OPTIMIZING THE JAVA VIRTUAL MACHINE INSTRUCTION SET BY ...

204
Algorithm SplitLeaves(node n)
For each branch, b, leaving n
If the node, m, at the end of b is a leaf and the label on b starts with
a character in Sigma
Create a new node, p
Create a branch, c, from p to m
Change the target of branch b to node p
Find the position, x, of the first sentinel character in the label on b
using a binary search on the list of sentinel character positions
Label c with all characters from x to the end of the string
Change the label on b so that it only includes those characters
before x
Else
SplitLeaves(b.Target)
End For
Figure 9.2: Algorithm SplitLeaves
score can be recorded in the node and utilized later or a single variable can be used
to record the best string and its score encountered so far.
Examining Figure 9.4 reveals that the most contributory substring is abab. This
is determined by identifying the string that is formed by traversing the tree from the
root to the node that has the highest score. For this example, a score of 12 is the
highest value present in the tree. The occurrences of the most contributory string,
abab are illustrated below.
ababab$abab#
9.5 Handling Overlapping Occurrences
It is important to observe that the single underlined and single over-lined occurrences
of abab are overlapping. While this may not be a problem for some applications,
overlapping occurrences of the string are of little use for others. In particular, any
application that “uses up” the string once it is identified will destroy the subsequent
overlapping occurrence(s). Multicode substitution is an example of an instance where
the identified string is used up. Consequently it is highly desirable to overcome