Information Technology Reference
In-Depth Information
100
Sorted method
Unsorted method
80
60
40
20
0
20000
40000
60000
80000
100000
Threshold
F
IGURE
6.
Evaluation time(inmilliseconds) for bulk insertionmethods as
threshold is varied.
100
80
Unsorted method
60
40
Sorted method
20
1
10
100
1000
10000
100000
Threshold
F
IGURE
6.
Evaluation time(inmilliseconds) for bulk insertionmethods as
threshold is varied.
Fig. 11.4
Choice of axis scaling. For these graphs showing the same data, in the lower graph the
logarithmic scaling on the
x
-axis allows the behaviour for small thresholds to be seen
informative. Even if one algorithm is twice as fast as the other, a log-log graph
will show one line just a little below the other.
In some cases, data that seems innately tabular can be represented as a graph.
Often a bar graph is suitable because the items being compared are not ordered, as
shown in the graph in Fig.
11.5
. (Such data should not be represented by a line chart,
in which the points are connected into a continuous line, which would imply that
the axes were related by a function.) A richer example is shown in Fig.
11.6
.Forthe
more complex problem of comparing space and time simultaneously, the graph in
Fig.
11.7
would serve well.
