Transcript mem9

Linked Data Structures:
Linked data structures - versatile data structures to model complex
real world situations and entities. List, stacks, trees, graphs etc. just
to mention some.
Consider a simple binary search tree for characters:
Analyze chapter9_1 program:
Average time complexity to search for an item is O(n log n)
Organization: Binary tree, left child has value less than the parent,
the right child has value greater than the parent.
Chapter 9, Slide 1
E
B
A
NULL NULL
K
D
NULL NULL
J
NULL
NULL NULL
Depth first traversal will produce alphabetically ordered sequence:
ABDEJK
Chapter 9, Slide 2
The nature of “links” -- usually pointers, but any reference will do:
Analyze chapter9_2 program:
The references are array indexes (arrays are modeled on memory!).
This will work fine for a tree with up to 100 nodes. A natural way to
serialize a binary tree.
Not necessarily are linked data structures created on the heap only:
Analyze chapter9_3 program that is a recursive descent parser for a
list of characters separated by commas and builds a binary search
tree on the stack.
Though, it is not very practical.
Most commonly, linked data structures are linked by pointers and are
build on the heap.
Chapter 9, Slide 3
Pointer-based linked data structures are “flexible”, which is mostly
good, however it is bad for “moving” the structure elsewhere in
memory, or “transmitting” it over a communication channel, or
“recording” it to auxiliary memory.
• compaction: we say that a linked data structure is compacted if it
occupies a contiguous segment of memory and all pointers (addresses)
are relative to the beginning of that segment.
• serialization: we say that a linked data structure is serialized if it
occupies several contiguous segments of memory and all pointers
(addresses) are relative to the beginning of that segment where the
pointer is stored.
Thus compaction is the extreme form of serialization. A serialized
structure can easily by “moved” in memory just by “moving” the
whole segment(s), “transmitted” byte by byte over a communication
channel, or “recorded” to auxiliary memory and later restored.
Chapter 9, Slide 4
Illustration of serialization+allocation from arena: chapter9_4
program.
The “relativized” addresses are a pair of short integers, the first is
segment+1 and the second is offset. Let us now visualize the arena
after each stage of the process.
First “building” the tree:
root
d
NULL
NULL
address
4398592
Chapter 9, Slide 5
root
d
NULL
c
NULL
NULL
address
4398592
root
address
4398592
d
e
c
NULL
NULL
NULL
NULL
address
4399604
Chapter 9, Slide 6
root
address
4398592
d
e
c
NULL
NULL
a
NULL
NULL
NULL
address
4399604
root
address
4398592
d
e
c
NULL
NULL
NULL
NULL
a
NULL
NULL
address
4399604
b
address
4399648
Chapter 9, Slide 7
The tree is build, now we start the relativization process:
root
(1,0)
d
(1,12)
(2,0)
c
(2,12)
(0,0)
(0,0)
(0,0)
a
(0,0)
(3,0)
(0,0)
(0,0)
address
4398592
e
address
4399604
b
address
4399648
Chapter 9, Slide 8
root
(1,0)
d
(1,12)
(2,0)
c
(2,12)
(0,0)
(0,0)
(0,0)
a
(0,0)
(3,0)
(0,0)
(0,0)
address
8904304
e
address
8905200
b
address
8905284
We deliberately designed the structure/class NODE so that it has
size of 12 bytes, but 3 bytes are wasted on padding:
a
NULL
NULL
Chapter 9, Slide 9
We can compact the nodes with no space wasted:
a
NULL
NULL
But then we cannot use p->lch or p->rch, we must have our
custom-made access functions: analyze chapter9_5 program.
croot
b NULL NULL a NULL
c
NULL e NULL NULL d
Chapter 9, Slide 10
After relativization:
croot
37
b
1
0
2
0
6
a
10 11
0
1
15
c
10
19 20
0
24
e
0
29
28
0
33
d
37 38
19
28
42
End of slides for chapter 9
Chapter 9, Slide 11