hierarchical data model

In the hierarchical data model, information is organized as a collection of inverted trees of records. The inverted trees may be of arbitrary depth. The record at the root of a tree has zero or more child records; the child records, in turn, serve as parent records for their immediate descendants. This parent-child relationship recursively continues down the tree.

The records consists of fields, where each field may contain simple data values (e.g. integer, real, text), or a pointer to a record. The pointer graph is not allowed to contain cycles. Some combinations of fields may form the key for a record relative to its parent. Only a few hierarchical DBMSs support null values or variable-length fields.



Applications can navigate a hierarchical database by starting at a root and successively navigate downward from parent to children until the desired record is found. Applications can interleave parent-child navigation with traversal of pointers. Searching down a hierarchical tree is very fast since the storage layer for hierarchical databases use contiguous storage for hierarchical structures. All other types of queries require sequential search techniques.

A DDL for hierarchical data model must allow the definition of record types, fields types, pointers, and parent-child relationships. And the DML must support direct navigation using the parent-child relationships and through pointers. Programs therefore navigate very close to the physical data structure level, implying that the hierarchical data model offers only very limited data independence.

The hierarchical data model is impoverished for expressing complex information models. Often a natural hierarchy does not exist and it is awkward to impose a parent-child relationship. Pointers partially compensate for this weakness, but it is still difficult to specify suitable hierarchical schemas for large models.