Databases Reference
In-Depth Information
are called foreign keys if their elements are values of the primary key of another
table [
38
].
Basic operations for data retrieval in the relational structure are set operations,
such as, union, intersection, and difference and further typical operation are [
69
]:
•
Selection: the restriction of the returned tuples to a subset defined by a condition
(also called filter). A term often used in context with selection is selectivity. It
is a measure of the size of the returned subset of tuples. High selectivity means
that the condition applies to only a small number of tuples, which are returned.
Low selectivity, in contrast, means that the condition is applicable to many tuples,
which are then returned in the result set.
•
Projection: returns a specified subset of attributes from a table or result set for
the selected tuples.
•
Joins: a conjunction of one table with itself or between different tables based on
conditions defined on attributes.
•
Grouping: collocation of tuples in the result set based on the values of specified
attributes.
•
Aggregation: summary of specified attributes according to a given function, such
as, average or summation, and a given grouping.
•
Sorting: ordering of tuples in the result set according to the values of specified
attributes.
Entry point for data retrieval can be any table defined in the database. Navigation
is possible through any compatible pair of attributes of one table or different tables
rather than through predefined sets linking the tables.
Discussion of Data Models in Transaction Processing
Similar to file system models, the early hierarchical and network data models and
applications relying on them were connected too closely. Data was manipulated
via standard host programming languages. As a result, changes to the data storage
in order to leverage latest advances in hardware technology induced changes in
the applications to preserve functionality. This process endangered a companies
investment in existing applications, that would have to be changed over and over
again, causing additional costs [
84
].
Hierarchical models are very efficient for large amounts of data and provide
high transaction throughput, but do not allow very flexible relations and require
extensive application programming to use the database. Network models are more
flexible regarding data access paths and many-to-many relationships are easier to
implement than in the hierarchical model, but their structure can easily become
very complicated [
151
, Chap. 20]. In both, network as well as hierarchical models,
the database structure has to be designed around the access paths that applications
use most often. Logical relationships within hierarchical databases provide the
possibility to define additional access paths if needed, incurring some overhead
through extra pointers, but reducing redundancy.
Search WWH ::
Custom Search