Section 2.5.2. The organization of a CIF dictionary

The precision and efficiency of a data definition language are directly related to the scope of the attribute vocabulary. In other words, the lexical richness of the DDL depends on the number and the specificity of the available language attributes. The breadth of these attributes, in terms of the number of separate data characteristics that can be specified, largely controls the precision of data definitions. However, it is the functionality of attributes that determines the information richness and enables higher-level definition complexity. For example, the attributes that define child–parent relationships between data and key pointers to items in list packets are essential to understanding the data hierarchy and to its validation. Attributes provide the semantic tools of a dictionary.

The choice and scope of attributes in the DDL are governed by both semantic and technical considerations. Attributes need to have a clear purpose to facilitate easy definition and comprehension, and their routine application in automatic validation processes. A CIF dictionary is much more than a list of unrelated data definitions. Each definition conforms to the CIF syntax, which requires each data block in the dictionary to be unique. However, the functionality of a dictionary involves elements of both relational and object-oriented processes. For example, attributes in one definition may refer to another definition via _list_link_parent or _list_link_child attributes, so as to indicate the dependency of data items in lists. In this way the DDL, and consequently the dictionaries constructed from the DDL, invoke aspects of relational and object-oriented database paradigms. It is therefore useful to summarize these briefly here.

A relational database model (Kim, 1990) presents data as tables with a hierarchy of links that define the dependencies among tables. These explicit relationships enable certain properties of data to be shared, and, for related data values, to be derived. The structure of data links in a relational database is usually defined separately from the component data. This is an important strength of this approach. However, when data types and dependencies change continually, static relationships are inappropriate, and there is a need for non-relational extensions.

The object-oriented database model (Gray et al., 1992) allows data items and tables to be defined without static data dependencies. A data item may be considered as a self-contained `object' and its relationships to other objects handled by `methods' or `actions' defined within the objects. A database may have a base of statically defined explicit relationships with a dynamic layer of relationships provided by presenting some (or all) items as objects. Objects have well defined attributes, some of which may involve relationships with other data items, but objects need not have preordained links imposed by the static database structure.

The attributes and the functionality of CIF dictionaries incorporate aspects of both the basic relational and the object-oriented model. These provide the flexibility and extensibility associated with object-oriented data, as well as the relational links important to data validation and, ultimately, data derivation.