Computational Mathematics with Python

Basics

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 1 / 307
Introduction and Motivation Concepts Practical Information

1 Introduction and Motivation

Python vs Other Languages
Examples and Demo

2 Concepts
Basic Types
Variables
Lists
For Loop

3 Practical Information

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 2 / 307
Introduction and Motivation Concepts Practical Information

1 Introduction and Motivation

Python vs Other Languages
Examples and Demo

2 Concepts
Basic Types
Variables
Lists
For Loop

3 Practical Information

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 3 / 307
Introduction and Motivation Concepts Practical Information

Why Python?

Python is. . .
I Free and open source
I It is a scripting language, meaning that it is interpreted
I It is modern: object oriented, exception handling, dynamic typing etc.
I Plenty of libraries, in particular scientific ones: linear algebra;
visualisation tools: plotting, image analysis; differential equations solving;
symbolic computations; statistics ; etc.
I Many possible usages: Scientific computing (of course :-)), scripting, web
sites, text parsing, etc.
I Used by YouTube, Google, NASA, Los Alamos, NSA among others

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 4 / 307
Introduction and Motivation Concepts Practical Information

Python vs language XX

Java, C++ Object oriented compiled languages. Very limited and

extremely verbose. Low level compared to python. Few
scientific libraries.
C, FORTRAN Very low level compiled language. Useful in some CPU
critical situations.
php, ruby Other interpreted languages. PHP is web oriented. Ruby is
as flexible as python but has no scientific library.
MATLAB Tool for matrix computation that evolved for scientific
computing. The scientific library is huge but it is not a
programming language. Extremely expensive.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 5 / 307
Introduction and Motivation Concepts Practical Information

Examples
Python may be used in interactive mode:
>>> x = 3
>>> y = 5
>>> print x + y
8

Here we solve " # " #

1 2 2
·x =
3 4 1
>>> M = array ( [ [ 1 . , 2 . ] ,
[3., 4.]])
>>> V = array ( [ 2 . , 1 . ] )
>>> x = solve (M , V )
>>> print x
[-3. 2.5]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 6 / 307
Introduction and Motivation Concepts Practical Information

More examples

Computing eiπ and 2100 :

>>> print exp ( 1j * pi ) # should return -1 : -)
( - 1 + 1 . 22464679915e - 16j )
>>> print 2 * * 100
1267650600228229401496703205376L
P∞ 1 π2
Computing ζ(x) = k=1 k x . For x = 2 we know that ζ(2) = 6 :
# for x = 2 :
>>> print scipy . special . zeta ( 2 . , 1 )
1 . 64493406685
>>> print pi * * 2 / 6
1 . 6449340668482264

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 7 / 307
Introduction and Motivation Concepts Practical Information

Demo

Demo

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 8 / 307
Introduction and Motivation Concepts Practical Information

1 Introduction and Motivation

Python vs Other Languages
Examples and Demo

2 Concepts
Basic Types
Variables
Lists
For Loop

3 Practical Information

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 9 / 307
Introduction and Motivation Concepts Practical Information

Numbers

A number may be an integer, a real number or a complex number. The

usual operations are
I + and - addition and substraction
I * and / multiplication and division
I ** power
2 * * ( 2 + 2 ) # 16
1j * * 2 # -1

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 10 / 307
Introduction and Motivation Concepts Practical Information

Strings

Strings are “lists” of characters, enclosed by simple or double quotes:

’ valid string ’
" string with double quotes "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 11 / 307
Introduction and Motivation Concepts Practical Information

Strings

Strings are “lists” of characters, enclosed by simple or double quotes:

’ valid string ’
" string with double quotes "

You may also use triple quotes for strings including multiple lines:
""" This is
a long ,
long string """

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 11 / 307
Introduction and Motivation Concepts Practical Information

Concept: Variable

Variables
A variable is a reference to an object. An object may have several
references. One uses the assignment operator = to assign a value to a
variable.

Example
x = [3 , 4 ] # a list object is created
y = x # this object now has two labels : x and y
del x # we delete one of the labels
del y # both labels are removed : the object is deleted

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 12 / 307
Introduction and Motivation Concepts Practical Information

Concept: Lists

Lists
A python list is an ordered list of objects, enclosed in square brackets. One
accesses elements of a list using zero-based indices inside square brackets.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 13 / 307
Introduction and Motivation Concepts Practical Information

List Examples

Example
L1 = [1 , 2 ]
L1 [ 0 ] # 1
L1 [ 1 ] # 2
L1 [ 2 ] # raises IndexError

L2 = [ ’a ’ , 1 , [3 , 4 ] ]
L2 [ 0 ] # ’a ’
L2 [ 2 ] [ 0 ] # 3

L2 [ - 1 ] # last element : [3 , 4 ]
L2 [ - 2 ] # second to last : 1

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 14 / 307
Introduction and Motivation Concepts Practical Information

List Utilities
I range(n) creates a list with n elements, starting with zero:
print range ( 5 )
[0 , 1 , 2 , 3 , 4 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 15 / 307
Introduction and Motivation Concepts Practical Information

List Utilities
I range(n) creates a list with n elements, starting with zero:
print range ( 5 )
[0 , 1 , 2 , 3 , 4 ]

I len(L) gives the length of a list:

len ( [ ’a ’ , 1 , 2 , 34 ] ) # returns 4

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 15 / 307
Introduction and Motivation Concepts Practical Information

List Utilities
I range(n) creates a list with n elements, starting with zero:
print range ( 5 )
[0 , 1 , 2 , 3 , 4 ]

I len(L) gives the length of a list:

len ( [ ’a ’ , 1 , 2 , 34 ] ) # returns 4

I Use append to append an element to a list:

L = [ ’a ’ , ’b ’ , ’c ’]
L [ - 1 ] # ’c ’
L . append ( ’d ’)
L # L is now [ ’ a ’, ’b ’, ’c ’, ’d ’]
L [ - 1 ] # ’d ’

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 15 / 307
Introduction and Motivation Concepts Practical Information

Comprehensive lists
A convenient way to build up lists is to use the comprehensive lists
construct, possibly with a conditional inside.
Definition
The syntax of a comprehensive list is
[ < expr > for < x > in < list > ]

Example
L = [2 , 3 , 10 , 1 , 5 ]

L2 = [ x * 2 for x in L ] # [4 , 6 , 20 , 2 , 10 ]

L3 = [ x * 2 for x in L if 4 < x < = 10 ] # [ 20 , 10 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 16 / 307
Introduction and Motivation Concepts Practical Information

Comprehensive Lists in Maths

Mathematical Notation
This is very close to the mathematical notation for sets. Compare:

L2 = {2x; x ∈ L}

and
L2 = [ 2 * x for x in L ]

One big difference though is that lists are ordered while sets aren’t.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 17 / 307
Introduction and Motivation Concepts Practical Information

Operations on Lists

I Adding two lists concatenates (sammanfoga) them:

L1 = [1 , 2 ]
L2 = [3 , 4 ]
L = L1 + L2 # [1 , 2 , 3 , 4 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 18 / 307
Introduction and Motivation Concepts Practical Information

Operations on Lists

I Adding two lists concatenates (sammanfoga) them:

L1 = [1 , 2 ]
L2 = [3 , 4 ]
L = L1 + L2 # [1 , 2 , 3 , 4 ]

I Logically, multiplying a list with an integer concatenates the list with

itself several times: n*L is equivalent to L
|
+L+ {z
· · · + L}.
n times
L = [1 , 2 ]
3 * L # [1 , 2 , 1 , 2 , 1 , 2 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 18 / 307
Introduction and Motivation Concepts Practical Information

Concept: for loop

for loop
A for loop allows to loop through a list using an index variable. This
variable is successively equal to all the elements in the list.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 19 / 307
Introduction and Motivation Concepts Practical Information

Concept: for loop

for loop
A for loop allows to loop through a list using an index variable. This
variable is successively equal to all the elements in the list.

Example
L = [1 , 2 , 10 ]
for s in L :
print s * 2 ,
# output : 2 4 20

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 19 / 307
Introduction and Motivation Concepts Practical Information

Indentation

The part to be repeated in the for loop has to be properly indented:

for elt in my_list :
do_something ()
something_else ()
etc
print " loop finished " # outside the for block

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 20 / 307
Introduction and Motivation Concepts Practical Information

Repeating a Task

One typical use of the for loop is to repeat a certain task a fixed number
of time:
n = 30
for i in range ( n ) :
do_something # this gets executed n times

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 21 / 307
Introduction and Motivation Concepts Practical Information

1 Introduction and Motivation

Python vs Other Languages
Examples and Demo

2 Concepts
Basic Types
Variables
Lists
For Loop

3 Practical Information

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 22 / 307
Introduction and Motivation Concepts Practical Information

Python Shell

I Start a python session by typing scipython in a unix shell

I Check that it is working with: plot(rand(4));show()
I A window should appear with a graph; you should be able to type
other commands without having to close the graph window
I when you want to quit, write exit()
When you want to run python at home please follow the installation instruction
on http://www.maths.lth.se/na/python/install

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 23 / 307
Introduction and Motivation Concepts Practical Information

Executing Scripts

You often want to execute the contents of a file.

I We recommand to use Kate on the Linux machines (but any other
good editor will do)
I Save your files in (for example) in $HOME/course/
I Type (once) in ipython: cd course
I To execute the contents of a file named file.py just write
execfile(’file.py’) in ipython.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 24 / 307
Introduction and Motivation Concepts Practical Information

Getting Help

Some tips on how to use ipython:

I To get help on an object just type ? after it and then return
I Use the arrow keys to reuse the last executed commands
I We will see later that you may use the tabulation key for completion
in general

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 25 / 307
Introduction and Motivation Concepts Practical Information

Computational Mathematics with Python

Booleans

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 26 / 307
If Statement For Loop String Formatting Functions

4 If Statement

5 For Loop

6 String Formatting

7 Functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 27 / 307
If Statement For Loop String Formatting Functions

4 If Statement

5 For Loop

6 String Formatting

7 Functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 28 / 307
If Statement For Loop String Formatting Functions

Conditional Expressions

Definition
A conditional expression is an expression that may have the value True or
False.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 29 / 307
If Statement For Loop String Formatting Functions

Conditional Expressions

Definition
A conditional expression is an expression that may have the value True or
False.
Some common operators that yield conditional expressions are:
I ==, !=
I <, >, <=, >=
I One combines different boolean values with or and and
I not gives the logical negation of the expression that follows

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 29 / 307
If Statement For Loop String Formatting Functions

Boolean Expression Examples

Example
2 > = 4 # False
2 < 3 < 4 # True
2 < 3 and 3 < 2 # False
2 ! = 3 < 4 or False # True
2 < = 2 and 2 > = 2 # True
not 2 = = 3 # True
not False or True and False # True !

Note in the last example the rules when using not, and, or.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 30 / 307
If Statement For Loop String Formatting Functions

Concept: If statement
Concept: conditional statement
A conditional statement delimits a block that will be executed if the
condition is true. An optional block, started with the keyword else will be
executed if the condition is not fulfilled.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 31 / 307
If Statement For Loop String Formatting Functions

Concept: If statement
Concept: conditional statement
A conditional statement delimits a block that will be executed if the
condition is true. An optional block, started with the keyword else will be
executed if the condition is not fulfilled.

Example
We print the absolute value of x. Mathematically this is defined as x if
x ≥ 0 and −x if x < 0:

x = ... (
if x > = 0 : x if x≥0
print x |x| =
else :
−x else
print - x
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 31 / 307
If Statement For Loop String Formatting Functions

4 If Statement

5 For Loop

6 String Formatting

7 Functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 32 / 307
If Statement For Loop String Formatting Functions

The full for statement: break

break gets out of the for loop even if the list we are iterating is not
exhausted.
for x in x_values :
if x > threshold :
break
print x

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 33 / 307
If Statement For Loop String Formatting Functions

The full for statement: else

else checks whether the for loop was broken with the break keyword.
for x in x_values :
if x > threshold :
break
else :
print " all the x are below the threshold "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 34 / 307
If Statement For Loop String Formatting Functions

4 If Statement

5 For Loop

6 String Formatting

7 Functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 35 / 307
If Statement For Loop String Formatting Functions

Basic string formatting

I for strings:
course_code = " NUMA21 "
print " This course ’s name is % s " % course_code
# This course ’s name is NUMA21

I for integers:
nb_students = 16
print " There are % d students " % nb_students
# There are 16 students

I for reals:
average_grade = 3 . 4
print " Average grade : % f " % average_grade
# Average grade : 3 . 400000

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 36 / 307
If Statement For Loop String Formatting Functions

4 If Statement

5 For Loop

6 String Formatting

7 Functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 37 / 307
If Statement For Loop String Formatting Functions

Basics on Functions
Functions are useful to gather similar pieces of code at one place. Consider
the following mathematical function:

x 7→ f (x) := 2x + 1

The python equivalent is:

def f ( x ) :
return 2 * x + 1

I the keyword def tells python we are defining a function

I f is the name of the function
I x is the argument, or input of the function
I what is after return is called the output of the function
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 38 / 307
If Statement For Loop String Formatting Functions

enthought ®
Anatomy of a Function
Anatomy of a function

The keyword def Function arguments are listed

indicates the start separated by commas. They are passed
of a function. by assignment. More on this later.

def add(arg0, arg1):

Indentation is
used to indicate
a = arg0 + arg1
the contents of return a A colon ( : ) terminates
the function. It the function definition.
is not optional,
but a part of the
syntax. An optional return statement specifies
the value returned from the function. If
return is omitted, the function returns the
special value None.

(Source: E. Jones and T. Oliphant)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 39 / 307
If Statement For Loop String Formatting Functions

Calling a Function

Once the following function is defined:

def f ( x ) :
return 2 * x + 1

it may now be called using:

f(2) # 5
f(1) # 3
# etc .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 40 / 307
If Statement For Loop String Formatting Functions

Concluding example

Screenshot from Python editor drpython

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 41 / 307
If Statement For Loop String Formatting Functions

Computational Mathematics with Python

Slicing

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 42 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 43 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 44 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Creating sublists
Slicing
Slicing a list between i and j is creating a copy of its element starting
from element at index i and ending just before j.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 45 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Creating sublists
Slicing
Slicing a list between i and j is creating a copy of its element starting
from element at index i and ending just before j.

One simple way to understand slicing

L[i:j] means: create a list by removing the first i elements and
containing the next j − i elements (for j > i ≥ 0).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 45 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Creating sublists
Slicing
Slicing a list between i and j is creating a copy of its element starting
from element at index i and ending just before j.

One simple way to understand slicing

L[i:j] means: create a list by removing the first i elements and
containing the next j − i elements (for j > i ≥ 0).

Example
L = [ ’C ’ , ’l ’ , ’a ’ , ’u ’ , ’s ’]
L [ 1 : 4 ] # remove one element and take three from there :
# [ ’ l ’, ’a ’, ’u ’]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 45 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Partial slicing

One may omit the first or last bound of the slicing:

L = [ ’C ’ , ’l ’ , ’a ’ , ’u ’ , ’s ’]
L [ 1 : ] # [ ’ l ’, ’a ’, ’u ’, ’s ’]
L [ : 3 ] # [ ’ C ’, ’l ’, ’a ’]
L [ - 2 : ] # [ ’ u ’, ’s ’]
L [ : - 2 ] # [ ’ C ’, ’l ’, ’a ’]
L [ : ] # the whole list

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 46 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Partial slicing

One may omit the first or last bound of the slicing:

L = [ ’C ’ , ’l ’ , ’a ’ , ’u ’ , ’s ’]
L [ 1 : ] # [ ’ l ’, ’a ’, ’u ’, ’s ’]
L [ : 3 ] # [ ’ C ’, ’l ’, ’a ’]
L [ - 2 : ] # [ ’ u ’, ’s ’]
L [ : - 2 ] # [ ’ C ’, ’l ’, ’a ’]
L [ : ] # the whole list

Mathematical Analogy
This is similar to half lines in R. [−∞, a) means: take all numbers strictly
lower than a; this is similar to the syntax L[:j].

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 46 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[2:]
Rule of thumb
Taking all elements from index 2 included amounts to remove the
first two elements

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 47 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[2:]
Rule of thumb
Taking all elements from index 2 included amounts to remove the
first two elements

0 1 2 3 ··· -3 -2 -1
L[:2]
Rule of thumb
Taking all elements until index 2 excluded amounts to keep only the
first two elements

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 47 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[:-2]
Rule of thumb
Taking all elements until index -2 excluded amounts to remove the
last two elements

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 48 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[:-2]
Rule of thumb
Taking all elements until index -2 excluded amounts to remove the
last two elements

0 1 2 3 ··· -3 -2 -1
L[-2:]
Rule of thumb
Taking all elements from index -2 included amounts to keep only the
last two elements

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 48 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[2:-1]
Rule of thumb
L[i:-j] amounts to remove the first i elements and remove the last
j elements.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 49 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 2 3 ··· -3 -2 -1
L[2:-1]
Rule of thumb
L[i:-j] amounts to remove the first i elements and remove the last
j elements.

0 1 2 3 4 5 ··· -1
L[2:5]
Rule of thumb
L[i:j] amounts to remove the first i and keep the j − i next ones.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 49 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

0 1 ··· -5 -4 -3 -2 -1
L[-4:-1]
Rule of thumb
L[-i:-j] amounts to remove the last j and keep the i − j preceding
ones.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 50 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 51 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Tuples

Definition
A tuple is an immutable list. Immutable means that it cannot be modified.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 52 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Tuples

Definition
A tuple is an immutable list. Immutable means that it cannot be modified.

Example
my_tuple = 1 , 2 , 3 # our first tuple !
len ( my_tuple ) # 3 , same as for lists

my_tuple [ 0 ] = ’a ’ # error ! tuples are immutable

singleton = 1 , # note the comma

len ( singleton ) # 1

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 52 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Packing and unpacking

One may assign several variables at once by unpacking a list or tuple:

a, b = 0, 1 # a gets 0 and b gets 1
a , b = [0 , 1 ] # exactly the same effect
(a , b ) = 0 , 1 # same
[a , b ] = [0 , 1 ] # same thing

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 53 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Packing and unpacking

One may assign several variables at once by unpacking a list or tuple:

a, b = 0, 1 # a gets 0 and b gets 1
a , b = [0 , 1 ] # exactly the same effect
(a , b ) = 0 , 1 # same
[a , b ] = [0 , 1 ] # same thing

The swapping trick!

Use packing and unpacking to swap the contents of two variables.
a, b = b, a

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 53 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Formatting Strings

Format with several arguments. Use of tuple as a second argument is

mandatory.
print " My name is % s % s " % ( ’ Olivier ’ , ’ Verdier ’)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 54 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Returning Multiple Values

A function may return several values:

def argmin ( L ) : # return the minimum and index
...
return minimum , minimum_index

m , i = argmin ( [1 , 2 , 0 ] ) # m is 0 , i is 2
# or :
min_info = argmin ( [1 , 2 , 0 ] )
min_info [ 0 ] # 0
min_info [ 1 ] # 2

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 55 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

A final word on tuples

I Tuples are nothing else than immutable lists with a notation without
brackets

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 56 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

A final word on tuples

I Tuples are nothing else than immutable lists with a notation without
brackets
I In most cases lists may be used instead of tuples

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 56 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

A final word on tuples

I Tuples are nothing else than immutable lists with a notation without
brackets
I In most cases lists may be used instead of tuples
I The bracket free notation is nice but dangerous, you should use
parenthesis when you are not sure:
a , b = b , a # the swap tricks ; equivalent to :
(a , b ) = (b , a )
# but
1 , 2 = = 3 , 4 # returns (1 , False , 4 )
(1 , 2 ) = = (3 , 4 ) # returns False

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 56 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 57 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Convenient Notations

I Multiple assignment
a = b = c = 1 # a , b and c get the same value 1

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 58 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Convenient Notations

I Multiple assignment
a = b = c = 1 # a , b and c get the same value 1

I Increment operators
a + = 1 # same as a = a + 1
b * = 1 # same as b = b * 1

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 58 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Convenient Notations

I Multiple assignment
a = b = c = 1 # a , b and c get the same value 1

I Increment operators
a + = 1 # same as a = a + 1
b * = 1 # same as b = b * 1

I Multiple comparisons
a < b < c # same as : a < b and b < c
a = = b = = c # same as : a == b and b == c

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 58 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Function Documentation

Document your function using a string at the beginning:

def newton (f , x0 ) :
""" Compute a zero of ‘f ‘ with the Newton method ;
x0 is the initial guess """
...

Try this out and check the IPython help for that function!

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 59 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Functions are objects

Functions are objects like everything else. One may pass functions around
as arguments, change their names and delete them.
def square ( x ) :
""" Return the square of ‘x ‘ """
return x * * 2
square ( 4 ) # 16
sq = square # now sq is the same as square
sq ( 4 ) # 16
del square # ‘ square ‘ doesn ’t exist anymore
print newton ( sq , . 2 ) # passing as argument

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 60 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 61 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Named Arguments

Definition
The argument of a function may be accessed in any order, provided they
are named.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 62 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Example

Assume a function newton is defined as:

def newton (f , x0 , tol ) :
...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 63 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Example

Assume a function newton is defined as:

def newton (f , x0 , tol ) :
...

Now we can call it in various ways:

# note the order of the arguments in this call :
newton ( tol = 1e -3 , x0 = .1 , f = cos )

# here the first argument is not named

newton ( cos , tol = 1e -3 , x0 = . 1 ) # ok
newton ( cos , .1 , tol = 1e - 3 ) # ok

# but you can ’t have anonymous args after named ones

newton ( f = cos , .1 , tol = 1e - 3 ) # error !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 63 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Default Value

Definition
An argument may be given a default value with the equal sign.

Example
def newton (f , x0 , tol = 1e - 6 ) :
...
# the tolerance is set to the default 1e - 6
newton ( cos , . 2 )
newton ( x0 = .2 , f = cos ) # the same as above

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 64 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

8 Slicing
Definitions
Examples

9 Tuples
Definition
Usage

10 Convenient Syntax
Multiple Assignments and Comparisons
Tuples and Functions

11 Function arguments

12 Plotting

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 65 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Plot

I You have already used the command plot. It needs a list of x values
and a list of y values. If a single list is given, the list of x values
range(len(y)) is assumed.
I You may use the keyword argument label to give your curves a
name, and then show them using legend.
x = .2
x1 = [ sin (. 3 * n * x ) for n in range ( 20 ) ]
x2 = [ sin ( 2 * n * x ) for n in range ( 20 ) ]
plot ( x1 , label = ’0 . 3 ’)
plot ( x2 , label = ’2 ’)
legend ()

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 66 / 307
Slicing Tuples Convenient Syntax Function arguments Plotting

Computational Mathematics with Python

Exceptions, Generators

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 67 / 307
Exceptions Generators Slices Freezing Parameters

13 Exceptions

14 Generators

15 More on slices

16 Freezing Parameters

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 68 / 307
Exceptions Generators Slices Freezing Parameters

13 Exceptions

14 Generators

15 More on slices

16 Freezing Parameters

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 69 / 307
Exceptions Generators Slices Freezing Parameters

Sound the alarm

Creating errors
Creating an error is called “raise an exception”. You may raise an
exception like this:
raise Exception ( " Something went wrong " )

Typical exceptions are

I TypeError
I ValueError
You already know SyntaxError and ZeroDivisionError.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 70 / 307
Exceptions Generators Slices Freezing Parameters

Review the alarms

Reviewing the errors

You may review the errors using try and except:
try :
< some code that might raise an exception >
except ValueError :
print " Oops , a ValueError occurred "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 71 / 307
Exceptions Generators Slices Freezing Parameters

Review the alarms

Reviewing the errors

You may review the errors using try and except:
try :
< some code that might raise an exception >
except ValueError :
print " Oops , a ValueError occurred "

Flow Control
An exception stops the flow and looks for the closest enclosing try block.
If it is not caught it continues searching for the next try block.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 71 / 307
Exceptions Generators Slices Freezing Parameters

Error messages

Golden rule
Never print error messages, raise an exception instead

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 72 / 307
Exceptions Generators Slices Freezing Parameters

Error messages

Golden rule
Never print error messages, raise an exception instead

def factorial ( n ) :
if n < 0 :
raise ValueError ( " A positive integer is expected " )
...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 72 / 307
Exceptions Generators Slices Freezing Parameters

13 Exceptions

14 Generators

15 More on slices

16 Freezing Parameters

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 73 / 307
Exceptions Generators Slices Freezing Parameters

Definition
Definition
A generator generates objects (to be passed to a for loop). Similar to a list
except that the objects need not exist before entering the loop.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 74 / 307
Exceptions Generators Slices Freezing Parameters

Definition
Definition
A generator generates objects (to be passed to a for loop). Similar to a list
except that the objects need not exist before entering the loop.

Example
A typical generator is xrange: works like range but produces a generator
instead of a sequence.
for i in xrange ( 100000000 ) :
if i > 10 :
break

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 74 / 307
Exceptions Generators Slices Freezing Parameters

Definition
Definition
A generator generates objects (to be passed to a for loop). Similar to a list
except that the objects need not exist before entering the loop.

Example
A typical generator is xrange: works like range but produces a generator
instead of a sequence.
for i in xrange ( 100000000 ) :
if i > 10 :
break

Infinite
Note that, just as in mathematics, generators may be infinite.
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 74 / 307
Exceptions Generators Slices Freezing Parameters

Create Generators

Creation of generators is possible with the keyword yield:

def odd_numbers ( n ) :
" generator for odd numbers less than n "
for k in xrange ( n ) :
if k % 2 :
yield k

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 75 / 307
Exceptions Generators Slices Freezing Parameters

Create Generators

Creation of generators is possible with the keyword yield:

def odd_numbers ( n ) :
" generator for odd numbers less than n "
for k in xrange ( n ) :
if k % 2 :
yield k

Then you call it as:

g = odd_numbers ( 10 )
for k in g :
... # do something with k

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 75 / 307
Exceptions Generators Slices Freezing Parameters

Generator Tools

I enumerate is used to enumerate another generator:

A = [ ’a ’ , ’b ’ , ’c ’]
for i , x in enumerate ( A ) :
print i , x ,
# result : 0 a 1 b 2 c

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 76 / 307
Exceptions Generators Slices Freezing Parameters

Generator Tools

I enumerate is used to enumerate another generator:

A = [ ’a ’ , ’b ’ , ’c ’]
for i , x in enumerate ( A ) :
print i , x ,
# result : 0 a 1 b 2 c

I reversed creates a generator from a list by going backwards:

A = [0 , 1 , 2 ]
for elt in reversed ( A ) :
print elt ,
# result : 2 1 0

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 76 / 307
Exceptions Generators Slices Freezing Parameters

List Filling Pattern

Common programming pattern:

L = []
for k in xrange ( n ) :
L . append ( some_function ( k ))

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 77 / 307
Exceptions Generators Slices Freezing Parameters

List Filling Pattern

Common programming pattern:

L = []
for k in xrange ( n ) :
L . append ( some_function ( k ))

use instead:
L = [ function ( k ) for k in xrange ( n ) ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 77 / 307
Exceptions Generators Slices Freezing Parameters

Complicated List Filling

L = [0 , 1 ]
for k in range ( n ) :
# call various functions here
# that compute " result "
L . append ( result )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 78 / 307
Exceptions Generators Slices Freezing Parameters

Complicated List Filling

L = [0 , 1 ]
for k in range ( n ) :
# call various functions here
# that compute " result "
L . append ( result )

Use a generator instead:

def result_generator ( n ) :
for k in xrange ( n ) :
# call various functions here
# that compute " result "
yield result

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 78 / 307
Exceptions Generators Slices Freezing Parameters

Complicated List Filling

L = [0 , 1 ]
for k in range ( n ) :
# call various functions here
# that compute " result "
L . append ( result )

Use a generator instead:

def result_generator ( n ) :
for k in xrange ( n ) :
# call various functions here
# that compute " result "
yield result

...and if you really need a list:

L = list ( result_generator ( n )) # no append needed !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 78 / 307
Exceptions Generators Slices Freezing Parameters

List from a generator

To convert a generator to a list:

# for example :
g = xrange ( 10 )

L = list ( g )
# now L is a list with 10 elements

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 79 / 307
Exceptions Generators Slices Freezing Parameters

“Comprehensive” generator

Just as we had comprehensive lists, there is also comprehensive generators:

g = ( n for n in xrange ( 1000 ) if not n % 100 )
# generates 100 , 200 ,

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 80 / 307
Exceptions Generators Slices Freezing Parameters

Zipping generators

How to make one generator out of two?

xg = x_generator ()
yg = y_generator ()

for x , y in zip ( xg , yg ) :
print x , y

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 81 / 307
Exceptions Generators Slices Freezing Parameters

Zipping generators

How to make one generator out of two?

xg = x_generator ()
yg = y_generator ()

for x , y in zip ( xg , yg ) :
print x , y

The zipped generator stops as soon as one of the generators is exhausted.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 81 / 307
Exceptions Generators Slices Freezing Parameters

13 Exceptions

14 Generators

15 More on slices

16 Freezing Parameters

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 82 / 307
Exceptions Generators Slices Freezing Parameters

Out-of-bound slices

No errors for out-of-bound slices

Take notice that you never get index errors with out of bound slices, only,
possibly, empty lists.

Example
L = range ( 4 ) # [0 , 1 , 2 , 3 ]
L [ 4 ] # error !
L [ 1 : 100 ] # same as L [ 1 :]
L [ - 100 : - 1 ] # same as L [: - 1 ]
L [ - 100 : 100 ] # same as L [:]
L [ 5 : 0 ] # empty
L [ - 2 : 2 ] # empty WHY ?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 83 / 307
Exceptions Generators Slices Freezing Parameters

Altering lists

Slicing may be used to change lists.

L = [0 , 1 , 2 , 3 , 4 ]
L [ 0 ] = ’a ’ # standard access via index

I Replacement or Deletion
L [ 2 : 3 ] = [ 100 , 200 ] # [ ’ a ’, 1 , 100 , 200 , 3 , 4 ]
L [ 2 : 3 ] = [ ] # [ ’ a ’, 1 , 200 , 3 , 4 ]
L [ 3 : ] = [ ] # [ ’ a ’, 1 , 200 ]

I Insertion
L [ 1 : 1 ] = [ 1000 , 2000 ] # [ ’ a ’, 1000 , 2000 , 1 , 100 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 84 / 307
Exceptions Generators Slices Freezing Parameters

Strides

Definition
When computing slices one may also specify a stride which is the length of
the step from one index to the other. The default stride is one.

Example
L = range ( 100 )
L [ : 10 : 2 ] # [0 , 2 , 4 , 6 , 8 ]
L [ : : 20 ] # [0 , 20 , 40 , 60 , 80 ]
L [ 10 : 20 : 3 ] # [ 10 , 13 , 16 , 19 ]

Note that the stride may also be negative.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 85 / 307
Exceptions Generators Slices Freezing Parameters

Belonging to a list

Definition
One may use the keywords in and not in to determine whether an
element belongs to a list (similar to ∈ and ∈
/ in mathematics).

Example
L = [ ’a ’ , 1 , ’b ’ , 2 ]
’a ’ in L # True
3 in L # False
4 not in L # True

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 86 / 307
Exceptions Generators Slices Freezing Parameters

13 Exceptions

14 Generators

15 More on slices

16 Freezing Parameters

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 87 / 307
Exceptions Generators Slices Freezing Parameters

Mathematical View
In mathematics one has the following notation, for two sets A and C :

C A := {functions from A to C }

One remarkable fact is that

C A×B ≡ (C A )B ≡ (C B )A

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 88 / 307
Exceptions Generators Slices Freezing Parameters

Mathematical View
In mathematics one has the following notation, for two sets A and C :

C A := {functions from A to C }

One remarkable fact is that

C A×B ≡ (C A )B ≡ (C B )A

It is indeed obvious that

f : (a, b) 7→ f (a, b) ≡ a 7→ fa

where
fa : b 7→ f (a, b)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 88 / 307
Exceptions Generators Slices Freezing Parameters

Practical example
We want to use the function t 7→ sin(2πωt) for various frequencies ω.
Mathematically we have a function

(ω, t) 7→ sin(2πωt)

and for each ω we want the function

sinω : t 7→ sin(2πωt)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 89 / 307
Exceptions Generators Slices Freezing Parameters

Practical example
We want to use the function t 7→ sin(2πωt) for various frequencies ω.
Mathematically we have a function

(ω, t) 7→ sin(2πωt)

and for each ω we want the function

sinω : t 7→ sin(2πωt)

Python allows the following construction:

def make_sine ( freq ) :
" Make a sine function with frequency freq "
def mysine ( t ) :
return sin ( 2 * pi * freq * t )
return mysine

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 89 / 307
Exceptions Generators Slices Freezing Parameters

Computational Mathematics with Python

Linear Algebra

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 90 / 307
Vectors Matrices

17 Vectors
Vectors vs Lists
Plotting
Creating and Stacking

18 Matrices
Concept
Linear Algebra
Matrix Slices

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 91 / 307
Vectors Matrices

17 Vectors
Vectors vs Lists
Plotting
Creating and Stacking

18 Matrices
Concept
Linear Algebra
Matrix Slices

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 92 / 307
Vectors Matrices

The array type

Lists are almost like vectors but the operations on list are not the linear
algebra operation.
Definition
An array represents a vector in linear algebra. It is often initialised from a
list or another vector. Operations +, *, /, - are all elementwise. dot is
used for the scalar product.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 93 / 307
Vectors Matrices

Vector usage

Example
vec = array ( [ 1 . , 3 . ] ) # a vector in the plane
2 * vec # array ([ 2 . , 6 .])
vec * vec # array ([ 1 . , 9 .])
vec / 2 # array ([ 0 .5 , 1 . 5 ])
norm ( vec ) # norm
dot ( vec , vec ) # scalar product

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 94 / 307
Vectors Matrices

Vectors are similar to lists

I Access vectors via their indices
v = array ( [ 1 . , 2 . , 3 . ] )
v[0] # 1.

I The length of a vector is still obtained by the function len.

len ( v ) # 3

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 95 / 307
Vectors Matrices

Vectors are similar to lists

I Access vectors via their indices
v = array ( [ 1 . , 2 . , 3 . ] )
v[0] # 1.

I The length of a vector is still obtained by the function len.

len ( v ) # 3

I Parts of vectors using slices

v [ : 2 ] # array ([ 1 . , 2 .])

I Replace parts of vectors using slices

v [ : 2 ] = [ 10 , 20 ]
v # array ([ 10 . , 20 . , 3 .])

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 95 / 307
Vectors Matrices

Vectors are not lists! I

Operations are not the same:

I Operations + and * are different
I More operations are defined: -, /
I Many functions act elementwise on vectors: exp, sin, sqrt, etc.
I Scalar product with dot

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 96 / 307
Vectors Matrices

Vectors are not lists! II

I Vectors have a fixed size: no append method

I Only one type throughout the whole vector (usually float or
complex but beware of the int type!!)
I Vector slices are views:
v = array ( [ 1 . , 2 . , 3 . ] )
v1 = v [ : 2 ] # v is array ([ 1 . , 2 .])
v1 [ 0 ] = 0 . # if I change v1 ...
v # array ([ 0 . , 2 . , 3 .]) v is changed too !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 97 / 307
Vectors Matrices

More examples I

v1 = array ( [ 1 . , 2 . , 3 . ] ) # don ’t forget the dots !

v2 = array ( [2 , 0 , 1 . ] ) # one dot is enough
v1 + v2 ; v1 / v2 ; v1 - v2 ; v1 * v2

3 * v1
3 * v1 + 2 * v2

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 98 / 307
Vectors Matrices

More examples II

dot ( v1 , v2 ) # scalar product

cos ( v1 ) # cosine , elementwise

# access
v1 [ 0 ] # 1 .
v1 [ 0 ] = 10

# slices
v1 [ : 2 ] # array ([ 10 . , 2 .])
v1 [ : 2 ] = [0 , 1 ] # now v1 == array ([ 0 . , 1 . , 3 .])
v1 [ : 2 ] = [1 ,2 , 3 ] # error !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 99 / 307
Vectors Matrices

Vectors and plotting

The linspace method is a convenient way to create equally spaced arrays.

xs = linspace (0 , 10 , 200 ) # 200 points between 0 and 10
xs [ 0 ] # the first point is zero
xs [ - 1 ] # the last is ten

So for example the plot of the sine function between zero and ten will be
obtain by:
plot ( xs , sin ( xs ))

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 100 / 307
Vectors Matrices

Vectorised Functions

Note that not all functions may be applied on vectors. For instance this
one:
def const ( x ) :
return 1

We will see later how to automatically vectorise a function so that it works

componentwise on vectors.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 101 / 307
Vectors Matrices

Creating vectors

Some handy methods to quickly create vectors:

zeros zeros(n) creates a vector of size n filled with zeros
ones ones(n) is the same filled with ones
rand rand(n) creates a vector randomly filled with uniform
distribution between zero and one
empty empty(n) creates an “empty” vector of size n (try it!)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 102 / 307
Vectors Matrices

Concatenating Vectors
Since the + operation is redefined we need a means to concatenate
vectors. This is where the command hstack comes to help.
hstack([v1,v2,...,vn]) concatenates the vectors v1, v2, . . . , vn.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 103 / 307
Vectors Matrices

Concatenating Vectors
Since the + operation is redefined we need a means to concatenate
vectors. This is where the command hstack comes to help.
hstack([v1,v2,...,vn]) concatenates the vectors v1, v2, . . . , vn.
Symplectic permutation
We have a vector of size 2n. We want to permute the first half with the
second half of the vector with sign change:

(x1 , x2 , . . . , xn , xn+1 , . . . , x2n ) 7→ (−xn , −xn+1 , . . . , −x2n , x1 , . . . , xn )

def symp ( v ) :
n = len ( v ) / / 2 # use the integer division //
return hstack ( [ - v [ - n : ] , v [ : n ] ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 103 / 307
Vectors Matrices

17 Vectors
Vectors vs Lists
Plotting
Creating and Stacking

18 Matrices
Concept
Linear Algebra
Matrix Slices

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 104 / 307
Vectors Matrices

Matrices as Lists of Lists

Definition
Matrices are represented by arrays of lists of rows, which are lists as well.

# the identity matrix in 2D

id = array ( [ [ 1 . , 0 . ] , [ 0 . , 1 . ] ] )
# python allows this :
id = array ( [ [ 1 . , 0 . ] ,
[0., 1.]])
# which is more readable

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 105 / 307
Vectors Matrices

Accessing Matrix Entries

Matrix coefficients are accessed with two indices:

M = array ( [ [ 1 . , 2 . ] ,[ 3 . , 4 . ] ] )
M [0 , 0 ] # first line , first column : 1 .
M [ -1 , 0 ] # last line , first column : 3 .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 106 / 307
Vectors Matrices

Creating Matrices

Some convenient methods to create matrices are:

eye eye(n) is the identity matrix of size n
zeros zeros([n,m]) fills an n × m matrix with zeros
rand rand(n,m) is the same with random values
empty empty([n,m]) same with “empty” values

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 107 / 307
Vectors Matrices

Shape

The shape of a matrix is the tuple of its dimensions. The shape of an

n × m matrix is (n,m). It is given by the method shape:
M = eye ( 3 )
M . shape # (3 , 3 )

V = array ( [ 1 . , 2 . , 1 . , 4 . ] )
V . shape # (4 ,) <- tuple with one element

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 108 / 307
Vectors Matrices

Transpose

You may switch the two shape elements by transposing the matrix. The
transpose of a matrix Aij is a matrix B such that

Bij = Aji

A = ...
A . shape # 3 , 4

B = A . T # A transpose
B . shape # 4 , 3

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 109 / 307
Vectors Matrices

Matrix vector multiplication

The mathematical concept of reduction:
X
aij xj
j

is translated in python in the function dot:

angle = pi / 3
M = array ( [ [ cos ( angle ) , - sin ( angle ) ] ,
[ sin ( angle ) , cos ( angle ) ] ] )
V = array ( [ 1 . , 0 . ] )
Y = dot (M , V ) # the product M . V

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 110 / 307
Vectors Matrices

Matrix vector multiplication

The mathematical concept of reduction:
X
aij xj
j

is translated in python in the function dot:

angle = pi / 3
M = array ( [ [ cos ( angle ) , - sin ( angle ) ] ,
[ sin ( angle ) , cos ( angle ) ] ] )
V = array ( [ 1 . , 0 . ] )
Y = dot (M , V ) # the product M . V

Elementwise vs. matrix multiplication

The multiplication operator * is always elementwise. It has nothing to do
with the dot operation. A*V is a legal operation which will be explained
later on.
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 110 / 307
Vectors Matrices

Solving a Linear System

If A is a matrix and b is a vector you solve the linear equation
A·x =b
using solve which has the syntax x = solve(A,b).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 111 / 307
Vectors Matrices

Solving a Linear System

If A is a matrix and b is a vector you solve the linear equation
A·x =b
using solve which has the syntax x = solve(A,b).
Example
We want to solve (
x1 + 2x2 = 1
3x1 + 4x2 = 4

A = array ( [ [ 1 . , 2 . ] ,
[3., 4.]])
b = array ( [ 1 . , 4 . ] )
x = solve (A , b )
dot (A , x ) # should be almost b

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 111 / 307
Vectors Matrices

Slices

Slices are similar to that of lists and vectors except that there are now two
dimensions.
M[i,:] a vector filled by the row i of M
M[:,j] a vector filled by the column j of M
M[2:4,:] slice 2:4 on the lines only
M[2:4,1:4] slice on lines and columns

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 112 / 307
Vectors Matrices

Slices

Slices are similar to that of lists and vectors except that there are now two
dimensions.
M[i,:] a vector filled by the row i of M
M[:,j] a vector filled by the column j of M
M[2:4,:] slice 2:4 on the lines only
M[2:4,1:4] slice on lines and columns

Omitting a dimension
If you omit an index or a slice, SciPy assumes you are taking rows only.
M[3] is the third row of M
M[1:3] is a matrix with the second and third rows of M .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 112 / 307
Vectors Matrices

Altering a Matrix

You may alter a matrix using slices or direct access.

I M[2,3] = 2.
I M[2,:] = <a vector>
I M[1:3,:]= <a matrix>
I M[1:4,2:5]= <a matrix>
The matrices and vectors above must have the right size to “fit” in the
matrix M .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 113 / 307
Vectors Matrices

Computational Mathematics with Python

Linear Algebra 2

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 114 / 307
Matrices and Vectors Other Types

Computational Mathematics with Python

Linear Algebra 2

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 115 / 307
Matrices and Vectors Other Types

19 Matrices and Vectors

Dot Product
Rank and Shape
Building Matrices
Methods

20 Other Types
None
Complex Numbers

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 116 / 307
Matrices and Vectors Other Types

19 Matrices and Vectors

Dot Product
Rank and Shape
Building Matrices
Methods

20 Other Types
None
Complex Numbers

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 117 / 307
Matrices and Vectors Other Types

Dot multiplication

vector vector X
s= xi yi
i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 118 / 307
Matrices and Vectors Other Types

Dot multiplication

vector vector X
s= xi yi
i
matrix vector X
yi = Aij xj
j

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 118 / 307
Matrices and Vectors Other Types

Dot multiplication

vector vector X
s= xi yi
i
matrix vector X
yi = Aij xj
j

matrix matrix X
Cij = Aik Bkj
k

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 118 / 307
Matrices and Vectors Other Types

Dot multiplication

vector vector X
s= xi yi
i
matrix vector X
yi = Aij xj
j

matrix matrix X
Cij = Aik Bkj
k
vector matrix X
yj = xi Aij
i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 118 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Example

0 1 2 3 access shape rank kind

M[:2,1:-1] (2, 2) 2 matrix
4 5 6 7

8 9 10 11

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Example

0 1 2 3 access shape rank kind

M[:2,1:-1] (2, 2) 2 matrix
4 5 6 7 M[1,:] 4 1 vector
8 9 10 11

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Example

0 1 2 3 access shape rank kind

M[:2,1:-1] (2, 2) 2 matrix
4 5 6 7 M[1,:] 4 1 vector
8 9 10 11 M[1,1] ∅ 0 scalar

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Example

0 1 2 3 access shape rank kind

M[:2,1:-1] (2, 2) 2 matrix
4 5 6 7 M[1,:] 4 1 vector
8 9 10 11 M[1,1] ∅ 0 scalar
M[1:2,:] (1, 4) 2 matrix

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Rank of matrix slices

When slicing the rank of the result object is as follows:
access rank kind
index,index 0 scalar
slice,index 1 vector
slice,slice 2 matrix

Example

0 1 2 3 access shape rank kind

M[:2,1:-1] (2, 2) 2 matrix
4 5 6 7 M[1,:] 4 1 vector
8 9 10 11 M[1,1] ∅ 0 scalar
M[1:2,:] (1, 4) 2 matrix
M[1:2,1:2] (1, 1) 2 matrix
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 119 / 307
Matrices and Vectors Other Types

Reshaping

From a given tensor (vector or matrix) one may obtain another tensor by
reshaping.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 120 / 307
Matrices and Vectors Other Types

Reshaping Example

A = arange ( 6 )
0 1 2 3 4 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 121 / 307
Matrices and Vectors Other Types

Reshaping Example

A . reshape (1 , 6 )
0 1 2 3 4 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 121 / 307
Matrices and Vectors Other Types

Reshaping Example

2
A . reshape (6 , 1 )
3

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 121 / 307
Matrices and Vectors Other Types

Reshaping Example

0 1 2
A . reshape (2 , 3 )
3 4 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 121 / 307
Matrices and Vectors Other Types

Reshaping Example

0 1

A . reshape (3 , 2 ) 2 3

4 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 121 / 307
Matrices and Vectors Other Types

Reshaping Trick

Note that python can guess one of the new dimensions. Just give a
negative integer for the dimension to be guessed:
A = arange ( 12 ) # a vector of length 12

A . reshape (1 , - 1 ) # row matrix

A . reshape ( -1 , 1 ) # column matrix

A . reshape (3 , - 1 ) # 3 , 4 matrix
A . reshape ( -1 , 4 ) # same

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 122 / 307
Matrices and Vectors Other Types

Building Matrices

I Piling vectors
I Stacking vectors
I Stacking column matrices

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 123 / 307
Matrices and Vectors Other Types

Building Matrices

I Piling vectors
I Stacking vectors
I Stacking column matrices
The universal method to build matrices is concatenate. This function is
called by several convenient functions
I hstack to stack matrices horizontally
I vstack to stack matrices vertically
I column_stack to stack vectors in columns

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 123 / 307
Matrices and Vectors Other Types

Stacking Vectors

v1 = array ( [1 , 2 ] )
v2 = array ( [3 , 4 ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 124 / 307
Matrices and Vectors Other Types

Stacking Vectors

v1 = array ( [1 , 2 ] )
v2 = array ( [3 , 4 ] ) 1 2

3 4
vstack ( [ v1 , v2 ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 124 / 307
Matrices and Vectors Other Types

Stacking Vectors

v1 = array ( [1 , 2 ] )
v2 = array ( [3 , 4 ] ) 1 3

2 4
column_stack ( [ v1 , v2 ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 124 / 307
Matrices and Vectors Other Types

sum, max, min

Example
You may perform a number of
operations on arrays, either on 1 2 3 4
the whole array, or
column-wise or row-wise. The 5 6 7 8
most common are
I max A . sum ()
I min 36
I sum

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 125 / 307
Matrices and Vectors Other Types

sum, max, min

Example
You may perform a number of
operations on arrays, either on 1 2 3 4
the whole array, or
column-wise or row-wise. The 5 6 7 8
most common are
I max A . sum ( axis = 0 )
I min The result is a vector
I sum
6 8 10 12

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 125 / 307
Matrices and Vectors Other Types

sum, max, min

Example
You may perform a number of
operations on arrays, either on 1 2 3 4
the whole array, or
column-wise or row-wise. The 5 6 7 8
most common are
I max A . sum ( axis = 1 )
I min The result is a vector
I sum
10 26

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 125 / 307
Matrices and Vectors Other Types

19 Matrices and Vectors

Dot Product
Rank and Shape
Building Matrices
Methods

20 Other Types
None
Complex Numbers

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 126 / 307
Matrices and Vectors Other Types

None

I A function always returns

Example
a value. If you don’t
specify any, then the def f () :
object None is returned. " do nothing "
r = f ()
print r
None

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 127 / 307
Matrices and Vectors Other Types

None

I A function always returns

Example
a value. If you don’t
specify any, then the def f () :
object None is returned. ...
return None
I Sometimes you want to
explicitly return None to
get out of the function.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 127 / 307
Matrices and Vectors Other Types

None

I A function always returns

Example
a value. If you don’t
specify any, then the def f ( arg = None ) :
object None is returned. if arg is None :
arg = [ ]
I Sometimes you want to
...
explicitly return None to
get out of the function.
I None is also useful for
default arguments.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 127 / 307
Matrices and Vectors Other Types

Complex Numbers
Complex numbers are as easy to handle as real numbers. Just use the
syntax <number>j to refer to an imaginary number.
I = 1j
I * * 2 # -1
abs ( I ) # 1
I * ( 1 + I ) # -1 + i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 128 / 307
Matrices and Vectors Other Types

Complex Numbers
Complex numbers are as easy to handle as real numbers. Just use the
syntax <number>j to refer to an imaginary number.
I = 1j
I * * 2 # -1
abs ( I ) # 1
I * ( 1 + I ) # -1 + i

The real and imaginary parts of a complex number are given by the real
and imag properties

z = ( 1 + 1j ) * * 2
z . real # 0
z . imag # 2

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 128 / 307
Matrices and Vectors Other Types

Complex Numbers
Complex numbers are as easy to handle as real numbers. Just use the
syntax <number>j to refer to an imaginary number.
I = 1j
I * * 2 # -1
abs ( I ) # 1
I * ( 1 + I ) # -1 + i

The real and imaginary parts of a complex number are given by the real
and imag properties or with the numpy function real and imag for arrays
only.
zs = array ( [ 1 . , 3 . ] ) +
z = ( 1 + 1j ) * * 2
1j * array ( [ 2 . , 4 . ] )
z . real # 0
real ( zs ) # array ([ 1 . , 3 .])
z . imag # 2
imag ( zs ) # array ([ 2 . , 4 .])

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 128 / 307
Matrices and Vectors Other Types

Computational Mathematics with Python

Modules, Booleans(cont.), Files, Recursions etc.

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 129 / 307
Booleans Recursion File Handling

Computational Mathematics with Python

Modules, Booleans(cont.), Files, Recursions etc.

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 130 / 307
Booleans Recursion File Handling

Modules

21 Booleans

22 Recursion

23 File Handling

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 131 / 307
Booleans Recursion File Handling

First steps with Modules

Python comes along with many different libraries. You may also install
more of those.
numpy and Scipy are examples of such libraries.
You may either
I load some objects only:

from numpy import array , vander

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 132 / 307
Booleans Recursion File Handling

First steps with Modules

Python comes along with many different libraries. You may also install
more of those.
numpy and Scipy are examples of such libraries.
You may either
I load some objects only:

from numpy import array , vander

I or load the entire library:

from numpy import *

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 132 / 307
Booleans Recursion File Handling

First steps with Modules

Python comes along with many different libraries. You may also install
more of those.
numpy and Scipy are examples of such libraries.
You may either
I load some objects only:

from numpy import array , vander

I or load the entire library:

from numpy import *

I or choose to import all the library inside a single variable

import numpy
...
numpy . array (...)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 132 / 307
Booleans Recursion File Handling

How to avoid to “destroy” functions

A usual mistake
from scipy . linalg import eig
A = array ( [ [1 , 2 ] ,[3 , 4 ] ] )
( eig , eigvec ) = eig ( A )
.....
(c , d ) = eig ( B ) # raises an error , which ?

How to avoids these unintended effects:

Better with import ... as construct
import scipy . linalg as sl
A = array ( [ [1 , 2 ] ,[3 , 4 ] ] )
( eig , eigvec ) = sl . eig ( A ) # eig and sl . eig are different obje
.....
(c , d ) = sl . eig ( B )
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 133 / 307
Booleans Recursion File Handling

Modules

21 Booleans

22 Recursion

23 File Handling

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 134 / 307
Booleans Recursion File Handling

Boolean casting

Definition
Using an if statement with a non-boolean type casts it to a boolean. The
rules are as follow:

bool False True

string ’’ ’not empty’
number 0 6= 0
list [] [...] (not empty)
tuple () (..,..) (not empty)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 135 / 307
Booleans Recursion File Handling

Boolean casting

Definition
Using an if statement with a non-boolean type casts it to a boolean. The
rules are as follow:

bool False True

string ’’ ’not empty’
number 0 6= 0
list [] [...] (not empty)
tuple () (..,..) (not empty)
array Error!

Notice that almost all types will be silently cast to booleans except arrays.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 135 / 307
Booleans Recursion File Handling

Examples
Empty list test
# L is a list
if L :
print " list not empty "
else :
print " list is empty "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 136 / 307
Booleans Recursion File Handling

Examples
Empty list test
# L is a list
if L :
print " list not empty "
else :
print " list is empty "

Parity test
# n is an integer
if n % 2 :
print " n is odd "
else :
print " n is even "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 136 / 307
Booleans Recursion File Handling

Modules

21 Booleans

22 Recursion

23 File Handling

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 137 / 307
Booleans Recursion File Handling

Avoid Recursion!

Recursion Iteration
def f ( N ) : for i in xrange (
if N = = 0 : return 0 10000000 ) :
return f ( N - 1 ) pass

This is the simplest recursion In iterative programming those

program. In python it chokes loops may at least be 10000
for N ≥ 1000. Note that this times bigger! That is an empty
is the best case since this for loop of 10 000 000 is not
program doesn’t do anything. a problem.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 138 / 307
Booleans Recursion File Handling

Avoid Recursion!

Recursion Iteration
def f ( N ) : for i in xrange (
if N = = 0 : return 0 10000000 ) :
return f ( N - 1 ) pass

This is the simplest recursion In iterative programming those

program. In python it chokes loops may at least be 10000
for N ≥ 1000. Note that this times bigger! That is an empty
is the best case since this for loop of 10 000 000 is not
program doesn’t do anything. a problem.
Note that in some very special cases (tree traversal) recursion is almost
unavoidable.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 138 / 307
Booleans Recursion File Handling

Modules

21 Booleans

22 Recursion

23 File Handling

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 139 / 307
Booleans Recursion File Handling

File I/O

File I/O (in- and output) is essential when

I working with measured or scanned data
I interacting with other programs
I saving information for comparions or other postprocessing needs
I .....

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 140 / 307
Booleans Recursion File Handling

File objects

A file is a Python object with associated methods:

myfile = open ( ’ measurement . dat ’ , ’r ’) # creating a read - only file

and here a generator to extract the data:

for line in myfile :
data = line . split ( ’; ’)
print ’ time % s sec temperature % s C ’ % ( data [ 0 ] , data [ 1 ] )

or direct extraction into a list

data = list ( myfile )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 141 / 307
Booleans Recursion File Handling

File close method

A file has to be closed before it can be reread.

myfile . close () # closes the file object

It is automatically closed when

I the program ends

Before a file is closed, you won’t see any changes in it by an external

editor!

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 142 / 307
Booleans Recursion File Handling

File close method

A file has to be closed before it can be reread.

myfile . close () # closes the file object

It is automatically closed when

I the program ends
I the enclosing program unit (e.g. function) is left .
Before a file is closed, you won’t see any changes in it by an external
editor!

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 142 / 307
Booleans Recursion File Handling

File Modes

file1 = open ( ’ file1 . dat ’ , ’r ’) # read only

file2 = open ( ’ file2 . dat ’ , ’r + ’) # read / write
file3 = open ( ’ file3 . dat ’ , ’a ’) # append ( write to the end of th
file4 = open ( ’ file4 . dat ’ , ’w ’) # ( over -) write the file

The modes ’r’,’r+’,’a’ require that the file exists.

File append example
file3 = open ( ’ file3 . dat ’ , ’a ’)
file3 . write ( ’ something new \ n ’) # Note the ’\ n ’

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 143 / 307
Booleans Recursion File Handling

Module: pickle

read and write methods convert data to strings.

Complex data types (like arrays) cannot be written this way.
pickle solves this problem:
Pickle dump examples
import pickle
myfile = open ( ’ file . dat ’ , ’w ’)
a = rand ( 200 )
pickle . dump (a , myfile )

you can pickle any python object even code, e.g. functions.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 144 / 307
Booleans Recursion File Handling

Module: pickle

Pickle load examples

import pickle
myfile = open ( ’ file . dat ’ , ’r ’)
a = pickle . load ( myfile ) # restores the array

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 145 / 307
Booleans Recursion File Handling

Computational Mathematics with Python

Boolean arrays, more on iterations, from datatypes top classes

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 146 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 147 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 148 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Boolean Arrays

One may create views of an array using boolean arrays.

B = array ( [ [ True , False ] ,
[ False , True ] ] )
M = array ( [ [2 , 3 ] ,
[1 , 4 ] ] )
M[B] = 0
M # [[ 0 , 3 ] , [1 , 0 ]]
M [ B ] = 10 , 20
M # [[ 10 , 3 ] , [1 , 20 ]]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 149 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Creating boolean Arrays

It might be just as tedious to create the boolean array by hand than to
change the array directly. There are however many methods to create
boolean arrays.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 150 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Creating boolean Arrays

It might be just as tedious to create the boolean array by hand than to
change the array directly. There are however many methods to create
boolean arrays.
Any logical operator will create a boolean array instead of a boolean.
M = array ( [ [2 , 3 ] , [1 , 4 ] ] )
M > 2 # array ([[ False , True ] , [ False , True ]])
M = = 0 # array ([[ False , False ] , [ False , False ]])
N = array ( [ [2 , 3 ] , [0 , 0 ] ] )
M = = N # array ([[ True , True ] , [ False , False ]])
...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 150 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Creating boolean Arrays

It might be just as tedious to create the boolean array by hand than to
change the array directly. There are however many methods to create
boolean arrays.
Any logical operator will create a boolean array instead of a boolean.
M = array ( [ [2 , 3 ] , [1 , 4 ] ] )
M > 2 # array ([[ False , True ] , [ False , True ]])
M = = 0 # array ([[ False , False ] , [ False , False ]])
N = array ( [ [2 , 3 ] , [0 , 0 ] ] )
M = = N # array ([[ True , True ] , [ False , False ]])
...

This allows the elegant syntax:

M[M>2] = 0
# all the elements > 2 are replaced by 0

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 150 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Comparing Arrays

Note that because array comparison create boolean arrays one cannot
compare arrays directly.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 151 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Comparing Arrays

Note that because array comparison create boolean arrays one cannot
compare arrays directly.
The solution is to use the methods all and any:
A = array ( [ [1 , 2 ] ,[3 , 4 ] ] )
B = array ( [ [1 , 2 ] ,[3 , 3 ] ] )
A = = B # creates array ([[ True , True ] , [ True , False ]])
( A = = B ). all () # False
( A ! = B ). any () # True

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 151 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Boolean Operations
For the same reason as before you cannot use and, or nor not on boolean
arrays! Use the following replacement operators instead:

logic operator replacement for bool arrays

A and B A & B
A or B A | B
not A -A

a = array ( [ True , True , False , False ] )

b = array ( [ True , False , True , False ] )

a and b # error !
a & b # array ([ True , False , False , False ])
a | b # array ([ True , Trues , True , False ])
- a # array ([ False , False , True , True ])

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 152 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 153 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Multiple iteration
Comprehensive double for
# M is a matrix
M . shape # (3 , 4 )
flat = [ M [i , j ] for i in range ( M . shape [ 0 ] )
for j in range ( M . shape [ 1 ] ) ]
flat # a list of length 12

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 154 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Multiple iteration
Comprehensive double for
# M is a matrix
M . shape # (3 , 4 )
flat = [ M [i , j ] for i in range ( M . shape [ 0 ] )
for j in range ( M . shape [ 1 ] ) ]
flat # a list of length 12

zip
x_values = [1 , 2 , 3 , 4 ]
y_values = [ 10 , 20 , 30 , 40 ]
for x , y in zip ( x_values , y_values ) :
print " the value at % f is % f " % (x , y )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 154 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 155 / 307
 Assignment
Boolean Arrays Iteration Objects and References Objects and Types Classes

Variables !""#$%&'%()*+',('")-./'*()+'0'+'%*'"1))
are References

>>> x = [0, 1, 2] x 0 1 2
# y = x cause x and y to point
# at the same list y
>>> y = x

# changes to y also change x x

>>> y[1] = 6 0 6 2
>>> print x y
[0, 6, 2]

# re-assigning y to a new list x 0 6 2

# decouples the two lists
>>> y = [3, 4] y 3 4

(Source: E. Jones and T. Oliphant)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 156 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Equality vs Identity I

Definition
I Identity is the property of two variables to be the reference to the
same object.
The identity operator is is.
I Equality is the property of two objects to be equal.
The equality operator is ==.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 157 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Equality vs Identity II

Example
To test whether to variables are the same reference you may use is:
L = [ ’a ’ , ’b ’]
L2 = L
L is L2 # True

L2 = [ ’a ’ , ’b ’]
L is L2 # False
L = = L2 # True

Notice that is should not be used to compare two objects!

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 158 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Copy

I To create a copy of an object use the function copy of the module

copy:
c = [3 , 4 ]
d = copy ( c )
d[0] = 0
c # [3 , 4 ]
d # [0 , 4 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 159 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Copy

I To create a copy of an object use the function copy of the module

copy:
c = [3 , 4 ]
d = copy ( c )
d[0] = 0
c # [3 , 4 ]
d # [0 , 4 ]

I For an array it is more efficient to use the method copy:

A = arange ( 12 ). reshape (3 , 4 )
B = A . copy ()
B [1 , 2 ] = 0 # only changes B , not A

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 159 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 160 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Type of an object

Each object has a type that may be obtained using the function type.
x = [1 , 2 ] # list
type ( x ) # returns list
y = [2 , 3]
type ( x ) = = type ( y ) # True

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 161 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Checking the type

To check the type of a variable, always use isinstance:
L = [1 , 2 ]
if isinstance (L , list ) :
print " L is a list "

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 162 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Checking the type

To check the type of a variable, always use isinstance:
L = [1 , 2 ]
if isinstance (L , list ) :
print " L is a list "

We have already come across different types:

I float
I int
I complex
I list
I tuple
I module
I function
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 162 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Methods and Properties

Definition
Method and property A method or property is a function bound to an
object. The syntax is
< object > . method ( < arguments ... > )
# or
< object > . property

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 163 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Methods and Properties

Definition
Method and property A method or property is a function bound to an
object. The syntax is
< object > . method ( < arguments ... > )
# or
< object > . property

I lists: append(<obj>)
I arrays: shape, sum(), max() etc.
I complex: real, imag

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 163 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

24 Boolean Arrays
Modifying Arrays
Comparing Arrays
25 Iteration

26 Objects and References

Copying
27 Objects and Types
Type Checking
Methods and Properties
28 Classes
Motivation
The Complex Example
__init__ and self

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 164 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Why New Types?

We have already seen many types existing in python. For instance, float,
int and complex are types.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 165 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Why New Types?

We have already seen many types existing in python. For instance, float,
int and complex are types.

What if we want new structures to be represented in python? For example

I polynomials
I quaternions (for rotations)
I symbols (for symbolic calculations)
I etc.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 165 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Creating new Types

A new type is called a class. The syntax is as follows:

class Complex ( object ) :
def __init__ ( self , r , i ) :
self . r = r
self . i = i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 166 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Creating new Types

A new type is called a class. The syntax is as follows:

class Complex ( object ) :
def __init__ ( self , r , i ) :
self . r = r
self . i = i

I object is a keyword
I __init__ is called at the creation of the object
I self is the object itself

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 166 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Usage Example

One may use this class using the following syntax:

z = Complex (2 , 3 )
z.r # 2
z.i # 3

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 167 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Adding methods

A method is a function for a particular class.

class Complex ( object ) :
...
def module ( self ) :
return sqrt ( self . r * * 2 + self . i * * 2 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 168 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Adding methods

A method is a function for a particular class.

class Complex ( object ) :
...
def module ( self ) :
return sqrt ( self . r * * 2 + self . i * * 2 )

It is used as the other methods you know:

z = Complex (4 , 3 )
z . module () # 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 168 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

The __init__ method

I Called at the creation of the object

z = Complex (2 , 3 ) # here __init__ is called

I Often used to initialize the state of the object

I optional if no initialisation has to be done

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 169 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

The self argument

All the methods get a special first argument. It is often called self by
convention.
This argument contains the object itself. So a method is basically a usual
function which gets an object as a first argument.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 170 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

The self argument

All the methods get a special first argument. It is often called self by
convention.
This argument contains the object itself. So a method is basically a usual
function which gets an object as a first argument.
More generally, if an object obj is of class cls then the following calls are
equivalent:
obj . method ( arg1 , arg2 )
cls . method ( obj , arg1 , arg2 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 170 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

The self argument

All the methods get a special first argument. It is often called self by
convention.
This argument contains the object itself. So a method is basically a usual
function which gets an object as a first argument.
More generally, if an object obj is of class cls then the following calls are
equivalent:
obj . method ( arg1 , arg2 )
cls . method ( obj , arg1 , arg2 )

In the previous example the two calls are equivalent:

z = Complex (2 , 3 ) # the type of z is Complex
z . abs ()
Complex . abs ( z ) # exactly equivalent to z . abs ()

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 170 / 307
Boolean Arrays Iteration Objects and References Objects and Types Classes

Computational Mathematics with Python

Classes

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 171 / 307
Debugging Operator Overloading

29 Debugging
Motivation
Stack
Debugging Mode

30 Operator Overloading
Operators
Brackets

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 172 / 307
Debugging Operator Overloading

29 Debugging
Motivation
Stack
Debugging Mode

30 Operator Overloading
Operators
Brackets

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 173 / 307
Debugging Operator Overloading

What is a Bug?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 174 / 307
Debugging Operator Overloading

Bugs in Python

There are two kinds of bug:

I An exception is raised and not catched
I The code does not function properly

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 175 / 307
Debugging Operator Overloading

Bugs in Python

There are two kinds of bug:

I An exception is raised and not catched←Easy to fix
I The code does not function properly

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 175 / 307
Debugging Operator Overloading

Bugs in Python

There are two kinds of bug:

I An exception is raised and not catched←Easy to fix
I The code does not function properly←More difficult

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 175 / 307
Debugging Operator Overloading

Bugs in Python

There are two kinds of bug:

I An exception is raised and not catched←Easy to fix
I The code does not function properly←More difficult

We are only concerned with the first case in what folows.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 175 / 307
Debugging Operator Overloading

The Stack

When an exception is raised you see

the stack. It is all the functions that
called the function where the
exception was raised.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 176 / 307
Debugging Operator Overloading

The Stack

Example
def f () :
...
When an exception is raised you see g ()
def g () :
the stack. It is all the functions that
...
called the function where the h ()
exception was raised. def h () :
raise Exception ()

f ()

The stack is now f, g, h.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 176 / 307
Debugging Operator Overloading

Debugging in IPython

After an exception was raised, enter the debug mode by typing debug.
You are now in debug mode.
You may now inspect the current variables and work as usual.

Demo

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 177 / 307
Debugging Operator Overloading

Debug Commands

h help
q quit
l shows current line
u go up in the stack
d go down in the stack

Short Variable Names

If you want to inspect a variable with name, for example, h, you must use
!h.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 178 / 307
Debugging Operator Overloading

Trigger the Debug Mode

I The simplest way to trigger the debug mode is to throw an exception.

For example by writing 1/0 where you want the execution to stop
(but you can’t resume it).
I A nicer way is to use the following code
# at the beginning of the file :
from IPython . Debugger import Tracer
dh = Tracer ()
...
dh () # triggers the debug mode

The advantage with that approach is that you may now resume the
execution.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 179 / 307
Debugging Operator Overloading

29 Debugging
Motivation
Stack
Debugging Mode

30 Operator Overloading
Operators
Brackets

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 180 / 307
Debugging Operator Overloading

Operators
Operators (+, *, -, etc.) are defined as methods in python.

operator method
+ __add__
* __mul__
- __sub__
/ __div__

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 181 / 307
Debugging Operator Overloading

Operators
Operators (+, *, -, etc.) are defined as methods in python.

operator method
+ __add__
* __mul__
- __sub__
/ __div__

Example
Given two lists L1 and L2, the following are exactly equivalent:
L = L1 + L2
L = L1 . __add__ ( L2 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 181 / 307
Debugging Operator Overloading

Redefining Operators

In your own classes you may define the meaning of operators!

class polynomial ( object ) :
...
def __add__ ( self , other ) :
...
return ...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 182 / 307
Debugging Operator Overloading

Redefining Operators

In your own classes you may define the meaning of operators!

class polynomial ( object ) :
...
def __add__ ( self , other ) :
...
return ...

You may use that code in this way:

p1 = polynomial (...)
p2 = polynomial (...)
p = p1 + p2 # here p1 . __add__ is called

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 182 / 307
Debugging Operator Overloading

More Operators

Many more operators may be redefined:

I ==, !=
I <=, <, >, >=
I +=, -=, *=, /=
I %
I etc.
You will find a complete list on the Python Quick Reference Page
(http://rgruet.free.fr/PQR25/PQR2.5.html#SpecialMethods)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 183 / 307
Debugging Operator Overloading

Brackets
You may also redefine the brackets:
bracket method
() __call__
[] __getitem__

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 184 / 307
Debugging Operator Overloading

Brackets
You may also redefine the brackets:
bracket method
() __call__
[] __getitem__

Example
class polynomial ( object ) :
...
def __call__ ( self , x ) :
return self . eval ( x )

Which now may be used as:

p = polynomial (...)
p ( 3 .) # value of p at 3 .
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 184 / 307
Debugging Operator Overloading

Computational Mathematics with Python

Classes

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 185 / 307
Classes vs Functions Attributes Dos and don’t Training 9

31 Classes vs Functions
Problem
Solution

32 Attributes

33 Dos and don’t

34 Training 9
Formula
Pieces of Code
Full Code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 186 / 307
Classes vs Functions Attributes Dos and don’t Training 9

31 Classes vs Functions
Problem
Solution

32 Attributes

33 Dos and don’t

34 Training 9
Formula
Pieces of Code
Full Code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 187 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Functions

Functions are building blocks of computer programming.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 188 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Functions

Functions are building blocks of computer programming.

The disadvantage is: they are stateless.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 188 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Functions

Functions are building blocks of computer programming.

The disadvantage is: they are stateless.

A function
I takes an input
I computes, creates local variables
I cleans everything except the output
I returns the output
I ...and dies

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 188 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

def newton (f , x0 , tol ) :

...
return solution

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 189 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

def newton (f , x0 , tol ) :

...
return solution

But I want the number of iteration used.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 189 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

def newton (f , x0 , tol ) :

...
return solution

But I want the number of iteration used.

Bad solution:
def newton (f , x0 , tol ) :
...
return solution , nb_iterations

Bad solution because now it always returns the number of iterations.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 189 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

def newton (f , x0 , tol ) :

...
return solution

But I want the number of iteration used.

Bad solution:
def newton (f , x0 , tol ) :
...
return solution , nb_iterations

Bad solution because now it always returns the number of iterations.

It is as good as it gets with functions

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 189 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Input

# many arguments here :

def integrate (f , a , b , hmin , tol , ...) :

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 190 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Input

# many arguments here :

def integrate (f , a , b , hmin , tol , ...) :

Usage:
>>> integrate (f ,0 ,1 ,. 003 , 1e -6 ,...)
< bad result >
>>> integrate (f ,0 ,1 ,. 002 , 1e -6 ,...)
< bad result >

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 190 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Input

# many arguments here :

def integrate (f , a , b , hmin , tol , ...) :

Usage:
>>> integrate (f ,0 ,1 ,. 003 , 1e -6 ,...)
< bad result >
>>> integrate (f ,0 ,1 ,. 002 , 1e -6 ,...)
< bad result >

When the function dies, obviously it also forgets everything about the
input arguments!
You have to give the full set of input at every call.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 190 / 307
Classes vs Functions Attributes Dos and don’t Training 9

The solution: objects

An object can perform operations on its data and stay alive.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 191 / 307
Classes vs Functions Attributes Dos and don’t Training 9

The solution: objects

An object can perform operations on its data and stay alive.

Advantages are:
Output Methods return only the essential result, but intermediary
steps are available
Input The inputs are stored and one can modify one part of the
inputs

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 191 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

class Newton ( object ) :

def run ( self , x0 ) :
...
# store the nb_iterations in the object :
self . nb_iterations = ...
return solution

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 192 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

class Newton ( object ) :

def run ( self , x0 ) :
...
# store the nb_iterations in the object :
self . nb_iterations = ...
return solution

Usage:
solver = Newton ( f )
solution = solver . run (. 2 )
# now if I want the number of iterations :
nbIter = solver . nb_iterations

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 192 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Output

class Newton ( object ) :

def run ( self , x0 ) :
...
# store the nb_iterations in the object :
self . nb_iterations = ...
return solution

Usage:
solver = Newton ( f )
solution = solver . run (. 2 )
# now if I want the number of iterations :
nbIter = solver . nb_iterations

Possible because the the object solver stays alive (although the function
run dies).
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 192 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Example: Input

class Integrator ( object ) :

def __init__ ( self , f , a , b , tol , ...) :
self . f = f
self . a = a
...
def compute ( self ) :
...

Usage:
>>> sin_int = Integrator ( sin , 0 , 1 , 1e - 6 )
>>> sin_int . compute ()
< bad result >
>>> sin_int . tol = 1e - 3
>>> sin_int . compute ()

Now you can change arguments one at a time.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 193 / 307
Classes vs Functions Attributes Dos and don’t Training 9

31 Classes vs Functions
Problem
Solution

32 Attributes

33 Dos and don’t

34 Training 9
Formula
Pieces of Code
Full Code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 194 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Methods

Method
A method is a function bound to an object. The syntax is
< object > . method ( < arguments ... > )
# or
< object > . property

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 195 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Methods

Method
A method is a function bound to an object. The syntax is
< object > . method ( < arguments ... > )
# or
< object > . property

For instance:
I list: append(<obj>)
I array: shape, sum(), max() etc.
I complex: real, imag
I Complex: r, i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 195 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Class attributes

I Attributes in the class declaration are always class attributes (logical

but different from C++)
class Newton ( object ) :
tol = 1e - 8 # this is a * class * attribute !
N1 = Newton ( f )
N2 = Newton ( g )
N1 . tol # 1e - 8
N2 . tol = 1e - 4 # relax tolerance for g
N1 . tol # 1e - 8 only N2 was changed

I Class attributes are handy to simulate default values.

Newton . tol = 1e - 10 # now all the Newton classes have 1e - 5
N2 . tol # 1e - 4 because the object attribute is fetched

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 196 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Class and object syntax

I Attributes may be added, removed and accessed at any time by
anybody
class C :
pass # the class is empty
c = C ()
c . first = 1 # attributes created dynamically
c . second = 3

I You may put any kind of code inside a class statement! This code is
read only once.
class C :
2 + 2 # why not ?
def f ( self ) :
...
g = f # now the method g is added to the class

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 197 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Documentation

Document classes, functions or modules by adding a string in the

beginning of the indentation:
class myClass ( object ) :
""" This class does this and that """
def f ( self ) :
""" Does nothing """

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 198 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Debugging

For debugging purposes you may use the __repr__ method:

class Complex ( object ) :
...
def __repr__ ( self ) :
return " % f + % f . i " % ( self .a , self . b )

# in the console :
>>> z = Complex (1 , 2 )
>>> print z
1 + 2.i
>>> z
1 + 2.i

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 199 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Almost everything is an object

Almost everything is an object in Python.

def my_sine (...) :
...

def my_cosine (...) :

...

# this is allowed :
my_sine . derivative = my_cosine

Note that:
I functions, as anything else, are objects
I we added a new attribute to an existing object

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 200 / 307
Classes vs Functions Attributes Dos and don’t Training 9

31 Classes vs Functions
Problem
Solution

32 Attributes

33 Dos and don’t

34 Training 9
Formula
Pieces of Code
Full Code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 201 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Don’t

I No while
I No global
I No map
I No lambda

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 202 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Don’t

I No while
I No global
I No map
I No lambda

Don’t use recursive programming: it is slow and unreliable.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 202 / 307
Classes vs Functions Attributes Dos and don’t Training 9

No While

Don’t use while:

while < condition > :
...

because you don’t know whether it will ever stop.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 203 / 307
Classes vs Functions Attributes Dos and don’t Training 9

No While

Don’t use while:

while < condition > :
...

because you don’t know whether it will ever stop.

Always use this instead:

max_it = 100
for k in xrange ( max_it ) :
...
else :
raise Exception ( " No convergence in % d iterations " % k )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 203 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Function arguments

A function has access to variables outside their scope. Don’t use this
feature, the outcome is not reliable.
tolerance = 1e - 6
def algorithm () :
...
if abs ( error ) < tolerance # bad !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 204 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Function arguments

A function has access to variables outside their scope. Don’t use this
feature, the outcome is not reliable.
tolerance = 1e - 6
def algorithm () :
...
if abs ( error ) < tolerance # bad !

Give all the arguments as inputs:

def algorithm ( tolerance ) :
...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 204 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Errors

Don’t print out errors:

print " The algorithm did not converge " # bad !

it lacks information both for humans and computers.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 205 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Errors

Don’t print out errors:

print " The algorithm did not converge " # bad !

it lacks information both for humans and computers.

Throw an exception instead (with a message):

raise Exception ( " The algorithm did not converge " ) # good !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 205 / 307
Classes vs Functions Attributes Dos and don’t Training 9

For loops
Don’t use:
for k in range (...) :
...
element = my_list [ k ]

unless you have a good reason to do so.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 206 / 307
Classes vs Functions Attributes Dos and don’t Training 9

For loops
Don’t use:
for k in range (...) :
...
element = my_list [ k ]

unless you have a good reason to do so.

A better way is often:
for element in my_list :
...

or:
for element in my_generator :

because it is easier to read.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 206 / 307
Classes vs Functions Attributes Dos and don’t Training 9

No append
Avoid append. It is often used as:
my_list = [ ]
for k in xrange ( n ) :
... # compute some value here
my_list . append ( value )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 207 / 307
Classes vs Functions Attributes Dos and don’t Training 9

No append
Avoid append. It is often used as:
my_list = [ ]
for k in xrange ( n ) :
... # compute some value here
my_list . append ( value )

Instead, use a generator:

def generate_values ( n ) :
for k in xrange ( n ) :
...
yield value

# this is much more readable :

my_list = list ( generate_values ( n ))

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 207 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Documentation and Testing

Do give a short documentation to all your functions.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 208 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Documentation and Testing

Do give a short documentation to all your functions.

Always add some test function:

def algorithm (...) :
...

def test_algorithm () :
expected = 3
computed = algorithm (...)
assert expected = = computed

The keyword assert will raise an exception if the statement is not

evaluated to true.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 208 / 307
Classes vs Functions Attributes Dos and don’t Training 9

31 Classes vs Functions
Problem
Solution

32 Attributes

33 Dos and don’t

34 Training 9
Formula
Pieces of Code
Full Code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 209 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Understanding the Formula

The original formula was:
j−1
ci+1 − cij−1
cij =
xi+j − xi
with the initialisation:
ci0 = yi

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 210 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Understanding the Formula

The original formula was:
j−1
ci+1 − cij−1
cij =
xi+j − xi
with the initialisation:
ci0 = yi
So if we define the operator:

(∆j x)i = xi+j − xj

Then:
∆1 c j−1
cj =
∆j x

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 210 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Understanding the Formula

The original formula was:
j−1
ci+1 − cij−1
cij =
xi+j − xi
with the initialisation:
ci0 = yi
So if we define the operator:

(∆j x)i = xi+j − xj

Then:
∆1 c j−1
cj =
∆j x
How do we program the opertor ∆j in python?
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 210 / 307
Classes vs Functions Attributes Dos and don’t Training 9

∆j

Given a vector y the operator ∆j is in python:

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 211 / 307
Classes vs Functions Attributes Dos and don’t Training 9

∆j

Given a vector y the operator ∆j is in python:

y[j:] - y[:-j]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 211 / 307
Classes vs Functions Attributes Dos and don’t Training 9

∆j

Given a vector y the operator ∆j is in python:

y[j:] - y[:-j]

Isn’t it simple, short and readable?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 211 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Fetching the x and y values

The interpolation points are a list of tuples:

xy = [ ( 0 . , 1 .) , ( 1 . , 2 . 4 ) ,... ]

How to obtain the x and y values?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 212 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Fetching the x and y values

The interpolation points are a list of tuples:

xy = [ ( 0 . , 1 .) , ( 1 . , 2 . 4 ) ,... ]

How to obtain the x and y values?

Use arrays!!
xy = array ( xy ) # now xy is an nx2 matrix
xy [ : ,0 ] # x values
xy [ : ,1 ] # y values

That is what matrices are for.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 212 / 307
Classes vs Functions Attributes Dos and don’t Training 9

The Full Code of divdiff

x = xy [ : ,0 ] # x values
row = xy [ : ,1 ] # first row

for j in xrange (1 , len ( xy )) :

yield row [ 0 ]
row = ( row [ 1 : ] - row [ : - 1 ] ) / ( x [ j : ] - x [ : - j ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 213 / 307
Classes vs Functions Attributes Dos and don’t Training 9

Computational Mathematics with Python

Case Study: Initial Value Problem

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 214 / 307
Initial Value Problem

35 Initial Value Problem

Mathematical View
Class IVPSolver

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 215 / 307
Initial Value Problem

35 Initial Value Problem

Mathematical View
Class IVPSolver

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 216 / 307
Initial Value Problem

IVP

The problem:
Find a function u : [t0 , tf ] → Rn with the property

u̇(t) = f (t, u(t)) u(t0 ) = u0

is called an initial value problem (IVP).

f is called the right-hand side function, u0 the initial value.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 217 / 307
Initial Value Problem

Simple Example

The problem
u̇(t) = −3u(t) u(t0 ) = 5
has the solution
u(t) = e−3(t−t0 ) 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 218 / 307
Initial Value Problem

Discretisation

A numerical method computes approximations to the solution at discrete

times
t0 < t1 < . . . < tn = tf
with a step size hi = ti+1 − ti .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 219 / 307
Initial Value Problem

Euler’s Method

Euler’s explicit method is defined by

ui+1 = ui + hi f (ti , ui ) ti+1 = ti + hi

where ui is an approximation to u(ti ).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 220 / 307
Initial Value Problem

Runge-Kutta’s Classical Method

Here ui+1 is computed by passing some intermediate stages:

U1 = f (ti , ui )
hi hi
U2 = f (ti + , ui + U1 )
2 2
hi hi
U3 = f (ti + , ui + U2 )
2 2
U4 = f (ti + hi , ui + hi U3 )
hi
ui+1 = ui + (U1 + 2U2 + 2U3 + U4 )
6

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 221 / 307
Initial Value Problem

init

We initialize the problem

class IVPSolver ( object ) :
def __init__ ( self , f , u0 , t0 = 0 ) :
self . f = f # rhs function
self . u0 = u0 # initial value
self . t0 = t0 # initial time
self . ts = [ t0 ]
self . us = [ u0 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 222 / 307
Initial Value Problem

Class Attributes

We set some predefined (default) values

# default values for step size , simulation time and
# step number delimiter
h = . 01
time = 1 .
max_steps = 10000

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 223 / 307
Initial Value Problem

A dummy method

Some numerical methods use constant step sizes, other variable step sizes.
The change of step sizes is done by a method which is provided here as a
dummy method.
It will be replaced later by something meaningful, when variable stepsizes
are required.
def adjust_step_size ( self )
pass

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 224 / 307
Initial Value Problem

Step generator
We define a method independent step generator
def generator ( self , t , u , tf ) :
"""
Generates the (t , u ) values until t > tf
"""
for i in xrange ( self . max_steps ) :
t , u = self . step (t , u )
if t > tf :
break
yield t , u
self . adjust_stepsize ()
else :
raise Exception (
" Final time not reached within max_steps steps " )

This generator is method independent.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 225 / 307
Initial Value Problem

Solving the equation

Solving the IVP is done by the run method

def run ( self , time = None ) :
if time is None :
time = self . time
# start from the last time we stopped
t = t0 = self . ts [ - 1 ]
u = self . us [ - 1 ]
tus = list ( self . generator (t , u , t0 + time ))
self . ts . extend ( q [ 0 ] for q in tus )
self . us . extend ( q [ 1 ] for q in tus )
self . ats = array ( self . ts )
self . aus = array ( self . us )

Note the restarting capability.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 226 / 307
Initial Value Problem

Postprocessing

Postprocessing can be a plot or further computations:

def plot ( self ) :
"""
Plot the computed solution .
"""
if not hasattr ( self , ’ ats ’) :
raise Exception ( ’ No data to plot . ’)
plot ( self . ats , self . aus , ’. - ’)
xlabel ( ’ time ’)
ylabel ( ’ state ’)
if self . f . name is not None :
title ( self . f . name )

Note the exception. Other possibilities ....

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 227 / 307
Initial Value Problem

A method (heritage)

The stepper depends on the method:

class ExplicitEuler ( IVPSolver ) :
def step ( self , t , u ) :
return t + self .h , u + self . h * self . f (t , u )

Note, we use here a one step method.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 228 / 307
Initial Value Problem

Another method

Or the Runge–Kutta method

class RungeKutta4 ( IVPSolver ) :
def step ( self , t , u ) :
h = self . h
U1 = self . f (t , u )
U2 = self . f ( t + h / 2 . , u + h * U1 / 2 .)
U3 = self . f ( t + h / 2 . , u + h * U2 / 2 .)
U4 = self . f ( t + h , u + h * U3 )
return t +h , u + h / 6 . * ( U1 + 2 . * ( U2 + U3 ) + U4 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 229 / 307
Initial Value Problem

Overridding a class method

A method with local error estimation and step size control:
class RungeKutta34 ( IVPSolver ) :
tol = 1e - 6
def adjust_stepsize ( self ) :
self . h * = ( self . tol / self . error ) * * ( 1 / 4 )

def step ( self , t , u ) :

h = self . h
U1 = self . f (t , u )
U2 = self . f ( t + h / 2 . , u + h * U1 / 2 .)
U3 = self . f ( t + h /2 , u + h * U2 / 2 )
U3_1 = self . f ( t + h , u - h * U1 + 2 * h * U2 )
U4 = self . f ( t + h , u + h * U3 )
self . error = norm ( h / 6 * ( 2 * U2 + U3_1 - 2 * U3 - U4 ))
return t +h , u + h / 6 * ( U1 + 2 * ( U2 + U3 ) + U4 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 230 / 307
Initial Value Problem

Creating an instance and running

The rhs-function
def my_ode (t , u ) :
return - 3 * u
my_ode . name = " u ’= - 3 * u "

Creating an instance, solving and plotting:

rK = RungeKutta4 ( my_ode , -5 , 0 )
rK . run ()
rK . plot ()
ee = ExplixitEuler ( my_ode , -5 , 0 )
ee . run ()
ee . plot ()

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 231 / 307
Initial Value Problem

Linear Systems

More general rhs functions:

def make_lin ( A ) :
if np . isscalar ( A ) :
def lin (t , u ) :
return A * u
else :
def lin (t , u ) :
return dot (A , u )
lin . exact = ....
lin . name = ....
return lin

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 232 / 307
Initial Value Problem

Computational Mathematics with Python

Broadcasting

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 233 / 307
Performance Broadcasting

36 Performance

37 Broadcasting
Mathematical View
Broadcasting Arrays

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 234 / 307
Performance Broadcasting

36 Performance

37 Broadcasting
Mathematical View
Broadcasting Arrays

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 235 / 307
Performance Broadcasting

What is slow?

Any repeated task in python is slow.

Slow:
for i in range ( n ) : Fast:
for j in range ( n ) : v = dot (A , x )
v . append ( A [i , j ] *
x[j])
Faster because, under the
Slow because interpreted hood, it is compiled code.
code

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 236 / 307
Performance Broadcasting

Slow/Fast

Relatively slow: Fast, numpy operations:

I for loops I dot
I append for lists I array operations
I generators I solve
I comprehensive lists/generators I etc.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 237 / 307
Performance Broadcasting

Vectorization

To improve the performance, one has often to vectorize, i.e., to replace for
loops by numpy functions.
# v is a vector
# we want to shift its values by 5

# slow :
for i in range ( len ( v )) :
v[i] += 5

# fast :
v += 5

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 238 / 307
Performance Broadcasting

36 Performance

37 Broadcasting
Mathematical View
Broadcasting Arrays

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 239 / 307
Performance Broadcasting

Why Broadcasting

Take two functions f (x), g(x) and create a new function

F (t, x) = f (x) + g(t)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 240 / 307
Performance Broadcasting

Why Broadcasting

Take two functions f (x), g(x) and create a new function

F (t, x) = f (x) + g(t)

You have actually broadcast the function f and g as follows:

f → f̄ (t, x) = f (x)

g → ḡ(t, x) = g(t)
and now F = f̄ + ḡ.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 240 / 307
Performance Broadcasting

Simple Example
One of the simplest example of broadcasting in mathematics is constants
being broadcast to function.
If C is a scalar one often writes:

f := sin +C

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 241 / 307
Performance Broadcasting

Simple Example
One of the simplest example of broadcasting in mathematics is constants
being broadcast to function.
If C is a scalar one often writes:

f := sin +C

this is an abuse of notation since one should not be able to add functions
and constants. Constants are however implicitly broadcast to functions:

C̄ (x) := C ∀x

and now
f = sin +C̄

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 241 / 307
Performance Broadcasting

Mechanism

A mechanism to achieve broadcasting automatically is as follows:

1. I First reshape the function g to g̃(t, 0) := g(t)
I Then reshape f to f̃ (0, x) := f (x)
Now both f and g take two arguments.
2. I Extend f to f̄ (t, x) := f̃ (0, x)
I Extend g to ḡ(t, x) := g̃(t, 0)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 242 / 307
Performance Broadcasting

Several Variables

This construction extends readily to functions of several variables. Say you

want to construct
F (x, y, z) = f (x, y) ∗ g(z)

I f is broadcast to f̄ (x, y, z) = f (x, y)

I g is broadcast to ḡ(x, y, z) := g(z)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 243 / 307
Performance Broadcasting

Conventions

By convention a function is automatically reshaped by adding zeros on the

left:
g(x, y) −→ g(0, 0, . . . , 0, x, y)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 244 / 307
Performance Broadcasting

Array Broadcasting

Arrays are just particular cases of functions.

Broadcasting is done automatically in numpy.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 245 / 307
Performance Broadcasting

enthought ®

Array Broadcasting
Example
4x3 4x3
0 1 2 0 0 0 0 1 2 0 0 0
0 1 2 10 10 10 0 1 2 10 10 10
+ = + =

Adding a0 matrix
1 2 20 20 20
of shape 0 1 2
(4,3) and a matrix of20 size
20 20
(1,3). The second
0 1 2 30 30 30 0 1 2 30 30 30
matrix is extended to the shape (4,3).
4x3 3
0 0 0 0 1 2 0 0 0 0 1 2
0 1 2
10 10 10 10 10 10 0 1 2
+ = + = 10 11 12
20 20 20 20 20 20 0 1 2
20 21 22
30 30 30 30 30 30 0 1 2
stretch 30 31 32

4x1 3
0 and T. Oliphant) 0
(Source: E. Jones 1 2 0 0 0 0 1 2
10 10 10 10 0 1 2
+ = + =
20 20 20 20 0 1 2
30 30 30 30 0 1 2

stretch stretch
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 246 / 307
Performance Broadcasting

The Broadcasting Problem

Problem
Given an array of shape s1 broadcast it to the shape s2

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 247 / 307
Performance Broadcasting

The Broadcasting Problem

Problem
Given an array of shape s1 broadcast it to the shape s2
This is done in two steps:
1. If the shape s1 is shorter than the shape s2 then ones are added on
the left of the shape s1 . This is a reshaping.
2. When the shapes have the same length the array is extended to
match the shape s2 (if possible).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 247 / 307
Performance Broadcasting

The Broadcasting Problem

Problem
Given an array of shape s1 broadcast it to the shape s2
This is done in two steps:
1. If the shape s1 is shorter than the shape s2 then ones are added on
the left of the shape s1 . This is a reshaping.
2. When the shapes have the same length the array is extended to
match the shape s2 (if possible).
Example
You want to add a vector of shape (3,) to a matrix of shape (4,3). The
vector needs be broadcast.
1. (3,) → (1,3)
2. (1,3) → (4,3)
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 247 / 307
Performance Broadcasting

Example
v is a vector of length n
It is to be broadcast to the shape (m, n).
1. v is automatically reshaped to the shape (1, n)
2. v is extended to (m, n)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 248 / 307
Performance Broadcasting

Example
v is a vector of length n
It is to be broadcast to the shape (m, n).
1. v is automatically reshaped to the shape (1, n)
2. v is extended to (m, n)
M = array ( [ [ 11 , 12 , 13 , 14 ] ,
[ 21 , 22 , 23 , 24 ] ,
[ 31 , 32 , 33 , 34 ] ] )
v = array ( [ 100 , 200 , 300 , 400 ] )
M + v # works directly

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 248 / 307
Performance Broadcasting

Example
v is a vector of length n
It is to be broadcast to the shape (m, n).
1. v is automatically reshaped to the shape (1, n)
2. v is extended to (m, n)
M = array ( [ [ 11 , 12 , 13 , 14 ] ,
[ 21 , 22 , 23 , 24 ] ,
[ 31 , 32 , 33 , 34 ] ] )
v = array ( [ 100 , 200 , 300 , 400 ] )
M + v # works directly

Result:  
111 212 313 414
121 222 323 424
 
131 232 333 434
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 248 / 307
 Broadcasting Rules
Performance Broadcasting

The trailing axes of both arrays must either be 1 or have the same
Shape Mismatch
size for broadcasting to occur. Otherwise, a “ValueError:
framesa vector
To broadcast are not aligned”
of length exception
n to the shape (n, m)isthe
thrown.
automatic
reshaping will not work.
mismatch!
4x3 4
0 0 0 0 1 2 3
10 10 10
+ =
20 20 20
30 30 30

(Source: E. Jones and T. Oliphant)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 249 / 307
 Broadcasting Rules
Performance Broadcasting

The trailing axes of both arrays must either be 1 or have the same
Shape Mismatch
size for broadcasting to occur. Otherwise, a “ValueError:
framesa vector
To broadcast are not aligned”
of length exception
n to the shape (n, m)isthe
thrown.
automatic
reshaping will not work.
mismatch!
4x3 4
0 0 0 0 1 2 3
10 10 10
+ =
20 20 20
30 30 30

(Source: E. Jones and T. Oliphant)

1. one manually reshapes v to the shape (n, 1)

2. v is ready to be extended to (n, m)
Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 249 / 307
Performance Broadcasting

Shape Mismatch Example

M = array ( [ [ 11 , 12 , 13 , 14 ] ,
[ 21 , 22 , 23 , 24 ] ,
[ 31 , 32 , 33 , 34 ] ] )
v = array ( [ 100 , 200 , 300 ] )
M + v # error !
M + v . reshape ( -1 , 1 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 250 / 307
Performance Broadcasting

Shape Mismatch Example

M = array ( [ [ 11 , 12 , 13 , 14 ] ,
[ 21 , 22 , 23 , 24 ] ,
[ 31 , 32 , 33 , 34 ] ] )
v = array ( [ 100 , 200 , 300 ] )
M + v # error !
M + v . reshape ( -1 , 1 )

Result:  
111 112 113 114
221 222 223 224
 
331 332 333 334

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 250 / 307
Performance Broadcasting

Typical Examples

I M + C Matrix plus constant

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 251 / 307
Performance Broadcasting

Typical Examples

I M + C Matrix plus constant

I Multiply all the columns of a matrix by column dependent coefficients
in a vector V
M * V

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 251 / 307
Performance Broadcasting

Typical Examples

I M + C Matrix plus constant

I Multiply all the columns of a matrix by column dependent coefficients
in a vector V
M * V

I Multiply the rows of a matrix by coefficients in a vector V :

M * V . reshape ( -1 , 1 )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 251 / 307
Performance Broadcasting

Typical Examples

I M + C Matrix plus constant

I Multiply all the columns of a matrix by column dependent coefficients
in a vector V
M * V

I Multiply the rows of a matrix by coefficients in a vector V :

M * V . reshape ( -1 , 1 )

I Compute the tensor product Mi,j = Vi Wj

M = V . reshape ( -1 , 1 ) * W

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 251 / 307
Performance Broadcasting

Computational Mathematics with Python

Modules

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 252 / 307
Modules Dictionaries and Arguments Tests

38 Modules
Imports and run
__main__ variable

39 Dictionaries and Arguments

Dictionaries
Function Arguments

40 Tests
What are tests
nosetest

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 253 / 307
Modules Dictionaries and Arguments Tests

38 Modules
Imports and run
__main__ variable

39 Dictionaries and Arguments

Dictionaries
Function Arguments

40 Tests
What are tests
nosetest

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 254 / 307
Modules Dictionaries and Arguments Tests

Import

To load the contents of a file you may use import but the file is loaded
only once.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 255 / 307
Modules Dictionaries and Arguments Tests

Import

To load the contents of a file you may use import but the file is loaded
only once.
The various syntax are
I from module import something
I from module import *
I import module

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 255 / 307
Modules Dictionaries and Arguments Tests

IPython’s run command

I IPython has a special command named run which executes a file as if

you ran it directly in python.
I This means that the file is executed independently of what is already
defined in IPython.
I This is the recommended way to execute files.
You must import all you need in the executed file.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 256 / 307
Modules Dictionaries and Arguments Tests

Typical Example

from numpy import array

...
a = array (...)

And in IPython: run file

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 257 / 307
Modules Dictionaries and Arguments Tests

Typical Example

from numpy import array

...
a = array (...)

And in IPython: run file

Everything that is defined in the file is then imported in the IPython
workspace.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 257 / 307
Modules Dictionaries and Arguments Tests

Where are the commands?

Where are the numpy and scipy commands?

numpy array, arange, linspace, vstack, hstack, dot, eye,
zeros
numpy.linalg solve, lstsq, eig, det
pylab plot, legend, cla
scipy.integrate quad
copy copy, deepcopy

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 258 / 307
Modules Dictionaries and Arguments Tests

__name__ and __main__

In a given module the special variable __name__ is defined to the name of

the current module.
In the command line (in IPython) this variable is set to "__main__" which
allows the following trick:
# module
import ...

class ...

if __name__ = = " __main__ " :

# perform some tests here

The tests will be run only when the file is directly run, not when it is
imported.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 259 / 307
Modules Dictionaries and Arguments Tests

38 Modules
Imports and run
__main__ variable

39 Dictionaries and Arguments

Dictionaries
Function Arguments

40 Tests
What are tests
nosetest

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 260 / 307
Modules Dictionaries and Arguments Tests

Dictionaries

A dictionary is a structure similar to lists but where keys are (usually)

strings.
One accesses dictionaries with square brackets.
homework_passed = { ’ Svensson ’: True , ’ Karlsson ’: False }

homework_passed [ ’ Svensson ’] # True

# changing a value :
homework_passed [ ’ Svensson ’] = False
# deleting an item
del homework_passed [ ’ Svensson ’]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 261 / 307
Modules Dictionaries and Arguments Tests

Looping through Dictionaries

A dictionary is an object with the following useful mehods: keys, items,
values.
By default a dictionary is considered as a list of keys:
for key in homework_passed :
print " % s % s " % ( key , homework_passed [ key ] )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 262 / 307
Modules Dictionaries and Arguments Tests

Looping through Dictionaries

A dictionary is an object with the following useful mehods: keys, items,
values.
By default a dictionary is considered as a list of keys:
for key in homework_passed :
print " % s % s " % ( key , homework_passed [ key ] )

One may also use items to loop through keys and values:
for key , value in homework_passed . items () :
print " % s % s " % ( key , value )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 262 / 307
Modules Dictionaries and Arguments Tests

Looping through Dictionaries

A dictionary is an object with the following useful mehods: keys, items,
values.
By default a dictionary is considered as a list of keys:
for key in homework_passed :
print " % s % s " % ( key , homework_passed [ key ] )

One may also use items to loop through keys and values:
for key , value in homework_passed . items () :
print " % s % s " % ( key , value )

One may use the keys and values methods to copy or change the
dictionary:
dict_items = zip ( homework_passed . keys () ,
homework_passed . values ())
other_dict = dict ( dict_items ) # same dictionary

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 262 / 307
Modules Dictionaries and Arguments Tests

Function Argument List

Take the function newton

def newton (f , x0 , max_iter = 20 ) :
...

Recall that this function can be called by using

I positional arguments only: zero=newton(cos,.2,30)
I keyword arguments only:
zero=newton(f=cos,max_iter=30,x0=.2)
I or a mixed form: zero=newton(cos,maxiter=30,x0=.2)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 263 / 307
Modules Dictionaries and Arguments Tests

Function Argument List

def newton (f , x0 , max_iter = 20 ) :

...

Say that we are given a list with the arguments prepared:

L = [ cos , . 2 ]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 264 / 307
Modules Dictionaries and Arguments Tests

Function Argument List

def newton (f , x0 , max_iter = 20 ) :

...

Say that we are given a list with the arguments prepared:

L = [ cos , . 2 ]

One may transform this list into an argument list for positional arguments
with the single star operator:
newton ( L [ 0 ] , L [ 1 ] ) # ok but cumbersome
newton ( * L ) # does the same thing

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 264 / 307
Modules Dictionaries and Arguments Tests

Function Argument Dictionary

Take the function newton

def newton (f , x0 , max_iter = 20 ) :
...

Similarly, if one is given a dictionary one may use for keyword arguments
the double star operator:
D = { ’ x0 ’: .3 , ’f ’: cos }

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 265 / 307
Modules Dictionaries and Arguments Tests

Function Argument Dictionary

Take the function newton

def newton (f , x0 , max_iter = 20 ) :
...

Similarly, if one is given a dictionary one may use for keyword arguments
the double star operator:
D = { ’ x0 ’: .3 , ’f ’: cos }

One may call

newton ( D [ ’f ’] , D [ ’ x0 ’] ) # ok but cumbersome
newton ( * * D ) # better

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 265 / 307
Modules Dictionaries and Arguments Tests

Passing arguments

Also in the definition of functions you might find these constructs. This is
often used to pass parameters through a function
def letssee (f , x , * args , * * keywords ) :
return f (x , * args , * * keywords )

def look (x ,y ,z , u ) :
print y , z
print u
return x * * 2

A call gives
L = [1 , 2 ] 1 2
D = { ’u ’: 15 } 15
letssee ( look ,3 , *L , * * D ) Out [ 35 ] : 9

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 266 / 307
Modules Dictionaries and Arguments Tests

38 Modules
Imports and run
__main__ variable

39 Dictionaries and Arguments

Dictionaries
Function Arguments

40 Tests
What are tests
nosetest

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 267 / 307
Modules Dictionaries and Arguments Tests

Why Tests?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 268 / 307
Modules Dictionaries and Arguments Tests

Why Tests?

I Because you do them anyway

I Because it will keep your code alive.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 268 / 307
Modules Dictionaries and Arguments Tests

Automated Tests

Automated tests
I ensure a constant (high) quality standard of your code
I serve as a documentation of the use of your code
I document the test cases → test protocol

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 269 / 307
Modules Dictionaries and Arguments Tests

Example

A matrix property:

Two matrices A,B are called similar, if there excists a matrix S, such that
B = S −1 AS. A and B have the same eigenvalues.

A related piece of code:

def gen_similar (A , S ) :
return dot ( dot ( inv ( S ) , A ) , S )

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 270 / 307
Modules Dictionaries and Arguments Tests

Example (Cont.)

A test of this program can be

class test_sim ( object ) :
tol = 1 . e - 9
A = array ( [1 ,2 ,3 , 4 ] ). reshape (2 , 2 )
S = array ( [1 , -2 ,7 , - 4 ] ). reshape (2 , 2 )
def near_real ( self ,a , b ) :
return abs ( a - b ) < self . tol

def test_eig ( self ) :

"""
Check to see if similar
"""
B = gen_similar ( self .A , self . S )
assert self . near_real ( norm ( eig ( B ) [ 0 ] ) , norm ( eig ( self . A ) [

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 271 / 307
Modules Dictionaries and Arguments Tests

Example (Cont.)
even reaction on wrong input should be checked
class test_sim ( object ) :
...
S0 = zeros (( 2 , 2 ))
...

def test_eig_sing ( self ) :

"""
We check correct error raising
"""
try :
B = gen_similar ( self .A , self . S0 )
flag = False
except LinAlgError :
flag = True
assert flag

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 272 / 307
Modules Dictionaries and Arguments Tests

nosetests

There is a nice command which performs all tests automatically:

claus@Claus-Home:~/gen_sim$ nosetests gen_sim.py

..
--------------------------------------------------
Ran 2 tests in 0.390s

OK

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 273 / 307
Modules Dictionaries and Arguments Tests

Test Cases

You will want to put your tests together:

class Test_Similar ( object ) :
def setUp ( self ) :
# define the matrix here
self . A = < some matrix >
self . A_similar = gen_similar (A , S )

def test_eigenvalue ( self ) :

assert a rr ay _a lm ost_equal ( eigvals ( self . A ) , eigvals ( self . A_s

def test_singular ( self ) :

...

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 274 / 307
Modules Dictionaries and Arguments Tests

Test discovering

nosetest will discover all your test automatically provided the name of
the file/class/function starts with test:

nosetests test_file.py:Test_Similar.test_eigenvalue
nosetests test_file.py:Test_Similar
nosetests test_file.py

or even:
nosetests

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 275 / 307
Modules Dictionaries and Arguments Tests

Testing Tools

I numpy.testing.assert_array_almost_equal is very handy; use it

to compare vectors or even scalars:
expected = array ( [ 1 . , 2 . , 3 ] ) # or a scalar
computed = my_algorithm ()
a s s e r t _ a r r a y _ a l m o s t _ e q u a l ( computed , expected )

I nose has a number of assert functions: nose.tools.assert_true,

nose.tools.assert_raises, etc.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 276 / 307
Modules Dictionaries and Arguments Tests

Advice

When you develop, do not test your code in Ipython.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 277 / 307
Modules Dictionaries and Arguments Tests

Advice

When you develop, do not test your code in Ipython.

Write your tests in a file instead.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 277 / 307
Modules Dictionaries and Arguments Tests

Computational Mathematics with Python

Matlab and Concluding Remarks

Olivier Verdier and Claus Führer

Spring 2009

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 278 / 307
Matlab for Python users Matlab What we didn’t do in this course

41 Matlab for Python users

42 Matlab
Syntax
Lists, Dictionaries, Arrays
Linear Algebra
Functions
Environment

43 What we didn’t do in this course

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 279 / 307
Matlab for Python users Matlab What we didn’t do in this course

Matlab

Matlab has become the standard

language to teach and experiment
with scientific computing, especially
in Sweden and the US.
Matlab is a commerical product.
There are free clones: SciLab and
Octave.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 280 / 307
Matlab for Python users Matlab What we didn’t do in this course

Matlab

Matlab has become the standard

language to teach and experiment
with scientific computing, especially
in Sweden and the US.
Matlab is a commerical product.
There are free clones: SciLab and
Octave.
We will now focus on the difference
between Matlab and Python.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 280 / 307
Matlab for Python users Matlab What we didn’t do in this course

41 Matlab for Python users

42 Matlab
Syntax
Lists, Dictionaries, Arrays
Linear Algebra
Functions
Environment

43 What we didn’t do in this course

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 281 / 307
Matlab for Python users Matlab What we didn’t do in this course

I comments with %
% matlab comment # python comment

I strings with single quotes

I blocks require no identation but are ended with end

if z == 0 if z = = 0 :
dosomething dosomething
else else :
dosomethingelse dosomethingelse
end #

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 282 / 307
Matlab for Python users Matlab What we didn’t do in this course

Type

I There is no direct equivalent of Python types in Matlab.

I One may however use whos which gives information about a given
variable.
I All the numbers are floats by default (with or without the decimal
point).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 283 / 307
Matlab for Python users Matlab What we didn’t do in this course

Operators

The usual operators *, +, /, ==, etc. work as in Python.

The logical operators are different, as shown in the examples below.

2^3 % 8 2**3 # 8
2 ~= 3 % 1 2 ! = 3 # True

1 && 0 % 0 True and False # False

1 || 0 % 1 True or False # True
~1 % 0 not True # False

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 284 / 307
Matlab for Python users Matlab What we didn’t do in this course

Printing out

Matlab will always display (echo) what you defined unless you end the
assignment with a semicolon

% inside a program # inside a program

a = 10 % prints a =10 a = 10 # silent
a = 10; % silent print a # prints 10

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 285 / 307
Matlab for Python users Matlab What we didn’t do in this course

Operator differences

I No multiple comparison

4 > 3 > 2 % false ! 4 > 3 > 2 # True

I No increment operators (+=, *=,...)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 286 / 307
Matlab for Python users Matlab What we didn’t do in this course

Lists, Dictionaries, Arrays

I Python Lists correspond to MATLAB cell arrays

L ={[1 ,2 ,3 ,4] , ’ hello }

L = [ [1 ,2 ,3 , 4 ] , ’ hello ’]
L {1} % [1 ,2 ,3 ,4]

I Python dictionaries correspond to MATLAB structure arrays

D = { ’ key1 ’: 15 . , ’ key2 ’: - 25 } D . key1 =15

D [ ’ key1 ’] # 15 D . key2 = -25

I Python/numpy arrays correspond to MATLAB arrays (to some

extend):

A = array ( [ [1 ,2 , 3 ] ,[4 ,5 , 6 ] ] ) A =[1 ,2 ,3;4 ,5 ,6]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 287 / 307
Matlab for Python users Matlab What we didn’t do in this course

Creating matrices

The syntax is
M = [1 2 3; 4 5 6]
size ( M ) % 2 3

Blank or comma separates columns.

Semicolon or line break separates rows.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 288 / 307
Matlab for Python users Matlab What we didn’t do in this course

Creating matrices

The syntax is
M = [1 2 3; 4 5 6]
size ( M ) % 2 3

Blank or comma separates columns.

Semicolon or line break separates rows. Be careful, this role of the blank is
ambiguous:
M = [1 - 1] % what will this do ?
M = [1 -1] % and this ?

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 288 / 307
Matlab for Python users Matlab What we didn’t do in this course

Creating matrices

The syntax is
M = [1 2 3; 4 5 6]
size ( M ) % 2 3

Blank or comma separates columns.

Semicolon or line break separates rows. Be careful, this role of the blank is
ambiguous:
M = [1 - 1] % what will this do ?
M = [1 -1] % and this ?

Since you may not create vectors you must store vectors in row or column
matrices.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 288 / 307
Matlab for Python users Matlab What we didn’t do in this course

Transpose

The single quote is used for transposing matrices (and for delimiting
strings).
s = ’ abc ’
v = [1:3] % row matrix
v ’ % column matrix

Transpose is often used to transform row matrices to column matrices and

vice versa.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 289 / 307
Matlab for Python users Matlab What we didn’t do in this course

Shapes

I Shapes are not tuple.

I The shape is called size

size ( M ) shape ( M ) # or M . shape

size ( M )(1) % error ! shape ( M ) [ 0 ] # ok
( size ( M ))(1) % error ! ( shape ( M )) [ 0 ] # ok
size (M ,1) % ok

s = size ( M ) s = shape ( M )
s (1) % ok s [ 0 ] # ok

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 290 / 307
Matlab for Python users Matlab What we didn’t do in this course

Operators

Operators on matrices are linear algebra operators. Componentwise

operators are prefixed with a dot.

V = [1 2 3] V = array ( [1 ,2 , 3 ] )
W = [3 4 5] W = array ( [3 ,4 , 5 ] )
V * W % error !
V .* W % ok V * W # ok
V / W % error !
V ./ W % ok V / W # ok
V ^ W % error !
V .^ W % ok V * * W # ok

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 291 / 307
Matlab for Python users Matlab What we didn’t do in this course

Slices

I Slices are closed intervals

I Slice indices cannot be out of bound.
I Last index is denoted by end
I There are no half slices

V = [1 2 3] V = array ( [1 , 2 , 3 ] )
V (2: end ) - V (1: end -1) v[1:] - v[:-1]

V (2:) % error ! V [ 1 : ] # ok
V (2:100) % error ! V [ 1 : 100 ] # same as V [ 1 :]

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 292 / 307
Matlab for Python users Matlab What we didn’t do in this course

for loops

I for i=M goes through the columns of the matrix M

I 1:n creates a row matrix (similar to range in python)
I This allows the often used syntax:
for i =1: N

I Careful to create a column matrix!

% wrong :
c = 1:3 ’ % 3 is transposed first !

% right :
c = (1:3) ’

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 293 / 307
Matlab for Python users Matlab What we didn’t do in this course

Functions in Matlab

There are two kinds of functions in Matlab

File Functions
I automatically reloaded
I available from everywhere

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 294 / 307
Matlab for Python users Matlab What we didn’t do in this course

Functions in Matlab

There are two kinds of functions in Matlab

File Functions
I automatically reloaded
I available from everywhere

Local Functions
I available only in the file they are defined
I must be declared after the main file function

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 294 / 307
Matlab for Python users Matlab What we didn’t do in this course

Functions

For all matlab functions the syntax is:

function [ res ] = fname ( args ) def function_name ( args ) :

For file functions the function name is not used; what is used is the name
of the file instead (similar to shell scripting)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 295 / 307
Matlab for Python users Matlab What we didn’t do in this course

Functions

For all matlab functions the syntax is:

function [ res ] = fname ( args ) def function_name ( args ) :

For file functions the function name is not used; what is used is the name
of the file instead (similar to shell scripting)

Functions without arguments do not require parenthesis

rand () % random number rand () # random number

rand % same thing rand # the function object

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 295 / 307
Matlab for Python users Matlab What we didn’t do in this course

Single Return Values

function ret = f ()

To return the value one simply sets the variable ret to the value we want
to return.
ret = 2 % the function returns 2

Notice that the execution will continue after this, unless you write return
(but not return 2)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 296 / 307
Matlab for Python users Matlab What we didn’t do in this course

Multiple Return Values I

To return multiple values you may use structures

ret = {2 , 3}

Used as:
r = f
r {1} % 2
r {2} % 3
% but
f {1} % error !

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 297 / 307
Matlab for Python users Matlab What we didn’t do in this course

Multiple Return Values II

But the most common way is to use the special syntax:

function [ ret1 , ret2 ] = f ()

Note that this is not like returning a tuple. It is used as follows

a = f () % first value only !!!
a , b = f () % first and second values

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 298 / 307
Matlab for Python users Matlab What we didn’t do in this course

feval and @

Functions are not objects so you must use a special symbol to pass it
around.
function [ res ] = fname ( x )
...
end

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 299 / 307
Matlab for Python users Matlab What we didn’t do in this course

feval and @

Functions are not objects so you must use a special symbol to pass it
around.
function [ res ] = fname ( x )
...
end

You can’t pass this function using fname because fname executes the
function! You have to prevent the execution using the @ operator.
f = @fname
% now how to execute the function ?
f(x)

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 299 / 307
Matlab for Python users Matlab What we didn’t do in this course

Functions: differences with python

I Fixed return value(s)

I Multiple return values are not tuples
I File function name is not used
I Functions may not be defined in interactive mode
I Each function must be in its own file if it is to be publicly used
I Functions are not objects
I No default arguments
I No named arguments
I No * nor ** argument operators
I No docstring

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 300 / 307
Matlab for Python users Matlab What we didn’t do in this course

Mathematical Functions

Most of the mathematical functions, for computing and plotting, have the
same name in matlab and scipy. Examples are:
I plot, legend, grid, pcolor,. . .
I eig, eigs, svd, qr, . . .
I rand, . . .
I sin, cos, tan,. . .
I det, norm, . . .
I linspace, logspace, roll, . . .
I real, imag, . . .
I sum, max, min, mean, cumsum, cumprod, diff, . . .

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 301 / 307
Matlab for Python users Matlab What we didn’t do in this course

Variables

I There is no notion of reference in matlab. Variables are always copied

when reassigned.

x = [1 2 3; 4 5 6] x = array ( [ [1 ,2 , 3 ] ,[4 ,5 , 6 ] ] )
y = x y = x
y (1 ,1) = 0 y [0 , 0 ] = 0 # x is changed
x % unchanged x # [[ 0 ,2 , 3 ] ,[4 ,5 , 6 ]]

This is in particular important for function calls.

I Objects are not common (this feature was introduced recently in
Matlab).

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 302 / 307
Matlab for Python users Matlab What we didn’t do in this course

Modules

There are no modules nor namespaces in matlab. Everything that is in the

search path is available.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 303 / 307
Matlab for Python users Matlab What we didn’t do in this course

Exception handling

function a = divideme (b , c )
InfEx = MException ( ’ MATLAB : Inf ’ , ’ division by zero ’ );
a=b/c;
if a == inf def divideme (b , c ) :
throw ( InfEx ) return b / c
end try :
a = divideme (1 , 0 )
try except :
print ’ Division by zero ’
a = divideme (1 ,0)
catch InfEx
disp ( InfEx . message )
end

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 304 / 307
Matlab for Python users Matlab What we didn’t do in this course

License

I Matlab is neither free nor open source

I It requires a license which must be purchased
I The cheapest license are network ones, which means that you won’t
be able to use Matlab without internet access.

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 305 / 307
Matlab for Python users Matlab What we didn’t do in this course

41 Matlab for Python users

42 Matlab
Syntax
Lists, Dictionaries, Arrays
Linear Algebra
Functions
Environment

43 What we didn’t do in this course

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 306 / 307
Matlab for Python users Matlab What we didn’t do in this course

I Graphical Interface (GUIs)

I 3D plots
I Interface with C or Fortran code
I Python for controlling system commands
I ....

Olivier Verdier and Claus Führer Computational Mathematics with Python Spring 2009 307 / 307