Information Technology Reference
In-Depth Information
additional columns offer a simple way of allowing for extensions and refinements.
For example, new reserved columns, such as
color
, could be introduced to extend
the proposed visualisations and visually group the result set by a specified range
of colours.
In addition to this augmentation of view definitions, we further extend SQL
with
VIEWER
and
BROWSER
definitions, respectively. A viewer class defines
a set of alternative views, each of which is associated with a name displayed
for the user to switch between visualisations, and a combination of parameters
LIST
,
LINE
or
SPIRAL
and
MULTISELECT
. The first three determine which of
the visualisations shown in Fig. 3 will be used, where
LIST
represents the verti-
cal alignment of tags,
LINE
a horizontal, line-based visualisation and
SPIRAL
the
advanced tag cloud visualisation. Again, other visualisations could be supported
by introducing new parameters that represent the respective visualisations. If the
optional parameter
MULTISELECT
is provided, then the associated visualisation
must allow for multiple selection of tags. With multiple selection, a combination
of conjunctive queries between and disjunctive queries within views can be sup-
ported. Finally, a
BROWSER
defines a set of viewer classes and also provides a
display name for each of them.
By using these augmentations of SQL, we can build on established database
concepts and directly benefit from the rich support for SQL expressions and func-
tions such as
COUNT
. Moreover, the caching strategies and high performance of
query execution in many database management systems, such as MySQL, makes
it optimal for web search interfaces. The way in which the final presentation of
the tag clouds is generated as well as how the synchronisation between views on
selection of a particular tag works are discussed in the next section.
5 Framework
Having described our language extension, we now present a framework that can
process such browser specifications and generate a browser interface to search
and browse specific data collections. The framework is shown in Fig. 6 in terms
of its main components and their interactions. The browser defined in terms of
an extended SQL specification is provided as input to the framework
1
.The
framework processes this specification as follows. First, a document template
representing the browser's web interface is generated
2
. This document con-
tains one designated placeholder for each viewer, in which the tag clouds will be
inserted at application runtime. Second, a browser-specific SQL view manager
is created
3
based on the viewer specifications that will create the SQL views
specified for each viewer in the database. At run-time, the SQL view manager
queries these views
4
to retrieve the tags and their sizes which then provide the
necessary input for the tag cloud generator that is called to create the tag clouds
and the associated visualisations
5
. In a final step, these generated tag clouds
are inserted into the placeholders of the document template
6
, which yields the
final browser interface presented to the user.
When a user selects a tag in one of the tag clouds, the current view associated
with that tag cloud as well as the associated views of all other tag clouds are
