns in the array as there were defined
                                 columns at all levels.
OpenSchema(long i_Level,         Return an interface pointer to the schema
VARIANT_BOOL                     description for the records in the collection.
i_IncludeSystemColumns,
IGrooveRecordSchema
**o_ppCollectionSchema);
OpenTransaction(IGrooveTransaction Creates a transaction on the collection document.
**o_ppTransaction);
OpenWaffle(IGrooveWaffleListener Creates an IGrooveWaffle instance and adds it to
*i_pListener, IGrooveWaffle      the collections list of event listeners.
**o_ppWaffle);
SetCursorPosition(double         Sets the current position of the cursor to the row
i_RelativePosition);             with the specified relative position. The position
                                 should be a number between 0.0 (first row) and
                                 100.0 (last row).
SetExpansionMask(long i_Mask);   Sets the current value of the expansion mask.
                                 The mask is a stored in a DWORD, but only the
                                 first 10 (or so) bits are used. If a bit is set, all data
                                 the indicated level is expanded. The expansion
                                 mask is not persistent or shared —its effect is only
                                 on this collection object. The default value of the
                                 expansion mask is all 1s.
SetRecordExpansion(double        Sets the expansion state for a single row for this
i_RecordID, VARIANT_BOOL         scope. If Expand is true, the record will be
i_Expand);                       expanded, otherwise it will be collapsed. If
                                 EngineID is 0, then all rows encompassed by
                                 specified synthesized RecordID will be either
                                 expanded or collapsed.
Update(BSTR i_EngineURL,         Updates the collection. i_Operation is one of:
GrooveCollectionUpdateOp         OP_ADD, OP_DELETE, or OP_UPDATE.
i_Operation, void *
i_pUpdateRecord,
IGrooveElement *
io_pUpdateContext);
UseSort(BSTR i_SortName,         Sets the sort order for the collection to the named
VARIANT_BOOL                     sort order. The specified SortName must be one
i_RetainCursorPosition);         of the defined sort orders in the collection
                                 descriptor.
                                 If i_RetainCursorPosition is true and the current
                                 cursor position identifies a data record, the current
                                 collection's cursor is positioned to the same record
                                 in the new sort order. Otherwise, the cursor
                                 position is positioned to the first row in the new
                                 sort order.

Table 39 illustrates an interface 1504 (IGrooveCollectionListener) for a client of a collection manager that wishes to be notified whenever “significant” events happen within the collection. Significant events may occur at any time and include updating, addition, deletion, reparenting, or a change in ordinal position of a collection element. The collection manager listener interface includes the following methods:

TABLE 39
interface IGrooveCollectionListener : IGrooveDispatch
OnRecordChange(IGrooveElement Called when the data in this
*i_pElement);                 element has been updated
                              or the element has been
                              added, deleted, reparented,
                              or its ordinal position
                              has changed.
OnSortChange(void);           Called when the sort order
                              for the collection
                              changes.

Table 40 illustrates an interface 1506 (IGrooveCollectionCursor) for a client of a collection manager that wants to move a cursor within the collection. A collection may have one or more cursors active at any time. The collection manager cursor interface includes the following methods:

TABLE 40
interface IGrooveCollectionCursor : IGrooveDispatch
Move                            Moves the cursor in either an
(GrooveCollectionCursorPosition absolute or relative amount.
i_AbsolutePosition,             AbsolutePosition may have the values
GrooveCollectionNavigationOp    First, Last, or Current.
i_Navigator, long i_Distance,   Navigator may have the following
long                            values:
*o_pDistanceMoved);             Value
                                Description
                                NextAny, PriorAny
                                Move the cursor to the next/previous
                                source row, traversing down through
                                child rows and up through
                                parent rows.
                                NextPeer, PriorPeer
                                Move the cursor to the next/previous
                                source row at the same level,
                                stopping if a row at a
                                higher level is reached.
                                NextParent, PriorParent
                                Move the cursor to the next/previous
                                parent source row, traversing until the
                                root row is reached.
                                NextData, PriorData
                                Move the cursor to the next/previous
                                row that contains a data record.
                                NextUnread, PriorUnread
                                Move the cursor to the next/previous
                                unread row.
                                Distance sets the
                                numbers of iterations
                                to move the cursor, starting at
                                AbsolutePosition and moving through
                                Distance iterations of Navigator
                                movement.
                                Move returns the number of iterations
                                the cursor was actually moved.
OpenRecord                      Returns an interface pointer to the
(IGrooveRecord **               record the cursor is currently set at.
o_ppRecord);

The following tables illustrate allowed values for the enumerated data types listed in the above interfaces. In particular, Table 41, illustrates allowed values for the GrooveCollationOrder enumerated data type:

TABLE 41
GrooveCollationOrder
CollateAscending  Ordered by ascending data values.
CollateDescending Ordered by descending data values.
CollateOrdinal    Ordered by ordinal position.

Table 42 illustrates the allowed values for the GrooveCollectionNavigationOp enumerated data type:

TABLE 42
GrooveCollectionNavigationOp
NextAny     Move the cursor to the next source row,
            traversing down through child rows and up
            through parent rows.
PriorAny    Move the cursor to the previous source row,
            traversing down through child rows and up
            through parent rows.
NextPeer    Move the cursor to the next source row at the
            same level, stopping if a row at a higher level
            is reached.
PriorPeer   Move the cursor to the previous source row at
            the same level, stopping if a row at a higher
            level is reached.
NextParent  Move the cursor to the next parent source
            row, traversing until the root row is reached.
PriorParent Move the cursor to the previous parent source
            row, traversing until the root row is reached.
NextData    Move the cursor to the next row that contains
            a data record.
PriorData   Move the cursor to the previous row that
            contains a data record.
NextUnread  Move the cursor to the next unread row.
PriorUnread Move the cursor to the next unread row.

Table 43 illustrates the allowed values for the GrooveCollectionCursorPosition enumerated data type:

TABLE 43
GrooveCollectionCursorPosition
First   The first row in the collection.
Last    The last row in the collection.
Current The current row in the collection. This
        position is useful for performing relative cursor
        movement.

Table 44 illustrates the allowed values for the GrooveCollectionRowType enumerated data type:

TABLE 44
GrooveCollectionRowType
ROW_DATA   A row with data values.
ROW_HEADER A row header, for example, column break
           values.
ROW_FOOTER A row footer, for example, column break
           values and an aggregated result.

Table 45 illustrates the allowed values for the GrooveCollectionSynthType enumerated data type:

TABLE 45
GrooveCollectionSynthType
BreakUnique    Synthesized collection row indicates a change
               in value of categorized or sorted column. One
               of the other columns will have the new value.
BreakUnitDay   Synthesized collection row is a break on the
               change in units of days.
BreakUnitWeek  Synthesized collection row is a break on the
               change in units of weeks.
BreakUnitMonth Synthesized collection row is a break on the
               change in units of months.
BreakUnitYear  Synthesized collection row is a break on the
               change in units of years.
FuncTotal      Synthesized collection row is the result of an
               aggregate total function.
FuncCount      Synthesized collection row is the result of an
               aggregate count function.

Table 46 illustrates the allowed values for the GrooveCollectionUpdateOp enumerated data type:

TABLE 46
GrooveCollectionUpdateOp
OP_ADD            Add the record to the collection.
OP_DELETE         Delete the record from the collection.
OP_UPDATE         Change values of specific fields in this
                  record, which is already in the collection.
OP_REPARENT       Change this record's parent.
OP_CHANGE_ORDINAL Change the ordinal position of this
                  record in the collection.

Table 47 illustrates the allowed values for the GrooveCollectionWaffleSystem enumerated data type:

TABLE 47
GrooveCollectionWaffleSystemColumns
WAFFLE_ROWTYPE_COLUMN         One of the values for
                              GrooveCollectionRowType.
WAFFLE_SYNTHKIND_COLUMN       If not a data row, one of the values in
                              GrooveCollectionSynthType.
WAFFLE_RECORDID_COLUMN        A unique identifier for the record. The
                              RecordID must be unique within the
                              collection, but may not be unique in other
                              scopes.
WAFFLE_PARENT_RECORDID_COLUMN A reference to a parent record that contains
                              the recordID of a record in the collection. If
                              the record reference in the parent recordid is
                              deleted, this record will also be deleted from
                              the collection.
WAFFLE_LEVEL_COLUMN           The number of indention levels from the root
                              level of the hierarchy. The root level is 0.
WAFFLE_RELPOS_COLUMN          A number between 0.0 (first row) and 100.0
                              (last row).
WAFFLE_READ_COLUMN            A list of whoever has read this record. If this
                              field is not present, no users have read the
                              record.
WAFFLE_EXPANDED_COLUMN        A boolean indicator for whether the row is
                              collapsed or fully expanded.
WAFFLE_HASCHILDREN_COLUMN     A boolean indicator for whether the row has
                              children.

Table 48 illustrates the allowed values for the GrooveCollectionRecordID enumerated data type:

TABLE 48
GrooveCollectionRecordID
NULL_RECORD_ID The reserved value for the special null record
               id.

Table 49 illustrates the allowed values for the GrooveSortOrder enumerated data type:

TABLE 49
GrooveSortOrder
Ascending  Collate by ascending data values
Descending Collate by descending data values.

A software implementation of the above-described embodiment may comprise a series of computer instructions either fixed on a tangible medium, such as a computer readable media, e.g. a diskette, a CD-ROM, a ROM memory, or a fixed disk, or transmissible to a computer system, via a modem or other interface device over a medium. The medium can be either a tangible medium, including, but not limited to, optical or analog communications lines, or may be implemented with wireless techniques, including but not limited to microwave, infrared or other transmission techniques. It may also be the Internet. The series of computer instructions embodies all or part of the functionality previously described herein with respect to the invention. Those skilled in the art will appreciate that such computer instructions can be written in a number of programming languages for use with many computer architectures or operating systems. Further, such instructions may be stored using any memory technology, present or future, including, but not limited to, semiconductor, magnetic, optical or other memory devices, or transmitted using any communications technology, present or future, including but not limited to optical, infrared, microwave, or other transmission technologies. It is contemplated that such a computer program product may be distributed as a removable media with accompanying printed or electronic documentation, e.g., shrink wrapped software, pre-loaded with a computer system, e.g., on system ROM or fixed disk, or distributed from a server or electronic bulletin board over a network, e.g., the Internet or World Wide Web.

Although an exemplary embodiment of the invention has been disclosed, it will be apparent to those skilled in the art that various changes and modifications can be made which will achieve some of the advantages of the invention without departing from the spirit and scope of the invention. For example, it will be obvious to those reasonably skilled in the art that, although the description was directed to a particular hardware system and operating system, other hardware and operating system software could be used in the same manner as that described. Other aspects, such as the specific instructions utilized to achieve a particular function, as well as other modifications to the inventive concept are intended to be covered by the appended claims.

Claims

1. Apparatus for representing and managing an XML compliant document in a memory in a computer environment comprising a plurality of concurrently executing processes, the XML-compliant document being composed of a plurality of elements arranged in a nested relationship, the apparatus comprising: computer storage medium having: a plurality of element objects, each element object representing a respective element of the plurality of elements of the XML-compliant document, each element object providing a programming interface to information associated with the respective element of the plurality of elements;a mapping between elements of the plurality of elements and computer executable instructions that implement a respective element object of the plurality of element objects;computer-executable instructions that, when executed, provide a mechanism for arranging the plurality of element objects in a hierarchy representative of the nested relationship of the elements; andcomputer-executable instructions that, when executed, provide an application programming interface for applications to manipulate information associated with the plurality of elements, the application programming interface, in response to a call from an application program to manipulate an element of the plurality of elements, identifying from the mapping a respective element object for the element and returning to the application through the application programming interface a pointer to the respective element object, whereby the application can manipulate the element by invoking the computer executable instructions implementing the respective element object through the programming interface of the respective element object; anda schema document referenced by the XML-compliant document, the schema document containing element objects and the mapping,wherein: the mapping is contained within the schema document;the schema data is accessed using the computer-executable instructions that provide an application programming interface for applications to manipulate information associated with the plurality of elements of the XML-compliant document; andthe XML-compliant document is accessed from applications executing in each of the plurality of concurrently executing processes.

2. The apparatus as recited in claim 1 wherein at least some of the elements contain textual content and wherein element objects representing the elements provide the textual content through the programming interface associated with the element object.

3. The apparatus as recited in claim 1 wherein at least some of the elements contain attributes having values and wherein element objects representing the elements provide the attribute values through the programming interface associated with the element object.

4. The apparatus as recited in claim 3 wherein the attribute values contained in the at least some elements are typed.

5. The apparatus as recited in claim 3 further comprising an attribute index containing consistent pointers to all element objects containing attribute values.

6. The apparatus as recited in claim 1 wherein the arranging mechanism comprises database pointers and wherein a database pointer in a parent element object points to.

7. The apparatus as recited in claim 1 wherein the schema document is referenced by an XML processing statement in the data document.

8. The apparatus as recited in claim 1 further comprising a binary document object for representing a data document containing binary data.

9. The apparatus as recited in claim 1 further comprising a document object for representing the XML-compliant document.

10. The apparatus as recited in claim 9 wherein the document object contains links to other document objects so that the other document objects are sub-documents of the document object.

11. The apparatus as recited in claim 1 wherein each of the element objects exports a uniform interface containing methods for manipulating each of the element objects.

12. A method for representing and managing an XML-compliant document in a computer memory, the XML-compliant document comprising: information accessed from each of a plurality of concurrently executing processes, the XML-compliant document being composed of a plurality of elements arranged in a nested relationship, the method comprising: (a) creating a representation of a data document in the computer memory, the representation including;a plurality of element objects, each element object representing a corresponding part of the XML-compliant document and providing a programming interface, at least a portion of the plurality of element objects being stored as objects within a distributed shared memory system, anda mapping between each of a plurality of parts of the XML-compliant document and computer executable instructions that implement a corresponding element object of the plurality of element objects;(b) arranging the plurality of element objects in a hierarchy representative of the nested relationship of the elements;(c) accessing an element of the data document from a first of the concurrently executing processes through a first storage manager interface, the first storage manger interface accessing the mapping to generate a pointer to a corresponding element object representing the element, and the accessing further comprising using the pointer to access the programming interface of the corresponding element object of the plurality of element objects representing the element, whereby computer-executable instructions associated with the corresponding element object being-accessed manipulate data associated with the element through the distributed shared memory system;(d) accessing the element of the data document from a second of the concurrently executing processes through a second storage manager interface, the second storage manager interface accessing the mapping to generate a pointer to a corresponding element object representing the element, and the accessing further comprising using the pointer to access the programming interface of the corresponding element object of the plurality of element objects representing the element, whereby computer-executable instructions associated with the corresponding element object being accessed manipulate data associated with the element through the distributed shared memory system;(e) creating a schema document referenced by the data document in the computer memory, the schema document containing element objects and the mapping; and(f) accessing element objects of the schema document using the interfaces used to access the data document, wherein the XML-compliant document is the data document and the mapping is contained within the schema document, the schema document containing schema data represented by element objects, whereby the schema data is accessed using interfaces used to access the XML-compliant document.

13. The method as recited in claim 12 wherein at least some of the elements contain textual content and wherein element objects representing the elements contain the textual content.

14. The method as recited in claim 12 wherein at least some of the elements contain attributes having values and wherein element objects representing the elements contain the attribute values.

15. The method as recited in claim 14 wherein the attribute values contained in the at least some elements are typed.

16. The method as recited in claim 14 further comprising an attribute index containing consistent pointers to all element objects containing attribute values.

17. The method as recited in claim 12 wherein step (b) comprises creating a database pointer in a parent element object which pointer points to child objects of the parent element object in order to arrange the parent object and child objects in a hierarchical relationship.

18. The method as recited in claim 12 wherein step (e) comprises creating the schema document referenced by an XML processing statement in the data document.

19. The method as recited in claim 12 further comprising (e) creating a binary document object in the memory for representing a data document containing binary data.

20. The method as recited in claim 12 further comprising (e) creating a document object in the memory for representing the data document.

21. The method as recited in claim 20 wherein the document object contains links to other document objects so that the other document objects are sub-documents of the document object.

22. The method as recited in claim 12 wherein each of the element objects exports a uniform interface containing methods for manipulating each of the element objects.

23. A computer program product for representing and managing a first XML-compliant document in a memory in a computer environment comprising a plurality of concurrently executing processes, the first XML-compliant document being composed of a plurality of elements arranged in a nested relationship, the computer program product comprising a computer storage medium having computer readable program code thereon, including: program code creating a second XML-compliant document in the memory including a plurality of element objects, each element object representing a part of the first XML-compliant document, each of the element objects providing a programming interface to information associated with an element of the plurality of elements, the second XML-compliant document further comprising a registry mapping elements of the plurality of elements to element objects of the plurality of element objects; andprogram code arranging the plurality of element objects in a hierarchy representative of the nested relationship of the elements; andprogram code for accessing the registry based on an XML-compliant statement in the first XML-compliant document to select element objects in response to a request to manipulate elements of the first XML-compliant document,wherein: the second XML-compliant document comprises a schema document comprising schema data represented by element objects;the schema data is accessed using interfaces used to access the first XML-compliant document; andthe plurality of elements arranged in the nested relationship are accessible from each of the plurality of concurrently executing processes.

24. The computer program product of claim 23, further comprising program code for managing a telespace, the program code for managing the telespace accessing the program code for creating the data document to store telespace information in an XML-compliant document.

25. The computer program product of claim 24, further comprising program code for implementing a distributed memory system storing information associated with the XML-compliant document; and wherein at least a portion of the element objects comprises a method, accessible through the programming interface of the element object, for retrieving information about the XML-compliant document from the distributed memory system.