Download pptx

05 Array-Based Sequences
Hongfei Yan
Mar. 23, 2016
Contents
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
01 Python Primer (P2-51)
02 Object-Oriented Programming (P57-103)
03 Algorithm Analysis (P111-141)
04 Recursion (P150-180)
05 Array-Based Sequences (P184-224)
06 Stacks, Queues, and Deques (P229-250)
07 Linked Lists (P256-294)
08 Trees (P300-352)
09 Priority Queues (P363-395)
10 Maps, Hash Tables, and Skip Lists (P402-452)
11 Search Trees (P460-528)
12 Sorting and Selection (P537-574)
13 Text Processing (P582-613)
14 Graph Algorithms (P620-686)
15 Memory Management and B-Trees (P698-717)
05 Array-Based Sequences
5.1 Python’s Sequence Types
5.2 Low-Level Arrays
5.3 Dynamic Arrays and Amortization
5.4 Efficiency of Python’s Sequence Types
5.5 Using Array-Based Sequences
5.6 Multidimensional Data Sets
5.1 Python Sequence Classes
• Python has built-in types, list, tuple, and str.
• Each of these sequence types supports indexing to access an individual element of a
sequence, using a syntax such as A[i]
• Each of these types uses an array to represent the sequence.
• An array is a set of memory locations that can be addressed using consecutive indices, which, in
Python, start with index 0.
A
0 1 2
i
n
Public Behaviors
• A proper understanding of the outward semantics for a class is a
necessity for a good programmer.
• While the basic usage of lists, strings, and tuples may seem
straightforward, there are several important subtleties regarding the
behaviors associated with these classes
• (such as what it means to make a copy of a sequence, or to take a slice of a
sequence).
• Having a misunderstanding of a behavior can easily lead to inadvertent bugs
in a program.
• Therefore, we establish an accurate mental model for each of these classes.
• These images will help when exploring more advanced usage, such as
representing a multidimensional data set as a list of lists.
Implementation Details
• A focus on the internal implementations of these classes seems to go
against our stated principles of object-oriented programming. we
emphasized the principle of encapsulation, noting that the user of a
class need not know about the internal details of the implementation.
• While it is true that one only needs to understand the syntax and
semantics of a class’s public interface in order to be able to write legal
and correct code that uses instances of the class, the efficiency of a
program depends greatly on the efficiency of the components upon
which it relies.
Asymptotic and Experimental Analysis
• In describing the efficiency of various operations for Python’s
sequence classes, we will rely on the formal asymptotic analysis
notations.
• We will also perform experimental analyses of the primary operations
to provide empirical evidence that is consistent with the more
theoretical asymptotic analyses.
5.2 Low-Level Arrays
• To accurately describe the way in which Python represents the
sequence types, we must first discuss aspects of the low-level
computer architecture.
• The primary memory of a computer is composed of bits of
information, and those bits are typically grouped into larger units that
depend upon the precise system architecture. Such a typical unit is a
byte, which is equivalent to 8 bits.
• A computer system will have a huge number of bytes of memory, and
to keep track of what information is stored in what byte, the
computer uses an abstraction known as a memory address.
• Using the notation for asymptotic analysis, we say that any individual byte of memory
can be stored or retrieved in O(1) time.
• In general, a programming language keeps track of the association between an identifier
and the memory address in which the associated value is stored.
• For example, identifier x might be associated with one value stored in memory, while
y is associated with another value stored in memory.
• A group of related variables can be stored one after another in a contiguous portion of
the computer’s memory. We will denote such a representation as an array.
• We describe this as an array of six characters, even though it requires 12 bytes of
memory. We will refer to each location within an array as a cell, and will use an integer
index to describe its location within the array, with cells numbered starting with 0, 1, 2,
and so on.
• Each cell of an array must use the same number of bytes. This requirement is what allows
an arbitrary cell of the array to be accessed in constant time based on its index.
a programmer can envision a more typical high-level abstraction of an
array of characters as diagrammed in Figure 5.3.
5.2.1 Referential Arrays
• Python represents a list or tuple instance using an internal storage mechanism of an array of
object references. At the lowest level, what is stored is a consecutive sequence of memory
addresses at which the elements of the sequence reside. A high-level diagram of such a list is
shown.
• Although the relative size of the individual elements may vary, the number of bits used to store
the memory address of each element is fixed (e.g., 64-bits per address). In this way, Python can
support constant-time access to a list or tuple element based on its index.
12
The fact that lists and tuples are referential structures is significant to
the semantics of these classes. A single list instance may include
multiple references to the same object as elements of the list, and it is
possible for a single object to be an element of two or more lists, as
those lists simply store references back to that object.
5.2.2 Compact Arrays
• Primary support for compact arrays is in a module named array.
• That module defines a class, also named array, providing compact storage for
arrays of primitive data types.
• Compact arrays have several advantages over referential structures in
terms of computing performance.
• Most significantly, the overall memory usage will be much lower for a
compact structure because there is no overhead devoted to the explicit
storage of the sequence of memory references (in addition to the primary
data).
• Another important advantage to a compact structure for high-performance
computing is that the primary data are stored consecutively in memory.
Creating compact arrays of various types
• The constructor for the array class requires a type code as a first
parameter, which is a character that designates the type of data that
will be stored in the array.
Type Codes in the array Class
• Python’s array class has the following type codes:
5.2 Dynamic Arrays and
Amortization
5.3.1 Implementing a Dynamic Array
• Python list class provides a highly optimized implementation of
dynamic arrays, upon which we rely for the remainder of this book,
• it is instructive to see how such a class might be implemented.
• The key is to provide means to grow the array A that stores the
elements of a list.
• Of course, we cannot actually grow that array, as its capacity is fixed.
If an element is appended to a list at a time when the
underlying array is full, we perform the following steps:
Code Fragment 5.3: An implementation of a DynamicArray class,
using a raw array from the ctypes module as storage.
5.4 Efficiency of Python’s Sequence Types
Insertion
• In an operation add(i, o), we need to make room for the new element by
shifting forward the n - i elements A[i], …, A[n - 1]
• In the worst case (i = 0), this takes O(n) time
A
0 1 2
i
n
0 1 2
i
n
0 1 2
o
i
A
A
n
Element Removal
• In an operation remove(i), we need to fill the hole left by the removed
element by shifting backward the n - i - 1 elements A[i + 1], …, A[n - 1]
• In the worst case (i = 0), this takes O(n) time
A
0 1 2
o
i
n
0 1 2
i
n
0 1 2
i
A
A
n
Performance
• In an array based implementation of a dynamic list:
• The space used by the data structure is O(n)
• Indexing the element at I takes O(1) time
• add and remove run in O(n) time in worst case
• In an add operation, when the array is full, instead of throwing
an exception, we can replace the array with a larger one…
Growable Array-based Array List
• In an add(o) operation (without an
index), we could always add at the end
• When the array is full, we replace the
array with a larger one
• How large should the new array be?
Algorithm add(o)
if t = S.length - 1 then
A  new array of
size …
for i  0 to n-1 do
A[i]  S[i]
• Incremental strategy: increase the size by a
SA
constant c
nn+1
• Doubling strategy: double the size
S[n-1]  o
Comparison of the Strategies
• compare the incremental strategy and the doubling strategy
by analyzing the total time T(n) needed to perform a series of
n add(o) operations
• assume that we start with an empty stack represented by an
array of size 1
• call amortized time of an add operation the average time
taken by an add over the series of operations, i.e., T(n)/n
Incremental Strategy Analysis
• We replace the array k = n/c times
• The total time T(n) of a series of n add operations is proportional to
n + c + 2c + 3c + 4c + … + kc =
n + c(1 + 2 + 3 + … + k) =
n + ck(k + 1)/2
• Since c is a constant, T(n) is O(n + k2), i.e., O(n2)
• The amortized time of an add operation is O(n)
Doubling Strategy Analysis
• We replace the array k = log2 n times
• The total time T(n) of a series of n add operations
is proportional to
n + 1 + 2 + 4 + 8 + …+ 2k =
n + 2k + 1 - 1 =
3n - 1
• T(n) is O(n)
• The amortized time of an add operation is O(1)
geometric series
2
4
1
1
8
5.3.3 Python’s List Class
5.5 Using Array-Based Sequences
• 5.5.1 Storing High Scores for a Game
• 5.5.2 Sorting a Sequence
• 5.5.3 Simple Cryptography
5.5.1 Storing High Scores for a Game
5.5.1 Storing High
Scores for a Game
5.5.2 Sorting a Sequence
5.5.2 Sorting a Sequence
5.5.3 Simple Cryptography
• An interesting application of strings and lists is cryptography, the
science of secret messages and their applications. This field studies
ways of performing encryption, which takes a message, called the
plaintext, and converts it into a scrambled message, called the
ciphertext. Likewise, cryptography also studies corresponding ways of
performing decryption, which takes a ciphertext and turns it back into
its original plaintext.
• Arguably the earliest encryption scheme is the Caesar cipher, which is
named after Julius Caesar, who used this scheme to protect important
military messages.
• The Caesar cipher involves replacing each letter in a message with the letter
that is a certain number of letters after it in the alphabet.
5.6 Multidimensional Data Sets
• For example, the two-dimensional data
• might be stored in Python as follows.
data = [ [22, 18, 709, 5, 33], [45, 32, 830, 120, 750], [4, 880, 45, 66, 61] ]
Constructing a Multidimensional List
• initialize a one-dimensional list, rely on a syntax such as
data = [0] * n to create a list of n zeros.
• list comprehension syntax for a 2D list of integers
data = [ [0] c for j in range(r) ]
Tic-Tac-Toe
Summary