A.4 Property Set Definition Requirements (PSDR)

A.4.1.1 Property sets

A property set is defined in a "property set definition document", which is an SGML document that conforms to the "property set definition architecture" defined in this clause. The property set definition architecture meta-DTD is identified by the public identifier "ISO/IEC 10744:1997//NOTATION AFDR ARCBASE Property Set Definition Architecture//EN".

Note that, although they are technically architectural forms, the element types and attribute definitions that appear in this clause are normally used directly as the document type of property set definition documents. They are therefore referred to as element types in this document.

A property set consists of "property set components": classes, properties, enumerated values (of properties), and normalization rules (for comparing string property values). Some property set components are deemed to occur in all property sets; they are known as "intrinsic components". The property set definition architecture defines an element type for each component type; elements conforming to those types define components of the corresponding types.

The components of a property set can be organized into one or more separate "modules". Modules allow classes and properties to be grouped to reflect important distinctions among them that might be a basis for including or excluding them from grove plans.

NOTE 446 In the SGML property set, these distinctions include:

1. the applicable specification document (that is SGML, DSSSL, or HyTime)

2. whether the property exists only when an optional facility is used (e.g. implicit link); and

3. whether the property relates to the source data (e.g. general delimiters), to the abstraction described by the source data (e.g. elements and attributes), or to both (e.g. data characters).

A.4.1.2 Classes and properties

A class identifies a type of information, giving it a name and "providing" an ordered set of properties. Each property is also given a name, as well as a "property number".

Class names are unique within the property set that defines them; property names are unique within the class that provides them.

Property numbers are assigned to the properties provided by a class according to the order of their definition within the property set definition document. The first property provided by a class is assigned the number 1; the number of each subsequently defined property is determined by adding one to the number of the nearest previously defined property provided by the same class.

NOTE 447 For the purpose of determining the property number, it is immaterial in which module a property definition occurs, or whether it occurs in the content of a class definition or independently.

NOTE 448 Because intrinsic property definitions are considered present at the beginning of each property set definition document, the first few properties of every class are intrinsic properties; this has the effect that the property numbers assigned to the intrinsic properties are the same for all classes.

Property numbers remain constant for all groves, regardless of which properties are included in any given grove plan.

A property has a "declared datatype" which specifies the type of value nodes are permitted to exhibit for the property. A declared datatype is one of the following:

node

A node.

nodelist

An ordered list of zero or more nodes.

nmndlist

An ordered list of zero or more nodes, in which each node exhibits a "name property" whose value uniquely identifies the node within the list. A named node list is also called a "name space". The datatype of the name property must be string or node, and all of the name properties within a single named node list must have the same datatype.

Note that in effect, there are two varieties of named node list, "string-named" and "node-named". In the first, each node is identified by a string; in the second, each node is identified by another node. Collecting all the values of the "name" property of the nodes in a string-named node list yields a list of names that are unique within the list. Similarly, collecting all the values of the name properties of the nodes in a node-named node list yields a list of nodes that are unique within the list.

enum

A value which represents one of an enumerated set of values; an enumerator.

char

An abstract character. The concrete representations of "char" values in groves must allow any "char" value to be distinguished from all other "char" values, and for the semantics of each "char" to be determined. The concrete representation may or may not rely on the DCS, UNICODE, and/or other character sets and repertoires.

string

An ordered list of zero or more abstract characters.

strlist

An ordered list of zero or more strings.

integer

An integer.

intlist

An ordered list of zero or more integers.

boolean

A boolean value.

compname

A property set component name.

cnmlist

An ordered list of zero or more component names.

The node, enum, character, string, integer, and compname datatypes are known as "primitive datatypes"; the nodelist, strlist, intlist, and cnmlist datatypes are known as "list datatypes".

NOTE 449 The string datatype, though described as a list of characters, is still considered a primitive datatype.

A property whose declared datatype is node, nodelist, or nmndlist is said to be "nodal". Properties with other declared datatypes are "non-nodal".

A property whose declared datatype is string may have an associated "normalization rule", which should be consulted when making comparisons involving values exhibited for the property, including looking up a node by its name within a string-named node list.

A property definition can specify conditions under which the property will not apply to an instance of the class that provides it. A node for which a property is inapplicable exhibits a "null value" for the property. (Note that this is not the same as exhibiting an empty value for a list property.)

A.4.1.3 Nodes

A node is an ordered set of "property assignments", each of which associates a property name with a "value". The property names are the names of the properties provided by the node's class, the ordering of which determines the ordering of the property assignments.

A node is said to "exhibit" a value "for" a property if that node has a property assignment associating a value with the name of that property. Informally, the "properties of a node" are the properties for which the node exhibits a value. A node is said to be the "owner" of such properties.

A.4.1.4 Groves

Nodes exist in a directed graph in which each node may be connected to other nodes by labeled arcs. For each node occurring in the value exhibited by a node for a nodal property, there is one such arc from the exhibiting node to the value node, the label for which is the name of the nodal property.

NOTE 450 Informally, a node "points at" another node by having the node occur in the value of one of its properties.

The nodes occurring in the value of a nodal property are related to the node that exhibits the property by one of three possible "node relationship types": "subnode", "irefnode" (internal-reference node), or "urefnode" (unrestricted-reference node). Nodal properties can be qualified by their associated relationship type as "subnode properties", "irefnode properties", or "urefnode properties"; likewise, the arcs in the graph can be qualified by the relationship type assigned to the property to which the arc corresponds.

A node occurring in the value of a subnode property is called a "subordinate node" (or subnode) with respect to the node exhibiting the value, and the node exhibiting the value is the called the "origin" of the subnode. Each node may occur in the value of only one subnode property. The "siblings" of a node occurring in the value of a subnode property are the other nodes, if any, occurring in the value of the same subnode property. The subnode arcs of the graph connect nodes together into "subnode trees".

Irefnode, or "internal referenced node", arcs further connect nodes within one subnode tree. Such arcs may cause the graph to contain cycles and convergences.

A grove is a set of nodes connected together as a subnode tree and further connected by the irefnode arcs between the nodes. Each grove has exactly one node that is the root of the subnode tree; this node is the "grove root", and it is the only node within the grove that has no origin.

The remaining category of nodal properties (and therefore arcs) is urefnode, or "unrestricted referenced node". A urefnode arc connects a node to nodes in the same or other groves. A set of groves thus connected is a "hypergrove".

NOTE 451 A hypergrove should not be confused with a hyperdocument or the hypergrove that may result from the processing of a hyperdocument. Any set of groves connected by urefnode arcs is a hypergrove, not just those groves resulting from the processing of hyperdocuments.

A.4.1.5 Content trees

At most one property of a node is designated the "content property" of the node. The content property of a node must be either a subnode property or a string or character property.

If the content property of a node is a subnode property, the property is also the "children property" of the node. The nodes occurring in the value exhibited for the children property of a node are called the "children" of the node, and the node itself is called the "parent" of the children. A node that has children but does not have a parent is a "content tree root". The set of nodes reachable from a content tree root through children properties forms an ordered tree called a "content tree".

NOTE 452 If a node has a parent, then the parent is also the node's origin.

NOTE 453 The fact that a grove could contain a collection of disjoint content trees is the reason why it is called a grove.

NOTE 454 Tree addressing of a grove is normally based on content trees rather than on the subnode tree. That is because subnode tree ordering depends on the order of property definitions while content tree ordering depends on the order of node lists in property values -- that is, on the real information represented by the grove.

For this reason, when referring to a grove the unmodified term "tree" means "content tree".

If the content property of a node is a character or string property, the property is also known as the "data property" of the node. The data of the node is the value of the data property. The data of a node with a children property is the data of each of its children, separated by the value of the node's "data separator property", if it has one.

A.4.1.6 Grove plan application

For any given processing context, a grove plan may be used to specify what classes of node and what properties are significant within that context. This information can be used to prevent a process from wasting resources on inconsequential data while constructing a grove, or may be used as a mask over an existing grove, hiding those parts that are not of interest.