Method and apparatus for efficient management of XML documents

[0001] This invention relates to storage and retrieval of information and, in particular, to storage and retrieval of information encoded in Extended Markup Language (XML).

[0002] 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.

[0003] 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.

[0004] 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.

[0005] 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.

[0006] 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.

[0007] 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.

[0008] 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.

[0009] 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.

[0010] 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:

• • [0011] document text . . . <to >Recipient</to> . . . document text.

[0012] 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.

[0013] 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>.

[0014] 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.

[0015] 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.

[0016] 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.

[0017] 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.

[0018] 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.

[0019] 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.

[0020] 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.

[0038] 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.

[0039] 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.

[0040] 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.

[0041] 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.

[0042] 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.

[0043] 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.

[0044] 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.

[0045] 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:

• • [0046] 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:
    • [0047] <?xml version=“1.0” encoding=“UTF-8” standalone=“yes”?>
  • [0048] 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:
    • [0049] <!DOCTYPE Appl SYSTEM “app.dat”>
  • [0050] 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.

[0051] 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 property 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.

[0052] 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:

• • [0053] <?schema URI=“groovedocument:///GrooveXSS/$PersistRoot/sample.xml”?>

[0054] 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:

• • [0055] <!ELEMENT ThumbnailBox EMPTY >

[0056] 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:

[0057]

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

[0058] 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.

[0059] XML also permits custom definition statements similar to the #DEFINE statements used with some compilers. Commonly used definitions can be declared of within the DTD as “entities.” A typical entity definition could take the form:

• • [0060] <!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.

[0061] 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.

[0062] 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.

[0063] 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.

[0064] 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.

[0065] 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:

[0066]

<!-- The Document element is the root element in the schema -->
<!ELEMENT Document (Registry*, AttrGroup*, ElementDecl*)>
<!ATTLIST Document
	URL	CDATA	#REQUIRED
>
<!ELEMENT Registry TagToProg ID*>
<!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
>

[0067] 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.)

[0068] 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 Figure 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.

[0069] 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”.

[0070] 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.

[0071] 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:MA” 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.”

[0072] 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.

[0073] 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.

[0074] The following example shows the sample portions of a schema document for an XML document that defines a “telespace” that is previously described.

[0075]

<groove:Document URL=“TelespaceSchema.xml”
xmlns:groove=“um: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:AttGroup>
<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=“ECI” 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>

[0076] 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.: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.

[0077] 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.

[0078] 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 http://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.

[0079] 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.

[0080] Sub-document links follow the standard definition for simple links. An exemplary element definition of a link is as follows:

[0081]

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

[0082] 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.

[0083] 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 ftp://ftp.isi.edu/in-notes/rfc2557.txt/.

[0084] 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.

[0085]

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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==
--<<[[&&&]]>>--

[0086] 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.

[0087] 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.

[0088] 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.

[0089] 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.

[0090] 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 http://www.pdos.lcs.mit.edu/crl/.

[0091] 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.

[0092]

TABLE 1
DSM Method	Description
AddNotification(DSMRgn* i_pRgn, const	Adds a local event that will
IgrooveManualResetEvent * i_pEvent);	be signaled 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
i_CallbackParam, INCAddress	also atomically maps the
i_InitialOwner, DSMRId & io_RId,	new region and initiates
DSMRgn * & o_pRgn, void * & o_pData);	a StartWrite on the new
	region if Size is non-zero.
	Size is the initial size of
	the data in the new region.
	RId is identifier of the
	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
i_DatabaseNumber, TimeMillis	region resources.
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.
DSMRgnMapCallback * i_pCallback =	BroadcastGroup is the
NULL, void * i_pCallbackParam = NULL,	name of the group in which
BOOL * o_pMasterClient = NULL, UINT4	this DSM client belongs.
i_WaitTimeOut = 1000 UINT4 i_URCSize	URCSize is the size of the
= 1<<10, INCAddress * o_pAddress =	Unmapped Regions Cache.
NULL);	PAddress is the Inter-node
	Communication Address of
	this DSM client.
	pMasterClient specifies
	whether this DSM client
	is the Master(First) client.
Map(const DSMRId& i_RId, INT4	Maps the region to this
i_CallbackParam, BOOL i_InitialOwner);	client's memory space.
	RId is a valid identifier of
	the region to be mapped.
RemoveDatabase(UINT2	Removes the specified
i_DatabaseNumber);	database from the region
	mapping tables.
RemoveNotification(DSMRgn,* i_pRgn,	Removes interest in
const IGrooveManualResetEvent *	changes to data in a region.
i_pEvent);
Resize(DSMRgn* o_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
i_CallbackParam, void * & o_pData);	on the region's 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
INT4 i_CallbackParam, void * & o_pData);	read operation on 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
INT4 i_CallbackParam, void * & o_pData);	write operation on 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
i_CallbackParam, void * & o_pData);	on the region's 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.

[0093] 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.

[0094] 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.

[0095] 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.

[0096] 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.

[0097] 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.

[0098] 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.

[0099] 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.

[0100] 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.

[0101] 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.

[0102] The different DVO types are shown in Table 2.

[0103]

TABLE 2
DVO Type	Description
Binary	A kind of document that handles binary data.
Document
B-tree	The type of the root of a b-tree index. It contains a
Index	description of the index, as well as the address of the root
	index node.
Btree	A piece of a Btree index which can contain variable
Node	numbers of records, sorted by one or more keys.
Collection	A kind of document that handles Collection documents. In
Document	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	A type implementation of extendible hashing, as defined in
Hashing	“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.
Flat-	A specific kind of CollectionDocument used in shared
Colllection-	regions.
Document
Flat-	A specific kind of XMLDocument used in shared regions.
Document
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	A kind of index which supports key ordered sorting
Bucket	(integer, double, string)
Ordered	A type that provides a collated data vector. It has methods
Index	for adding, removing, and changing key/data pairs,
	managing index cursors, and managing parent and sub-
	indicies.
Ordered	Data types, called records and fields, that can be stored in
Index	ordered indices.
Type
Ordinal	A kind of index that support ordinal addressing. It is
Ordered	conceptually similar to vector that allows any entry to be
Index	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	A kind of ordered index that implements balancing using
Index	the red-black binary tree algorithm.
W32Binary-	A specific kind of binary document for 32-bit Windows
Document	platforms.
XML	A kind of document that handles XML documents. In
Document	addition to the Document methods, it has methods to handle
	schemas and indexes.

[0104] 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.

[0105] 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.

[0106] 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.

[0107] 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.

[0108] 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.

[0109] 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.

[0110] 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.

[0111] 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.

[0112] 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.

[0113] 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.

[0114] 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.

[0115] 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.

[0116] 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.

[0117]

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

[0118] 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.

[0119]

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

[0120] 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.

[0121]

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

[0122] 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.

[0123]

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

[0124] 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.

[0125]

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

[0126] Table 8 illustrates the methods for an interface 1102 (IGrooveCrossProcessSemaphore) for a client of a storage manager that needs to synchronize access among processes.

[0127]

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

[0128] 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 subclass 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:

[0129]

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 i_ReadOnly);	Starts a transaction. If
	ReadOnly is false, the
	database may be updated.
BeginIndependent (VARIANT_BOOL	Starts another transaction for
i_ReadOnly);	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.

[0130] 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:

[0131]

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

[0132] 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:

[0133]

TABLE 11
Interface IGrooveXMLDocument:IGrooveDocument
GenerateGrooveID (BSTR	Generates an 8 byte identifier
i_GrooveIDBase, double *	from the string identifier
o_pGrooveID);	i_GrooveIDBase.
ConvertGrooveIDToSerializedGrooveID	Converts an 8 byte identifier
(double i_GrooveID, BSTR *	to the string i_GrooveID.
o_pGrooveIDString);
ConvertSerializedGrooveIDToGrooveID	Converts a string version of a
(BSTR i_GrooveIDString, double *	Groove identifier to an
o_pGrooveID);	8 byte version.
CreateElement (BSTR i_Name,	Creates a new element with
IUnknown * i_pParent, IGrooveElement	the supplied Tag; the tag
** o_ppElement);	cannot be altered once
	created. If a Parent reference
	is supplied, the new element
	is created as a child of that
	parent.
CreateElementCopy (IGrooveElement *	Does a deep/shallow copy of
i_pSource, IGrooveElement *	the specified element and all
i_pParent, VARIANT_BOOL	of its children (recursively for
i_ShallowCopy, IGrooveElement **	deep; just the one level for
o_ppElement);	shallow), pulling the new
	element(s) in under the
	Parent element.
CreateElementFromSchema (BSTR	Creates an element that
i_Name, IGrooveElement * i_pParent,	conforms to the element's
IGrooveElement ** o_ppElement);	definition in the schema.
	Creates the element, its
	attributes, and any child
	elements.
CreateElementFromStream	Using a parser, creates an
(IGrooveByteInputStream * i_pStream,	element, reads from a byte
GrooveParseOptions i_ParseOptions,	input stream and creates
IUnknown * i_pParent, IUnknown *	elements and attributes
i_pLinkCallback, IGrooveElement **	from the text stream as
o_ppElement);	necessary, inserting them
	into the element, which is
	then returned to the caller.
	If a Parent reference is
	supplied, the new element is
	created as a child of that
	parent.
CreateLocator (IGrooveLocator **	Returns the interface to
o_ppLocator);	a new locator object.
FindElementByID (BSTR i_ID,	Looks for an element of the
IGrooveElement ** o_ppElement,	specified ID and returns a
VARIANT_BOOL * o_pFound);	boolean value if found.
OpenElementByID (BSTR i_ID,	Looks for an element of
IGrooveElement ** o_ppElement);	the specified ID.
OpenElementEnumByAttributeValue	Returns an enumeration of all
(BSTR i_ElementName, BSTR	of the elements within the
i_AttributeName, BSTR	document that have the named
i_AttributeValue, IGrooveElementEnum	attribute with the specified
**o_ppElementEnum);	value.
OpenElementEnumByAttributeValueAs	Returns an enumeration of all
Bool (BSTR i_ElementName, BSTR	of the elements within the
i_AttributeName, VARIANT_BOOL	document that have the named
i_AttributeValue, IGrooveElementEnum	attribute with the specified
**o_ppElementEnum);	boolean type value.
OpenElementEnumByAttributeValueAs	Returns an enumeration of
Double (BSTR i_ElementName, BSTR	all of the elements within
i_AttributeName, double	the document that have the
i_AttributeValue, IGrooveElementEnum	named attribute with the
**o_ppElementEnum);	specified double floating
	type value.
OpenElementEnumByAttributeValueAs	Returns an enumeration of
Long (BSTR i_AttributeName, long	all of the elements within
i_AttributeValue, IGrooveElementEnum	the document that have the
**o_ppElementEnum);	named attribute with the
	specified long integer
	type value.
OpenElementEnumByLocator (BSTR	Returns an element
i_LocatorText, IGrooveElementEnum **	enumerator with references
o_ppElementEnum);	to all elements satisfying
	the 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
i_Name, IGrooveElementEnum **	all of the elements within
o_ppElementEnum);	the document that have the
	specified tag name.
OpenMetaElement (IGrooveElement **	Returns the interface to the
o_ppElement);	meta element that defines
	this XML document.
OpenRootElement (IGrooveElement **	Opens the root element for the
o_ppRootElement);	XML document.

[0134] 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.

[0135]

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.

[0136] 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 http://www.w3.org/TR/xslt. The storage manager locator interface includes the following methods:

[0137]

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

[0138] 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:

[0139]

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

[0140] 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:

[0141]

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
VARIANT_BOOL i_ShallowCopy,	for deep; just the one level for shallow),
IGrooveElement ** o_ppElement);	putting 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
(GrooveByteInputStream ** 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 the 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 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.
FindContentElementByNameAnd	Within the context of this element, find an
Attribute (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 the 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 the type in the
	document schema.
OpenAttributeAsVARIANT (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.
OpenContentElementByNameAnd	Within the context of this element, find an
Attribute (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 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
(IGrooveMultiReaderElementQueue	the element and add a reader. This could
Reader ** 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.
IGrooveMultiReaderElementQueue
Writer ** o_ppQueue);
OpenMultiReaderElementReference	Returns the interface to the multi-reader
QueueReader	element reference queue reader object.
(IGrooveMultiReaderElementQueue
Reader ** o_ppQueue);
OpenMultiReaderElementReference	Returns the interface to the multi-reader
QueueWriter	element reference queue writer object.
(GrooveMultiReaderQueueOptions
i_Options,
IGrooveMultiReaderElementQueue
Writer ** o_ppQueue);
OpenName (BSTR * o_pName);	Returns the element's tag name.
OpenParent (IGrooveElement **	Get 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.
Open URI (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.
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 as text.
i_Value);
SetAttributeAsBinary (BSTR i_Name,	Sets any arbitrary 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.

[0142] 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.

[0143]

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

[0144] 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:

[0145]

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

[0146] 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:

[0147]

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);

[0148] 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:

[0149]

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.

[0150] 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:

[0151]

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

[0152] 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:

[0153]

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

[0154] 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 subclass 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:

[0155]

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

[0156] 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:

[0157]

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);

[0158] 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:

[0159]

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

[0160] 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:

[0161]

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

[0162] 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:

[0163]

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

[0164] 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.

[0165]

TABLE 27
interface IGrooveElementRPCServerThread:IGrooveElementUtilBase
(none)

[0166] 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:

[0167]

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

[0168] 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.

[0169]

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.

[0170] Table 30 illustrates the allowed values for the GrooveSerializeOptions enumerated data type.

[0171]

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.

[0172] Table 31 illustrates the allowed values for the GrooveParseOptions enumerated data type.

[0173]

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.

[0174] Table 32 illustrates the allowed values for the GrooveContentType enumerated data type.

[0175]

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.

[0176] Table 33 illustrates the allowed values for the GrooveXLinkShow enumerated data type.

[0177]

	TABLE 33
	GrooveXLinkShow
	GrooveXLinkShowNew	New.
	GrooveXLinkShowParsed	Parsed.
	GrooveXLinkShowReplace	Replace

[0178] Table 34 illustrates the allowed values for the GrooveXLinkActuate enumerated data type:

[0179]

	TABLE 34
	GrooveXLinkActuate
	GrooveXLinkActuateUser	User.
	GrooveXLinkActuateAuto	Auto.

[0180] Table 35 illustrates the allowed values for the GrooveXLinkSerialize enumerated data type.

[0181]

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

[0182] Table 36 illustrates the allowed values for the GrooveMultiReaderQueueOptions enumerated data type.

[0183]

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.

[0184] 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.

[0185] 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.

[0186] The following list shows the XML DTD contents for a collection descriptor document:

[0187]

<!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
>

[0188] 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.

[0189] 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.

[0190] 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.

[0191] Every collection must have at least one defined order. The order can be sorted collation or multi-level grouping with aggregate functions.

[0192] 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.

[0193] 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.

[0194] The GroupColumn identifies the result column to be grouped.

[0195] 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.

[0196] The following example shows a collection descriptor document for a simple document discussion record view with six collation orders:

[0197]

<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>

[0198] 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.

[0199] 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.

[0200] 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.

[0201] 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.

[0202] 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:

[0203]

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

[0204] 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:

[0205]

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
	(AttributeName)])
GetCursor(IGrooveCollection	Returns a copy of the cursor currently used by the
Cursor **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	Sets the record read/unread indicator for all
(VARIANT_BOOL_i_Read);	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 return's 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
	(relocation 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 an 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.

[0206] 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:

[0207]

TABLE 39
interface IGrooveCollection Listener:IGrooveDispatch
OnRecordChange(IGrooveElement	Called when the data in this element
*i_pElement);	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.

[0208] 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:

[0209]

TABLE 40
interface IGrooveCollectionCursor:IGrooveDispatch
Move(GrooveCollectionCursor	Moves the cursor in either an absolute
Position i_AbsolutePosition,	or relative amount.
GrooveCollectionNavigationOp	AbsolutePosition may have the values
i_Navigator, long i_Distance,	First, Last, or Current.
long *o_pDistanceMoved);	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.
	Move returns the number of iterations
	the cursor was actually moved.
OpenRecord (IGrooveRecord **	Returns an interface pointer to the
o_ppRecord);	record the cursor is currently set at.

[0210] 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:

[0211]

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

[0212] Table 42 illustrates the allowed values for the GrooveCollectionNavigationOp enumerated data type:

[0213]

	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.

[0214] Table 43 illustrates the allowed values for the GrooveCollectionCursorPosition enumerated data type:

[0215]

	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.

[0216] Table 44 illustrates the allowed values for the GrooveCollectionRowType enumerated data type:

[0217]

	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.

[0218] Table 45 illustrates the allowed values for the GrooveCollectionSynthType enumerated data type:

[0219]

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.

[0220] Table 46 illustrates the allowed values for the GrooveCollectionUpdateOp enumerated data type:

[0221]

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

[0222] Table 47 illustrates the allowed values for the GrooveCollectionWaffieSystem enumerated data type:

[0223]

TABLE 47
GrooveCollectionWaffleSystemColumns
WAFFLE_ROWTYPE—	One of the values for
COLUMN	GrooveCollectionRowType.
WAFFLE_SYNTHKIND—	If not a data row, one of the values in
COLUMN	GrooveCollectionSynthType
WAFFLE_RECORDID—	A unique identifier for the record. The
COLUMN	RecordID must be unique within the
	collection, but may not be unique in other
	scopes.
WAFFLE_PARENT—	A reference to a parent record that contains
RECORDID_COLUMN	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—	The number of indention levels from the root
COLUMN	level of the hierarchy. The root level is 0.
WAFFLE_RELPOS—	A number between 0.0 (first row) and 100.0
COLUMN	(last row).
WAFFLE READ—	A list of whoever has read this record. If this
COLUMN	field is not present, no users have read the
	record.
WAFFLE_EXPANDED—	A boolean indicator for whether the row is
COLUMN	collapsed or fully expanded.
WAFFLE—	A boolean indicator for whether the row has
HASCHILDREN—	children.
COLUMN

[0224] Table 48 illustrates the allowed values for the GrooveCollectionRecordID enumerated data type:

[0225]

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

[0226] Table 49 illustrates the allowed values for the GrooveSortOrder enumerated data type:

[0227]

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

[0228] 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.

[0229] 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.