Next: 3.4 Discussion and Exercises Up: 3. Linked Lists Previous: 3.2 DLList: A Doubly-Linked Contents Index
One of the drawbacks of linked lists (besides the time it takes to access elements that are deep within the list) is their space usage. Each node in a DLList requires an additional two references to the next and previous nodes in the list. Two of the fields in a Node are dedicated to maintaining the list, and only one of the fields is for storing data!
An SEList (space-efficient list) reduces this wasted space using
a simple idea: Rather than store individual elements in a DLList,
we store a block (array) containing several items. More precisely, an
SEList is parameterized by a block size
. Each individual
node in an SEList stores a block that can hold up to
elements.
For reasons that will become clear later, it will be helpful if we can
do Deque operations on each block. The data structure that we choose
for this is a BDeque (bounded deque),
derived from the ArrayDeque
structure described in Section 2.4. The BDeque differs from the
ArrayDeque in one small way: When a new BDeque is created, the size
of the backing array
is fixed at
and never grows or shrinks.
The important property of a BDeque is that it allows for the addition
or removal of elements at either the front or back in constant time. This
will be useful as elements are shifted from one block to another.
An SEList is just a doubly-linked list of blocks. In addition to
and
pointers, each node
in an SEList contains a BDeque,
.
An SEList places very tight restrictions on the number of elements
in a block: Unless a block is the last block, then that block contains
at least
and at most
elements. This means that, if an
SEList contains
elements, then it has at most
but, for every block except the last, at most a constant amount of
space is wasted in this array. The remaining memory used by a block is
also constant. This means that the wasted space in an SEList is only
. By choosing a value of
within a constant factor
of
, we can make the space-overhead of an SEList approach
the
lower bound given in Section 2.6.2.
The first challenge we face with an SEList is finding the list item
with a given index
. Note that the location of an element consists
of two parts:
that contains the block that contains the element
with index
; and
of the element within its block.
To find the block that contains a particular element, we proceed the same way as we do in a DLList. We either start at the front of the list and traverse in the forward direction, or at the back of the list and traverse backwards until we reach the node we want. The only difference is that, each time we move from one node to the next, we skip over a whole block of elements.
Remember that, with the exception of at most one block, each block
contains at least
elements, so each step in our search gets
us
elements closer to the element we are looking for. If we
are searching forward, this means that we reach the node we want after
steps. If we search backwards, then we reach the node we
want after
steps. The algorithm takes the smaller
of these two quantities depending on the value of
, so the time to
locate the item with index
is
.
Once we know how to locate the item with index
, the
and
operations translate into getting or setting a particular
index in the correct block:
The running times of these operations are dominated by the time it takes
to locate the item, so they also run in
time.
Adding elements to an SEList is a little more complicated. Before
considering the general case, we consider the easier operation,
,
in which
is added to the end of the list. If the last block is full
(or does not exist because there are no blocks yet), then we first
allocate a new block and append it to the list of blocks. Now that
we are sure that the last block exists and is not full, we append
to the last block.
Things get more complicated when we add to the interior of the list
using
. We first locate
to get the node
whose block
contains the
th list item. The problem is that we want to insert
into
's block, but we have to be prepared for the case where
's block already contains
elements, so that it is full and
there is no room for
.
Let
denote
,
,
,
and so on. We explore
looking for a node
that can provide space for
. Three cases can occur during our
space exploration (see Figure 3.4):
|
steps) find a node
whose block
is not full. In this case, we perform 
shifts of an element from
one block into the next, so that the free space in
becomes a
free space in
. We can then insert
into
's block.
steps) run off the end of the list
of blocks. In this case, we add a new empty block to the end of the
list of blocks and proceed as in the first case.
steps we do not find any block that is not full.
In this case,
is a sequence of
blocks
that each contain
elements. We insert a new block
at the end of this sequence and spread the original
elements so that each block of
contains exactly
elements. Now
's block contains only
elements so it has
room for us to insert
.
The running time of the
operation depends on which of
the three cases above occurs. Cases 1 and 2 involve examining and
shifting elements through at most
blocks and take
time.
Case 3 involves calling the
method, which moves
elements and takes
time. If we ignore the cost of Case 3
(which we will account for later with amortization) this means that
the total running time to locate
and perform the insertion of
is
.
Removing an element from an SEList is similar to adding an element.
We first locate the node
that contains the element with index
.
Now, we have to be prepared for the case where we cannot remove an element
from
without causing
's block to become smaller than
.
Again, let
denote
,
,
,
and so on. We examine
in order to look
for a node from which we can borrow an element to make the size of
's block at least
. There are three cases to consider
(see Figure 3.5):
|
steps) find a node whose block contains
more than
elements. In this case, we perform 
shifts of an
element from one block into the previous one, so that the extra element
in
becomes an extra element in
. We can then remove the
appropriate element from
's block.
steps) run off the end of the list of
blocks. In this case,
is the last block, and there is no need
for
's block to contain at least
elements. Therefore,
we proceed as above, borrowing an element from
to make an extra
element in
. If this causes
's block to become empty,
then we remove it.
steps, we do not find any block containing more than
elements. In this case,
is a sequence
of
blocks that each contain
elements. We gather
these
elements into
so that each
of these
blocks contains exactly
elements and we remove
, which is now empty. Now
's block contains
elements and we can then remove the appropriate element from it.
Like the
operation, the running time of the
operation is
if we ignore the cost of
the
method that occurs in Case 3.
Next, we consider the cost of the
and
methods
that may be executed by the
and
methods. For the
sake of completeness, here they are:
The running time of each of these methods is dominated by the two
nested loops. Both the inner and outer loops execute at most
times, so the total running time of each of these methods
is
. However, the following lemma shows that
these methods execute on at most one out of every
calls to
or
.
calls
to
and
is performed, then the total time
spent during all calls to
and
is
.
is fragile if
's block does not contain
elements (so that
is either the last node, or contains
or
elements). Any node whose block contains
elements is
rugged. Define the potential of an SEList as the number
of fragile nodes it contains. We will consider only the
operation and its relation to the number of calls to
.
The analysis of
and
is identical.
Notice that, if Case 1 occurs during the
method, then
only one node,
has the size of its block changed. Therefore,
at most one node, namely
, goes from being rugged to being
fragile. If Case 2 occurs, then a new node is created, and this node
is fragile, but no other node changes size, so the number of fragile
nodes increases by one. Thus, in either Case 1 or Case 2 the potential
of the SEList increases by at most one.
Finally, if Case 3 occurs, it is because
are all fragile nodes. Then
is called and these
fragile nodes are replaced with
rugged nodes. Finally,
is added to
's block, making
fragile. In total the
potential decreases by
.
In summary, the potential starts at 0 (there are no nodes in the list).
Each time Case 1 or Case 2 occurs, the potential increases by at
most 1. Each time Case 3 occurs, the potential decreases by
.
The potential (which counts the number of fragile nodes) is never
less than 0. We conclude that, for every occurrence of Case 3, there
are at least
occurrences of Case 1 or Case 2. Thus, for every
call to
there are at least
calls to
. This
completes the proof.
The following theorem summarizes the performance of the SEList data structure:
and
, an SEList with block size
supports the operations
and
in
time per operation; and
and
in
time per operation.
and
operations results in a total of
time spent during all calls to
and
.
The space (measured in words)3.1 used by an SEList
that stores
elements is
.
The SEList is a trade-off between an ArrayList and a DLList where
the relative mix of these two structures depends on the block size
.
At the extreme
, each SEList node stores at most three values,
which is not much different than a DLList. At the other extreme,
, all the elements are stored in a single array, just like in
an ArrayList. In between these two extremes lies a trade-off between
the time it takes to add or remove a list item and the time it takes to
locate a particular list item.
![\includegraphics[width=\textwidth ]{figs-python/selist-add-a}](img1305.png)
![\includegraphics[width=\textwidth ]{figs-python/selist-add-b}](img1306.png)
![\includegraphics[width=\textwidth ]{figs-python/selist-add-c}](img1307.png)
![\includegraphics[scale=0.90909]{figs-python/selist-remove-a}](img1351.png)
![\includegraphics[scale=0.90909]{figs-python/selist-remove-b}](img1352.png)
![\includegraphics[scale=0.90909]{figs-python/selist-remove-c}](img1353.png)