Deciding Whether to Use the R-Tree Access Method

The R-tree secondary access method is specifically designed to index data with the following two special properties:

• The data is multidimensional.
• On a given dimension, a data object spans some width. That is, it corresponds to an interval or range, not a point.

Examples of these types of data include:

• Two-dimensional spatial objects, such as points, lines, and polygons
• Geographic mapping information, defined in terms of latitude and longitude, that includes pointlike objects, such as cities; linelike objects, such as roads and rivers; and regionlike objects, such as counties, states, and land masses
• Video or audio clips, each with a start and stop time

  If you create a time range user-defined data type, you can search for overlapping clips more efficiently with an R-tree index than with a B-tree index.

• Color information that includes hue, brightness, and saturation
• Multidimensional views of standard relational quantitative data, such as age, salary, sales commission, hire date, and so on

An R-tree index works on data with only one of these properties (multi-dimensional points or ranges along a single dimension) but data corresponding to points on a single dimension is better indexed with a B-tree index.

Unlike other data structures, such as a grid-file and a quad-tree, the R-tree access method does not require that data values be in a known bounded area.

If you are developing a DataBlade module that includes a user-defined data type of a multidimensional or interval nature, you might want to use the R-tree access method to index columns of this data type.

The type of data most suited to B-tree indexes (the other indexing method included in Dynamic Server) is ordered numeric values in one dimension. Do not use B-tree indexes to index range or interval data. The following types of data are suited to being indexed with the B-tree access method and not the R-tree access method:

• Numerical data, such as employee IDs
• Character data, such as last names and product names

After you decide to use the R-tree access method to index a user-defined data type, you must create a new operator class. Creating a New Operator Class describes this process. The next section describes issues you should be aware of when you design the user-defined data type.