Persistent storage model when adding a value

The overall model of persistent storage consists of four different modules. The programs using the storage subsystem appear at the top of the figure. Your program first creates a new disk dictionary. The act of creating the dictionary results in the modules being created automatically. Once the dictionary is created, all of the member function calls you make deal directly with the dictionary.

Even though the figure shows different modules, for the most part you are not calling them directly. You do need to call modules directly, however, when extending the system,. For more information on extending the system, see "Extending persistent storage" on page 124.

The default index manager supports enumeration of all of the keys in a dictionary, but does not guarantee any specific ordering of the keys. You basically get the first key, then the next, and so on, in random order. Because keys have no defined order, you cannot request a range of keys in one lookup--for example keys greater than 10 and less than 20.

If you had an application in which it was important to specify a range of keys (implying a ranking or ordering of keys), you could implement a sorted index extension to the basic protocol offered by disk dictionaries. See the TSortedIndex example in "Extending persistent storage" on page 124. The TSortedIndex example shows how to enumerate dictionary keys in a particular order and also shows how to extend the persistent storage system in general.

The following actions occur when you add a key-value pair to a dictionary:

• Your program writes the key-value pair to the disk dictionary.
• The dictionary passes the value to the record manager.
• The record manager writes the value to disk and passes a record handle back to the dictionary.
• The dictionary associates the key with the record handle and passes the key and record handle to the index manager. The index manager writes the key and record handle to disk.