Python Basics

March 22, 2018

Contents
Introduction 4

Before Numpy 4
The Environment and Choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Launching the Jupyter Notebook . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Launching the IPython QT Console with the Canopy Editor . . . . . . . . . . 6
Launching the IPython QT Console From the Terminal . . . . . . . . . . . . . 8
Launching the Native Python Console From the Terminal . . . . . . . . . . . 9
Ending Your Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Data Types and Simple Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Hello World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
float, complex, long, int, str, and boolean . . . . . . . . . . . . . . . . . . . 10
Data Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Tuples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Dictionaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Formatting Strings and Gathering User Input . . . . . . . . . . . . . . . . . . . . . . . 21
Formatting Strings and Printing . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Gathering User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Flow Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
If, elif, and else . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
For Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
While Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
The Statements break and continue . . . . . . . . . . . . . . . . . . . . . . . . 29
Exceptions: try, except, and finally Blocks . . . . . . . . . . . . . . . . . . . . . . . 30
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Object Oriented Python: Writing a Class . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Writing a Simple Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1
 After Numpy 33
NumPy Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
The N-Dimensional Array and Available Types . . . . . . . . . . . . . . . . . 33
Array Creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Working With Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Graphics and More with Matplotlib 40

Signals and Systems Tools and Examples 40

The Scipy Module scipy.signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Using scikit-dsp-comm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
More Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
A Simple DSP Class Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
The class Code Base . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Lowpass and Bandpass Examples . . . . . . . . . . . . . . . . . . . . . . . . . 47

References 49

In [1]: %pylab inline

#%matplotlib qt
#from __future__ import division # use so 1/2 = 0.5, etc.
import sk_dsp_comm.sigsys as ss
import scipy.signal as signal
from IPython.display import Image, SVG

Populating the interactive namespace from numpy and matplotlib

In [2]: %config InlineBackend.figure_formats=['svg'] # SVG inline viewing

#%config InlineBackend.figure_formats=['pdf'] # render pdf figs for LaTeX

2
In [34]: print('Hello World')

Hello World

In [35]: 2*pi

Out[35]: 6.283185307179586

In [36]: arange(0,1,.1)

Out[36]: array([ 0. , 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9])

3
Introduction
This tutorial is structured around the idea that you want to get up and running with Python using
PyLab as quickly as possible. The first question I asked my myself before I started using PyLab
was why consider Python? What makes it a vialble alternative to other languages available for sci-
entific and engineering computations and simulations? OK, everyone has favorites, and presently
MATLAB is very popular in the signals and system community. Is there a need to change? This
is a debate that lies outside the scope of this tutorial, but the ability to use open-source tools that
work really, really well is very compelling.
To answer the first question, why consider Python, I can say:

1. The NumPy library

2. combined with Matplotlib

3. The SciPy library of modules, particularly signal, provides reasonable suppost for signals
and systems work. Additional libraries of modules are also available

Before Numpy
I have been saying a lot about using Python with Numpy as a means to do scientific and engineer-
ing analysis, simulation, and visualization. The fact of the matter is, Python is a good language
for doing many other things outside the computational realm.
Numpy plus Scipy are key elements to the attractiveness of using Python, but before getting
too carried away with the great scientific computing abiliies of the language, you should learn
some basics of the language. This way you will feel more comfortable at coding and debugging.
Before exploring the core language, I will spend time going over the environment and various
choices.

The Environment and Choices

How you choose to work with Python is up to you. I do have some strong suggestions. But
first I want to review four options in order of most recommended to least recommended. My
recommendations assume you are just starting out with Python, so I have a bias towards the
Jupyter notebook.
The first thing you want to do is get a version of Python with scientific support included. When
this notebook was first created I was using Canopy, but now my preference is to use Anaconda.
To learn more about the Jupyter notebook and its furture see Jupyter.

Launching the Jupyter Notebook

Regardless of the operating system, Windows, Mac OS, or Linux, you want to get a terminal
window open. It is best if the terminal window is opened at the top level of your user account,
so you will be able to navigate to any folder of interest. Note: In Windows 10x I recoimment the
use of powershell. This is done by clicking the file menu from the file manager and then selecting
powershell. It turns out with the notebook interface you can easily navigate to a location interest
and then launch an existing notebook or create a new notebook.

In [39]: Image('Python_Basics_files/LaunchNotebook2.png',width='90%')

4
 Out[39]:

From the above you can see that the notebook is all set. Note that the first cell is only relevant
if you intend to render your notebook to pdf using the LaTeX backend. This requires that you

5
install Pandoc and then an appropriate install of the TeX/LaTeX type setting system. The Pandoc
Web Site provides details.
The second cell issues commands to fine tune the configuration of the notebook. The first line
makes sure the workspace is populated with %pylab, which gives you full access to NumPy and
Matplotlib without having to type the module name or namespace. Note: commands that begin
with % are known as IPython magics, which in general allow you to perform OS operations outside
the default (see option four) Python console. The option inline directs all plots to show up right in
the notebook. If you prefer pop-up plots, enable the second line. The resolution of the embedded
png plots can be controlled using the third line.
The fifth line, if uncommented and run as a magic (put % at start) with change the render
mode from png to pdf. This will then result in a link to plots that opens them in a pdf viewer. For
LaTeX to pdf rendering, this will create crisp vector graphics. I recommend using this only when
you get ready to export a notebook to pdf. You will have to use Run All from the Cell menu to
convert all graphics to pdf and then switch back later to again have regular inline plots.
The two import lines just bring my ssd (Signals and Systems for Dummies ssd.py module into
the workspace). Note: for this to be sucessful ssd.py must be in the same folder as the notebook
you are working from. Once you import a module you can navigate to another location in your file
system. By the way, IPython magics make general OS path manipulation a breeze. Some of then
don’t even require that you forst type %. You do need to know basic Linux/Unix OS commends. I
show you a few examples below:

In [4]: pwd # check your path

Out[4]: '/Users/markwickert/Documents/Courses/Tablet/Python_Basics_saved'

In [5]: # Move up one level

%cd ..

/Users/markwickert/Documents/Courses/Tablet

In [3]: %ls

Python Basics.aux* Python Basics.tex*

Python Basics.ipynb* Python Basics.toc*
Python Basics.log* Python Basics_files/
Python Basics.out* Python_Basics_fig1.pdf*
Python Basics.pdf* Python_Basics_fig1.png*
Python Basics.synctex.gz* Python_Basics_figs.graffle/

If you are reading the present document in pdf format, you should consider downloading the
notebook version so you can follow along with interactive calculations and experiments, as you
learn Python Basics.
Moving on to the QT console...

Launching the IPython QT Console with the Canopy Editor

The second and third options are actually closely related. Both of these options have you working
at a commandline console, much like Octave or Matlab. All the features of IPython are available

6
at the QT console. The environment is very very nice. This is how I got started with Python, as
the notebook interface was not fully released.
OK, as the second choice for getting started with Python, I recommend the qtconsole in com-
bination with the Canopy. To bring up this environment simply launch the Canopy app (Win-
dows, Mac OS, or Linux), and then click the Editor button:

In [13]: Image('Python_Basics_files/LaunchCanopyEditor.png',width='90%')

Out[13]:

From the above figure you can see the top window is a code editor with Python syntax high-
lighting and other features. This is an Enthought (makers of Canopy) product. They plan to some-
day have a debugger included with the editor.
The lower window is the IPython console. By enabling the the Keep Directory Synced to Editor
option you can freely move around to import code modules from various locations and always
have the path in command console where you want it. By default when Canopy opens the editor
it starts the qtconsole woth pylab. It also by default has all plots going to pop-up windows.
The inline plots mode for the qtconsole is available, but not that great compared to the IPython
notebook.

7
Launching the IPython QT Console From the Terminal
If you prefer to use your own editor (many good choices out there) or if you are not using Canopy,
you can always start the qtconsole from the terminal. For quick calculations, where I don’t care to
have documentation created, this is my favorite interface. As I said earlier, I think starting with
the notebook is best, and it documents your work, which can be very useful for assignments. With
the qtconsole documentation is on your own. I wrote the Dummies book using this interface.
As with the notebook, you want to open the terminal already pointing to the folder where
your Python files of interest reside (yes you can always navigate using cd later). The next step
(Windows, Mac OS, Linux) is shown below:

In [15]: Image('Python_Basics_files/Launchqtconsole.png',width='90%')

Out[15]:

Note: The option --pylab is used to start up pylab as was done in earlier environments. Ev-
erything else you see in the above figure is very similar to the Canopy editor with qtconsole.

8
Launching the Native Python Console From the Terminal
In the beginning there was and there still is, the basic Python intaractive console. For a Python
beginner, wanting to learn how to do scientific/engineering calculatins in Python, this is the least
desireable way to go. Chances are you will have occasion to use this environment soon enough,
so no rush right now. If you decide to play with an embedded Linux device Rasberry Pi or Beagle-
Bone Black, and use Python to program it, this is where you will find yourself. Take this as good
news, as Python has many uses.
To launch the Python console start a terminal (command prompt) session as before and simply
type python:

In [14]: Image('Python_Basics_files/LaunchPythonTerm.png',width='90%')

Out[14]:

Ending Your Session

Not mentioned up to this point, is how you end a Python session.

• In the notebook you use the File menu and select Close and Halt
• On both the qtconsole and the traditional Python console you type exit()

Note: The () are required since exit() is a function that takes no arguments.
Now its finally time to start discussing some language details...

Data Types and Simple Calculations

Hello World
The classic first program in any language is Hello World. In Python this is accomplished quite
simply via:

In [85]: print('Hello Python World!')

9
Hello Python World!

As I roll through basics be aware that comments begin with # and multiline comments look
like

"""
A multiline comment
The second line
The third line
"""

if I need access to a particular Python module I need to import it first:

# Here I will import the SciPy signal processing module and

# give it a short name of signal
import scipy.signal as signal

More of discussion of import and modules will occur later. Until I start talking about NumPy
I will keep all he topics limited to what you can do with native Python. Note: If you need to
contiune a line you can use \ (backslash). You cannot break a string this way. You can also break
lines at commas.

float, complex, long, int, str, and boolean

The Native Python data types are actually rather few. When Numpy is brought in this changes,
but for now that is not the case.

Float In signals and systems work the float type is actually is actually a double as found in the
C language. This means it consumes 64 bits (8 bytes) of memory. A float is created by simply
including a decimal point in the number.

In [71]: a = 1.2
b = 4.603
a*b

Out[71]: 5.523599999999999

To be sure you can use the built-in function type(). To compare several calculation I will string
together several calls to type() with parenthesis and commas in between. This way I can display
the rults all on one line. Note: I have actually created a compount type known as a tuple. More
on tuples later.

In [87]: (type(a), type(2), type(2*a)) # note the upcasting to float

Out[87]: (float, int, float)

In [88]: type((type(a), type(2), type(2*a))) # still have to explain tuples

Out[88]: tuple

10
 The native operations available with float types are given in the following table. The table
order is from lowest to highest precedence.

In [20]: SVG('Python_Basics_files/Operations_table.svg')

Out[20]:

Native Type Standard Arithmetic Operators

Name Operator Quick
Example
>>> 3 + 4.5
Addition +
7.5
>>> 10 - 7.3
Subtraction -
2.7
>>> 3.4 * 39.1
Multiply *
132.94
/Note one number must be >>> 45.2/89.1
Division
a float in Python 2.7 0.5072951739618407
// In Python 3.3 or >>> 67//8 or 67/8
Integer Division
/In Python 2.7 (see note) 8
%Note with numpy you >>> 13 % 8
Remainder
typically usemod(a,b) 5
>>> 3**8
Exponentiation **
6561
Note:In Python 2.7 Python 3.3 division behavior is available if you make
your first import (be careful with this):
>>> from __future__ import division

A frustration with Python 2.7 (what I am currently using in this IPython notebook), is that
when do perform simple division such as

x = 6/7

thinking you will form a float, you instead get and integer. In the example above you get 0. In
Python 3.3 this is resolved. I am making a big deal about this because over and over again I get
tripped up.
So what can you do? In Python 2.7 I find it best to just remember to put a decimal point on one
of the two numbers when working with ratios of integers in math calculations. A hard conversion
to float is possible using the native function float(), e.g.,

x = 6/7.
# or
x = 6/float(7)

The best option now is to use Python 3.6 or alternatively import the 3.x division rule into
Python 2.7 as follows:

11
In [98]: from __future__ import division
(6/7,6//7) # display two results, again using a tuple

Out[98]: (0.8571428571428571, 0)

Note: It is strongly recommended that this import be placed before any other imports. Also,
with this division change, when you really want integer division you need to use //. More on the
int type coming up.

Complex Another standard type to Python is complex. For √signals and systems work, complex
plays a significant role. The constant 1j gives you access to −1, and allows you to readily form
complex quantities. In the following example I will again create a tuple just for the convenience
of displaying multiple results without using a formatted print statement.

In [19]: z1 = complex(1.,3) # z = complex(x,y) builds a complex type

z2 = 5. - 20j
(z1 + z2, z1 - z2, z1*z2, z1/z2)

Out[19]: ((6-17j), (-4+23j), (65-5j), (-0.12941176470588237+0.08235294117647059j))

The convenience of built-in complex arithmeic is very nice. I need to mention however, that
getting access to functions beyond the operators listed in the table above, requires the import of
specific code modules. The math and cmath bring in additional functions for float or real numbers
and complex numbers respectively. Don’t get too excited about jumping in to use these modules.
With NumPy, which will be talked about later, the use of math and cmath is taken care of for you.
AN with NumPy you will have full vector/matrix support. I just mention it here so you know it
does exist, and if for some strange case you don’t want to use NumPy, this is what you will have
work with.

Int and Long For integer math, indices, and loop counters, Python has the types int and long.
The int type is a signed integer similar to int in the C language. The size of int depends upon
the machine you are running on. If you import the sys module you can find out more information
about int for your machine:

In [93]: import sys

sys.maxint

Out[93]: 9223372036854775807

On a 64-bit OS the maximum value should be like 264−1 − 1, accounting for the fact that one
bit is needed for the sign and since zero is represented you have to stop one value short of 263 .
The native math capability of Python goes one step further via the long type. The long type
offers unlimited size! Furthermore if you are working with an int type and perform an opera-
tion that exceeds the maximum size, it will converted to a long integer for you. Loop counters
however, are bound to the maximum size on int. There are work arounds for this too.

In [94]: x = 34
(type(x),x) # display two results, again using a tuple

12
Out[94]: (int, 34)

In [95]: y = x**32
(type(y),y) # display two results, again using a tuple

Out[95]: (long, 10170102859315411774579628461341138023025901305856L)

In [99]: 1-y

Out[99]: -10170102859315411774579628461341138023025901305855L

Notice in the above examples that long integers are displayed with L appended to the end.

Other Bases In computer engineering you often need to work with other bases.

Bitwise Operations Along with the display of integers in other formats, Python also sup-
ports the bitwise operations shown in the following table.

In [22]: SVG('Python_Basics_files/Bitwise_table.svg')

Out[22]:

Native Bitwise Arithmetic Operators

Name Operator Quick
Example
>>> ~0x001 + 0x100
Bitwise not ~
254 sincehex(~0x001) = -0x2
>>> bin(0b0001<<2)
Shift left <<
'0b100' or4
>>> bin(0b11001>>3)
Shift right >>
'0b11' or3
Exclusive OR bin(0b1010^0b1111)
^
(XOR) '0b101' or5
>>> bin(0b101100 & 0xf)
AND &
'0b1100' or12
>>> bin(0b1000 | 0b0001)
OR |
'0b1001' or9

Note:If you search the Internet you will find little helper functions to
allow you to represent hex values with proper sign extension.

13
String String creation and manipulation in Python is a breeze. In signals and systems work
string manipulation often shows up when working with formatted printing, on the screen or in a
text file, and in labels for plots.
The ability to mix fixed text with formatted numbers, on the fly, is very handy when you have
to tabulate analysis and simulation results. Formatted print strings will discussed when I discuss
the print() function. Presently the basic of type str are discussed.
Formally a string in Python is a sequence of immutable characters. Immutable means the values
of the string cannot be changed. You can easily create a new string from an existing string, and
this is where you can introduce changes you may desire.
A string can be indicated using: (1) single quotes (2) double quotes, or (3) triple quotes to create
a string block.

xa = 'Bat'
xb = "Bat"
xc =\
"""
Many bats flying
through the air.
"""

In [124]: xa = 'Bat'
xb = "Bat"
xc =\
"""
Many bats flying
through the air.
"""
# Use a tuple to display some results
(xa,type(xa),xb,xc)

Out[124]: ('Bat', str, 'Bat', '\nMany bats flying \nthrough the air.\n')

Note: The multi-line string has embedded line feed, \n, characters.
Single and double quotes are interchangeable and are useful when you want to preserve quotes
that belong in a string, e.g.,

In [108]: xd = "It's a beautiful day"

xd

Out[108]: "It's a beautiful day"

Don’t be afraid to experiment!

String can be concatenated or added quite easily:

In [114]: 'Hello,' + ' ' + 'what is your name?'

Out[114]: 'Hello, what is your name?'

14
 The number of characters in a string can be found using the len() function, while individual
characters can be accessed using brackets, e.g., xd[3].
Indexing can be used to obtain substrings. The indices are integers which run from 0 to
len()-1. To generate substrings use brackets, i.e.,

In [121]: len(xd)

Out[121]: 20

The table below sumarizes basic string manipulation, including the fun topic of slicing. Slicing
returns with native Python lists, tuples, and NumPy ndarrays.

In [24]: SVG('Python_Basics_files/StringOperations.svg')

Out[24]:

15
 Native string operations
Name Operation Quick
Example
xa + xb
Concatenate/ >>> ‘Hello’ + ‘ ‘ + ‘World’
to concatenate
adding ‘Hello World’
strings
x*n or
Replicate/ n*x >>> ‘Bat’ * 3
multiply to replicate a ‘BatBatBat’
string ntimes
x[n] >>> x = ‘Bat’
x[-1] >>> x[1]
the end value ‘a’
Indexing x[-2] >>> x[-1]
the second t
from the end
x[n:m] >>> x = ‘Bright colors’
the substring >>> x[1:6]
from nto m-1 ‘right’
x[:m] >>> x[7:]
the substring ‘colors’
from 0to m-1 >>> x[:-1]
x[n:] ‘Bright color’
>>> x[::2]
the substring
‘Bih oos’
Slicing: from nto the
In the above the third argument is the
Many forms end
optional stride(the default, if not
possible x[n:-1]
given is 1) factor which controls the
the substring
step size as you run from 0 to the
from nto
end-1 end in this case, since only::is
x[n:-2] given.
the substring
from nto
end-2
x[n:m:k]
Note:Indexing and slicing will work the same way when wiring with
Python listsand tuples, and the Numpyndarray.

There are many functions for searching and modifying strings. Too many to cover here. If you
feel the need, do some searching on your own. As a specific example, consider breaking a string
down into substrings and then put back together in a differnt form. Below I use find() to do
some simple string parsing to assit in the tear-apart:

In [120]: xd[0:5] + xd[xd.find('beau'):xd.find('day')-1]

Out[120]: "It's beautiful"

16
Boolean The boolean type has it place in making logical decisions in program flow. In Python
the boolean type holds either True (1) or False (0). You will see booleans in action when I discuss
program flow. Logical operation as used in program flow control return booleans. A few simple
examples follow:

In [161]: b1 = True
b1 > 1

Out[161]: False

In [162]: 34 > 0 and 4 < 3

Out[162]: False

In [163]: 34 > 0 or 4 < 3

Out[163]: True

Data Structures
Python’s native data structures of interest here are lists, tuples, and briefly dictionaries. All
three of these data structures are sequences that can be thought of as containers for Python objects.
The most import object you will be using is the ndarray, which I have made mention of many
times. Although note mentioned in the section on string, they are also sequences of characters.

Lists
Simply put, a list is a mutable (changable) sequence of objects. A list can be created using brack-
ets:

In [146]: l1 = [1,'abc',23.4]
l1

Out[146]: [1, 'abc', 23.4]

Indexing and slicing of lists works the same as with strings. In fact a list can hold strings as
you see in the above example.
When I introduce for loops a little bit later, you will encounter a list object containing integers.
With regard to for loops, the native function range(), is frequently used to create a list of integers.
Consider the examples below:

n1 = range(start,stop,step) # = [start,start+step,start+2*step,...]
n2 = range(20) # = [0,1,2,...,19]

If start is omitted the sequence starts at 0. If step is omitted the step size is 1. Note step may
be negative.
The fact that lists are mutable means I can write

n1 = range(10)
n1[4] = 20 # replace the 5th element with 20, not a problem

17
In [169]: n1 = range(10)
n1
Out[169]: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
In [170]: n1[4] = 20
n1
Out[170]: [0, 1, 2, 3, 20, 5, 6, 7, 8, 9]
In [136]: (type(n1),len(n1))
Out[136]: (list, 10)
List can contain lists, and so on. Below I create n2 as a list of two lists made from subsequences
of the original n1.
In [137]: n2 = [n1[:5],n1[5:]]
n2
Out[137]: [[0, 1, 2, 3, 20], [5, 6, 7, 8, 9]]
Indexing and slicing into n2 now requires two indices:
In [144]: n2[1][:3]
Out[144]: [5, 6, 7]
There are methods (functions) associated with list objects. In engineering/scientific applications
of Python you need to be aware of lists, but explicit use of lists beyond range() (or the memory
conserving xrange()) is minimal, as NumPy’s ndarray is more powerful. As a simple example
consider sort() which places the list elements in ascending order:
In [171]: n3 = sort(n1) # Note n1.sort() sorts in-place
n3
Out[171]: array([ 0, 1, 2, 3, 5, 6, 7, 8, 9, 20])

List Comprehensions Indexing through lists and performing calculations is a frequent task,
at least without NumPy. Python allows you to combine looping and list manipulation into one
operation.
new_list = [function_of_index for index in range(n1,n2+1)]
#or to list the values in the terminal immediately
[function_of_index for index in range(n1,n2+1)]
Below is a simple example that returns a list of numbers corresponding to 3 + 4n + n2 for
0 ≤ n ≤ 10.
In [173]: [3+4*n+n**2 for n in range(0,11)]
Out[173]: [3, 8, 15, 24, 35, 48, 63, 80, 99, 120, 143]
When you use list comprehensions you can write very terse Python code. I encourage you to
explore list comprehensions as you feel more comfortable with the language. With NumPy the list
comprehension still provides a convenient way to fill a list or array with values of interest, but
again NumPy has its own ways too, that most likely are even faster.

18
Tuples
A tuple is like a list, but it is immutable (not changable). Your first reaction to this might be ’what
good is it if I can’t change it’. It turns out that the immuatabilty aspect is perfect for the needs of
engineering/scientific computing.
Creating a tuple can be done using parenthesis much like you do with lists. One signifi-
cant difference is that a single element tuple requires a comma so as not to be confused with the
ordinary use of parenthesis.

In [147]: t1 = (1,2,23.5,'abcd')
t1

Out[147]: (1, 2, 23.5, 'abcd')

In [159]: t2 = (27)
t2 # This is not a one element tuple

Out[159]: 27

In [160]: t3 = (34.5,) # the comma does it, its a one element tuple
t3

Out[160]: (34.5,)

In [156]: type(t3)

Out[156]: tuple

Trying to change a value of a tuple element fails, as you can see from the following:

In [151]: t1[1] = 56

---------------------------------------------------------------------------
TypeError Traceback (most recent call last)

<ipython-input-151-4a896955a572> in <module>()
----> 1 t1[1] = 56

TypeError: 'tuple' object does not support item assignment

A typical use of the tuple is as a return from a function call. Each element of the tuple can
be an object such as a list or with NumPy an ndarray. You can then unpack the tuple into its
constituent objects, say a frequency array and a frequency response array. Further analysis follows.
Suppose you have a function that returns a tuple of two lists. List 1, denoted l1, containing
numbers and list 2, denoted l2, containing characters, you can unpack the tuple into two lists as
follows:

19
In [152]: # First set up the scenario by artificially creating
# a tuple containing two lists
l_composite = ([0,1,2,3,4],['a','b','c','d','e'])
# Break the tuple apart
l1,l2 = l_composite

In [153]: l1

Out[153]: [0, 1, 2, 3, 4]

In [154]: l2

Out[154]: ['a', 'b', 'c', 'd', 'e']

Tuples have functions, such as len(),

Dictionaries
A dictionary is a mutable (changable) sequence of values that is addressable using a name/key. The
key needs to be unique, but the value does not. Dictionaries like lists are mutable.
The motivation for introducing dictionaries at this time is because some of the numerical al-
gorithms in SciPy return dictionaries. If you should need to use one of these algorithms, then you
will need to know something about dictionaries.
To create a dictionary using braces to create {key : value} pairs.

In [180]: weekdays = {'monday' : 1, 'tuesday' : 2, 'wednesday' : 3,

'thursday' : 4, 'friday' : 5}
type(weekdays)

Out[180]: dict

You can now access the dictionary elements using the keys:

In [177]: weekdays['wednesday']

Out[177]: 3

Dictionaries have a collection of associated functions. For example, you can list the keys using
the keys() method:

In [179]: weekdays.keys()

Out[179]: ['tuesday', 'thursday', 'friday', 'wednesday', 'monday']

If you have a dictionary but don’t know whats inside, you can list() it as tuples. The order
of the list is the hash ordering, which is an internal order scheme for fast retrieval.

In [181]: weekdays.items()

Out[181]: [('tuesday', 2),

('thursday', 4),
('friday', 5),
('wednesday', 3),
('monday', 1)]

20
Variables
You have already seen variables in action, but there are some fine points you need to know about
them. Variable names can contain characters, numbers, and underscores. Variables cannot begin
with a number.
Since everything in Python is an object, all objects have an address. If you declare a structure
variable (say a list) it is given an address. If you later set the list variable name equal to the first
variable you will not be creating a new object. Rather you create a new reference to the same
object. Python does have a copy method for these instances when you really do want a copy.

In [184]: a1 = [0,23.4]
a2 = a1
(a1,a2)

Out[184]: ([0, 23.4], [0, 23.4])

In [185]: a2[0] = 56.8

(a1,a2)

Out[185]: ([56.8, 23.4], [56.8, 23.4])

Notice in the above reassignement of the first element of list a2, the values held by a1 have
followed. In other words a1 references the same object. To insure you actually make a copy, you
can use some form of copy method. For Numpy ndarrays seen later, there is a copy() method.
For lists you can use a4 = list(a3) to make a copy:

In [190]: a3 = [2,17]
a4 = list(a3)
(a3,a4)

Out[190]: ([2, 17], [2, 17])

In [193]: a4[1] = 20
(a3,a4,'<== It works!')

Out[193]: ([2, 17], [2, 20], '<== It works!')

Formatting Strings and Gathering User Input

Strings and gathering user input may seem unrelated, but they come together when you to write
interactive programs in Python. I have placed this section here so that it can put to use in the
section on flow control coming up next.

Formatting Strings and Printing

Being able to control the format of numers displayed both on the screen and in files allows you to
effectively communicate the results of your Python analysis and simulation. Python supports two
approaches: string interpolation and formatting strings. I will be showing just string interpolation, as
to me it is very easy to pick up, as it follows from a background in C. Formatting strings do give
more control.

21
 To print a string to the console/terminal, IPython qtconsole, or IPython notebook, you use the
print() with a formatted string as the argument, or in many cases all rolled up into one statement.
A string interpolation expression takes the form format % values, where format is a string and
values is a tuple of values. The % character present in the string format indicates a value should
placed in the string using a format specification.
Consider the following simple example:

In [197]: v1 = 3.141516
v2 = 2*v1
print('v1 = %6.4f and v2 = %2.4e' % (v1,v2))

v1 = 3.1415 and v2 = 6.2830e+00

Note: The values to be formatted are contained in the tuple following the % character. The
formatting for the values always follows a % as well. Don’t be confused, there are multiple uses of
% in string interpolation.
If you simply wanted a string for use in plot labels, etc. you can write:

In [198]: str1 = 'v1 = %6.4f and v2 = %2.4e' % (v1,v2)

str1

Out[198]: 'v1 = 3.1415 and v2 = 6.2830e+00'

As I said from the start, string interpolation is very much like string formatting in C. The format
string specifications are given is the table below.

In [19]: SVG('Python_Basics_files/Printformat_table.svg')

Out[19]:

Format specifications for strings and use inprint()

Character Type Example
d Integer print(‘x = %d’ % x)
Engineering notation with e or
e,E print(‘x = %2.4e’ % x)
E respectively for the exponent
f Floating point print(‘x = %6.4f’ % x)
General f or e depending upon
g need. Decimal point and print(‘x = %2.4g’ % x)
trailing zeros may be ommitted
o Octal, not too common print(‘x = %o’ % x)
s String print(‘x = %s’ % x)
Hexadecimal lower or upper
x,X print(‘x = %x’ % x)
case, i.e., 0x or 0X

Note: d, f, and sar e the most common format types. Thef

specification is very nice forfloats.

22
 When for loops are introduced in the next section you will see how nice tabular lists of data
can be prepared. As a quick example which iterates over the list [-23, 34, 1004] consider:

In [203]: # Use of format specifications; also not line continue via \

for k in [-23,34,1004]:
print('Decimal: %o, Decimal padded: %4d, Hex: %x,\
Hex string: %s' % (k,k,k,hex(k)))

Decimal: -27, Decimal padded: -23, Hex: -17, Hex string: -0x17
Decimal: 42, Decimal padded: 34, Hex: 22, Hex string: 0x22
Decimal: 1754, Decimal padded: 1004, Hex: 3ec, Hex string: 0x3ec

Gathering User Input

User input may be provided interactively or in the case of a Python script via command line
areguments. Both are of interest, with the latter perhaps being more relevant to Python applica-
tions running under the control of another program.

Interactive User Input The function used to accept user inputs is

val = input('format string')

Below is a simple example:

In [200]: val = input('Enter a number: ')

Enter a number: 234.5

In [202]: (val,type(val))

Out[202]: (234.5, float)

Running Scripts with Command Line Arguments as Inputs There are times when you may
want to write a Python script that you can run from the command line or perhaps have another
program call. As an example, I have written GUI apps in another language that bring together
both command line C++ executables and Python script outputs.
A Python script is a *.py file containing code you might ordinarily type at the Python or
IPython prompt. You run the script right from the terminal provided by your OS:

Marks-MacBook-Pro:IPython_notebooks wickert$ python my_script.py arg1 arg2 arg_etc

Note: You can also run scripts from IPython using the %run magic, i.e.,

In [28]: %run my_script arg1 arg2 arg_etc

23
 Note: one or more command line argument may be supplied following the script file name.
The script is actually a Python code module that may contain functions as well as a script body,
that will run from the command line. Any functions in the module can be used by importing the
modules namespace into your Python (IPython) environment using:

import my_script

A sample script that reads four command line arguments is given below. This script imports
methods from the sys module for reading the command line arguments and the os module to
allow the full path to the script to be discerned. Having the full path comes in handy when you
want to read or write files from your script and you have called the script from another directory,
say even via another program.

#!/usr/bin/python

"""
cmd_line_test.py
A simple command line script program taking four arguments:
string = a file_name, e.g. data_set.txt
int = an interger loop variable
float = a calculations variable
float = a second calculations variable

Note all command line arguments are read as strings, so no

quotes are required.

Mark Wickert, October 2014

"""

# import needed modules and packages

from sys import argv, exit
import os
import numpy as np

#Get the app path for use later

app_path = os.path.dirname(os.path.realpath(__file__))

"""
Sample command line:
>>>python cmd_line_test.py cmd_test_results.txt 5 109.8 -34.567
"""
# Read command line arguments and convert as needed
if len(argv) < 4+1: # argv[0] is the script name itself
print('error: Need 4 command line arguments!')
print('User provided only %d.' % len(argv))
exit(1)
else:
out_file = argv[1]

24
 N_loops = int(argv[2])
value1 = float(argv[3])
value2 = float(argv[4])
# Do something with the collected inputs
print('Echo commandline inputs back to user:')
print('argv[0] = %s' % argv[0])
print('argv[1] = %s' % out_file)
print('argv[2] = %d' % N_loops)
print('argv[3] = %6.4f' % value1)
print('argv[4] = %6.4f' % value2)
# For reading and writing files you may want the full path
print('FYI, the path to your script is:')
print('%s' % app_path)
# Create an empty N_loops x 2 2D array
output_data = np.zeros((N_loops,2))
for k in xrange(N_loops):
output_data[k,0] = value1 + k*10.0
output_data[k,1] = (value1 + k*10.0)/value2
np.savetxt(app_path + out_file,output_data)
Running the above script from the terminal prompt results in:
Marks-MacBook-Pro:IPython_notebooks wickert$ python cmd_line_test.py
sample_output.txt 20 1823.69 -38276.76
Echo commandline inputs back to user:
argv[0] = cmd_line_test.py
argv[1] = sample_output.txt
argv[2] = 20
argv[3] = 1823.6900
argv[4] = -38276.7600
FYI, the path to your script is:
/Users/wickert/Documents/Documents/IPython_notebooks
A quick look at the file sample_output.txt reveals a nice list of two columns separated by a
space.
1.823690000000000055e+03 -4.764483723282744027e-02
1.833690000000000055e+03 -4.790609236518451192e-02
1.843690000000000055e+03 -4.816734749754159051e-02
...
The complementary Numpy function loadtxt() (discussed later) can easily load a text file
into ndarrays, using a variety of options.
Note: this script has also used a numpy method that makes it easy to write ndarrays to a text
file. More will be said about reading and writing ndarrays to files in the NumPy chapter.

Flow Control
The control of program flow is fundamental to moving on just using Python with NumPy. A lot
of good analysis can be done without flow control, but sooner or later you need to include some
looping and decision logic.

25
 The key operators used for decision logic in Python are shown in the table below.

In [17]: SVG('Python_Basics_files/LogicalOperations_table.svg')

Out[17]:

Logical and boolean comparison operators

Operator Type Example
X < Y Less than 5 < 6 returns True
X <= Y Less than or equal to 5 <= 4.9 returns False
X > Y Greater than 5 > 6 returns False
X >= Y Greater than or equal to 5 >= 4.9 returns True
X == Y Equal to (same value) 5 == 5.01 returns False
X != Y Not equal (also X <> Y in 2.X) 5 != 5.001 returns True
not X If X is false then True; else False not(5 < 6) r eturns False
(5 < 6) or(5 != 6) r eturns
X or Y If X is false then Y; else, X
True
(5 < 4) and(5 != 6) r eturns
X and Y If X is false then X; else Y
False

This is also where one of the unusual aspects of Python comes to light, that of code indenting.
Indenting and unindenting code by 4 four spaces is the standard. Python code editors are set up
this way, or you can make it so if not.
Indenting must be consistent all the way through a code block inthe IPython notebook or in
general in code module file. It is easy to mess up your indenting, so be careful. This is an area that
a newcomer is likely to get frustrated with. Hang in there, it gets better with practice.
In this section I cover if, elif, else blocks, for loops, and while loops. What I will leave
for self study is try, else, and finally blocks.

If, elif, and else

In Python the core flow control structure is if, elif, else:

if condition1:
block1
elif condition2:
block2
...
elif conditionN:
blockN
else:
elseblock

26
 All code blocks must be indented (by convention 4 spaces and not a tab) from the if, elif,
else statements. A condition can be passed over by including the pass statement in place of an
actual block. Coding continues following the elseblock by outdenting. No blank lines required.
AT first this seems strange, but you get used to it. The Canopy code editor as well as the editor
used for code in the IPython notebook help get you up to speed.

In [210]: my_value = 10
if my_value <= 4:
print('I am in the first block!')
elif my_value > 4 and my_value <= 8:
print('I am in the second block!')
else:
print('I am in the default block!')

I am in the default block!

In [217]: modeA = 'Green'

modeB = 'hot'
if (modeA.lower() == 'green') and (modeB.lower() != 'cold'):
print('What I am looking for!')
else:
print('No match!')
print('Entered a new block due to outdent')

What I am looking for!

Entered a new block due to outdent

For Loops
The for loop in Python is differnt from that found in most other languages.

for element in sequence:

ForCodeBlock
# Outside for loop due to outdent. Carry on with the program flow

What you see in the above says that a for loop is governed by the for element in sequence
statement. The words for and in must appear. How you choose to handle element and sequence
is up to you. The simplest configuration is to let element = k, and index variable and define a
sequence (list) of integers using the Python native range() function:

for k in range(10)
print('Index k = %d' % k)

As defined above k steps over the values in the list, which here has values 0,1,2,...,10-1.
The use of range() is convenient since it can generate a sequence of values to iterate over. The
xrange() function is better still because it does not have to allocate memory for the entire list.
The list you interate over can be most anything. In signals and systems work you typically
have a sequence (list) of numbers, integer or floating point. Below I fill a list with floats manually,
but once NumPy is on board you will fill nrrays by some other means.

27
Get Two For One by Using the Iterator enumerate When you process float values in a loop your
frequently need to use both the sequence index and the value itself. The loop interation contruct
that I really like makes use of the Python iterator enumerate. Consider:

x = [0, 0.1, 0.2, 0.3, 0.4, 0.5]

for n, xn in enumerate(x):
print('n = %d and xn = %4.1f' % (n, xn))

The iterator enumerate returns both an index to x and the value at the corresponding index, in
that order. Check it out in the notebook:

In [10]: # enumerate() Demo

x = [0, 0.1, 0.2, 0.3, 0.4, 0.5] # An input list of floats
y = [0.0 for i in range(6)] # A list filled using list comprehension
for n, xn in enumerate(x):
y[n] = xn**2 # note how I make use of both n and xn
print('n = %d, xn = %4.1f, and y[n] = %4.2f' % (n, xn, y[n]))

n = 0, xn = 0.0, and y[n] = 0.00

n = 1, xn = 0.1, and y[n] = 0.01
n = 2, xn = 0.2, and y[n] = 0.04
n = 3, xn = 0.3, and y[n] = 0.09
n = 4, xn = 0.4, and y[n] = 0.16
n = 5, xn = 0.5, and y[n] = 0.25

Another useful interator is reversed(). You can run everything in reverse:

In [18]: # reversed() Demo

x = [0, 0.1, 0.2, 0.3, 0.4, 0.5] # An input list of floats
y = [0.0 for i in range(6)] # A list filled using list comprehension
for n in reversed(range(len(x))):
y[n] = x[n]**2 # without xn I have to access the list directly
print('n = %d, xn = %4.1f, and y[n] = %4.2f' % (n, xn, y[n]))

n = 5, xn = 0.5, and y[n] = 0.25

n = 4, xn = 0.5, and y[n] = 0.16
n = 3, xn = 0.5, and y[n] = 0.09
n = 2, xn = 0.5, and y[n] = 0.04
n = 1, xn = 0.5, and y[n] = 0.01
n = 0, xn = 0.5, and y[n] = 0.00

While Loops
The while loop is very similar to the for loop, but the loop control point is different. Interation is
controlled by a logical compare at the top of the loop and exit the loop occurs when the condition
becomes false. The previous iteration of the loop takes place before you actually jump out of the
loop. Also, you must manage the variable used to form the condition.

28
while condition:
while_block

While condition is True looping continues. Infinite looping is also possible

# A never ending loop

while True
while_block

As a simple example consider:

In [13]: i = 5 # initialize the looping variable

while i <= 10: # form the True/False condition
print(i)
i += 1 # incrementthe looping variable

5
6
7
8
9
10

The Statements break and continue

When looping using for or while, you can use break to jump out of the loop and continue
with normal program flow. You can also use continue to skip the remainder of the code in the
for_block or while_block and come around to the next iteration.

In [16]: #break and continue Demo

print('Here I break at 2')
for i in range(4):
if i == 2: break
print('i = %d' % i)
print('Here I continue at 2')
for i in range(4):
if i == 2: continue
print('i = %d' % i)

Here I break at 2
i = 0
i = 1
Here I continue at 2
i = 0
i = 1
i = 3

29
Exceptions: try, except, and finally Blocks
Another aspect of flow control is the ability of a program handle runtime errors without crashing
your program. For the purposes of this intro tutorial. I consider exception handling to be a more
advanced topic. I am however including some discussion on this topic incase you are looking at
code samples you may find on the internet.
In Python exceptional handling is taken care of using try, except, and finally blocks. The
idea behind exception handling is to have the program catch that an exception has been raised or
thrown, then handle it in a safe way, and finally let the user know something about what happened.
To be completed later...

Functions
To me the heart and soul of any programming language is the ability to write reusable functions.
In Python functions are written using a def construct.

def function_name(arguments): # arguments are optional

"""
Function docstring to describe the purpose and variable input/output
"""
function_body # The function body must be indented
return one_or_more variables # The use of return is optional

Note: The return statement does not have to appear at the end of the function. You can actually
return from multiple locations if you need to. The bottom line is the function does end when it
reaches a return statement.
Arguments to the left can be given default values. If say two arguments are given default
values and you want to overide the lasgt value only, you must explicity refer to the last value in
the function call and give it a value:

def my_f1(a,b,c=5,d=25):
function_body
return a + b + c + d

# Using the function

x = my_f1(2.3,-4.7)
27.6
y = my_f1(2.3,-4.7,d=20)
22.6

In [19]: def my_function(a,b,c=8):

"""
A simple example function that takes three arguments:
a = arg1
b = arg2
c = arg3, which has a default value of 8
x = a + 31.5*b/c for c != 0 otherwise
a + 31.5*b/1000
"""

30
 # Conditional evaluation
if c == 0:
x = a + 31.25*b/1000
else:
x = a + 31.25*b/c
return x

In [20]: # test the function using the default c and with c = 0

(my_function(10.,20.),my_function(10.,20.,0))

Out[20]: (88.125, 10.625)

Object Oriented Python: Writing a Class

Basics
Object oriented programming (OOP) is quite easy in Python. So, what is it and how do you do it?
The following subsection walks through a simple examples. Once NumpY is introduced a signal
processing I develop the start of a simple filter class.
An object is a collection of data, say scalar numbers, lists, Numpy ndarrays, and functions.
To create a new object type you first have to create a class. The class defines what data types and
functions the object will contain. An object is said to encapsulate the data and functions that operate
on the data.
Objects can inherent data and functions from an existing class, if you wish. This can be a very
useful property, as it can save you the trouble of starting from scratch if some other class type has
much of what you need in your new class.

Writing a Simple Class

In a separate code module or right here in the Notebook, you write a class as follows:

In [20]: # Simple starter class entitled Entity for holding name,

# date, and time, and having some methods
# Import some modules needed for your class
import time # a Python standard library module
import datetime # a Python standard library module

class Entity(object): # object is the default to inherit from

"""
A simple starter class

Mark Wickert October 2014

"""
# You begin by initializing the class. This is the class
# constructor:
def __init__(self, me): #Note self refers to the object itself
self.name = me
# current time since epoch in float seconds
self.time = time.time()

31
 # year, month, date stucture
self.date = datetime.date.fromtimestamp(self.time)

# This is a special method that can be implemented to provide

# a string representation of the object.
def __str__(self):
string1 = 'Person %s started at %10.2fs, \n' \
% (self.name,self.time)
string2 = 'which corresponds to year %d, month %d, and day %d.' \
% (self.date.year,self.date.month,self.date.day)
return string1 + string2

# This is a special method that can be implemented to provide

# the official representation (repr) of the object.
# Without it you just get an object address when you type
# the object name and press enter.
def __repr__(self):
return str(self)

# Create a method to re-set the Entity name.

def set_name(self,new_name):
self.name = new_name

# Time in seconds the Entity has been in service

def service_time(self):
return time.time() - self.time

Note: You see self everywhere when you write a class. All objects and data must be preceeded
by self and every class method (function) must begin with a reference to self. Forgetting self
somewhere in your class definition is a fairly common error. Be on the look out for this error.

In [22]: # Create a new object of type Entity having name Joe

person1 = Entity('Joe')

In [23]: # Use the repr method to give the representation of person1

person1

Out[23]: Person Joe started at 1414386667.63s,

which corresponds to year 2014, month 10, and day 26.

In [24]: # Change the name of person1 using the setter method

person1.set_name('John')

In [25]: # Verify that the name change took place

person1

Out[25]: Person John started at 1414386667.63s,

which corresponds to year 2014, month 10, and day 26.

In [26]: str(person1)

32
Out[26]: 'Person John started at 1414386667.63s, \nwhich corresponds to year 2014, month 10, an

In [27]: person1.service_time()

Out[27]: 17.271279096603394

After Numpy
With Python basics taken care of, now its time to move on to the real focus of using Python for sci-
ence and engineering. NumPy (Numerical Python) is an open-source Python library for numerical
computing. When you combine NumPy with MatPlotLib and SciPy, and the IPython console or
notebook app, and you really have a very powerful set of tools. The full NumPy documentation.
The writing for the NumPy section is far from complete. At present I have placed many tables.

1. Numpy Fundamentals
2. Working with 1D Arrays

a. Signals

b. Systems

3. Working with 2D Arrays (Matrices)

4. The signal processing functions of ssd.py and digitalcom.py
5. A DSP Class using NumPy and Matplotlib

NumPy Fundamentals
The N-Dimensional Array and Available Types
The essence of the ndarray is shown in the figure below. Numerical operations with ndarrays
mostly run at full compiled code speed. This is particulary true if the math operations you perform
on an array do not change the size of the array.
Once an array is created you can access the attributes and many other methods using the .
operator. To see the complete list type:

dir(numpy)

In [25]: SVG('Python_Basics_files/Ndarray_stucture.svg')

Out[25]:

33
 The structure of the ndarray

The element data type,

dtype
e.g., float64or int8

attributes
dim count How many dimensions
Elements per dimension:
dimensions
(n, m, …)
Bytes per row, p,
strides
bytes per element,q
The data memory block
data with row and element
boundaries

q
e0 e1 e2 e3 e4 e5 e6 … … …
p

From outside

User/Python e0 e1 e2 e3 Note: C-ordering

view of 2D is the default as
e4 e5 e6 e7 shown. Fortran
array with
e8 e9 e10 e11 ordering (by
n = ?and
column) is also
m = 4 ... … … … possible

As a quick example consider:

In [33]: # Here I use the `array()` method (see Array Creation below)
A = array([1., 34.,-345, 98.2])
A

Out[33]: array([ 1. , 34. , -345. , 98.2])

In [34]: A.dtype

Out[34]: dtype('float64')

In [36]: A.shape # This a 1D array

Out[36]: (4,)

34
 When using PyLab, which makes the IPython environment similar to MATLAB, you work with
ndarrays in a very natural manner. The default data type for floats is double precision or 64 bit
(128 bits for complex). Many other data types can be used to make more efficient use of memory.
The table below lists the types and makes mention of how you declare types and perform casting
from one type to another.

In [27]: SVG('Python_Basics_files/ndarray_types_table.svg')

Out[27]:

Available ndarraydata types set by dtype

Type Description
bool Boolean (True or False) stored as a bit
int8 Byte (-128 to 127)
int16 Integer (-32768 to 32767)
int32 Integer (-2**31 to 2**31 - 1)
int64 Integer (-2**63 to 2**63 - 1)
uint8 Unsigned integer (0 to 255)
uint16 Unsigned integer (0 to 65535)
uint32 Unsigned integer (0 to 2**32 - 1)
uint64 Unsigned integer (0 to 2**64 - 1)
float16 Half precision float: sign bit, 5b expo, 10b mantissa
float32 Single precision float: sign bit, 8b expo. 23b mantissa
float64 Double precision float: sign bit, 11b expo, 52b mantissa
Complex number, represented by two 32-bit floats (real &
complex64
imag)
Complex number, represented by two 64-bit floats (real &
complex128
imag)
Note:The highlighted types are the defaults on a 64-bit OS. Type casting
is possible using methods such asy = uint32(x), etc.

Array Creation
The table below provides examples of commly used methods to create ndarrays.

In [28]: SVG('Python_Basics_files/Creating_ndarrays_table.svg')

Out[28]:

35
 Creating NumPyndarrays
Method Description Example
>>> a = array([1,2,3,4])
This is the core method used to
will create an int64 array
create ndarrays from a list. The
>>> a =
array() dtypear gument is good for array([1,2,3,4],dtype=float16)
setting the per element data will create a float16 array
type

>>> a = ones(20)
ones(n1) Will create an array of specified
a 20 element 1D array
or length (n1 or n1xn2, etc)
>>> a = ones((5,4))
ones((n1,n2)) containing all ones as 1D,2D, …
a 5x4 2D array of ones
Similar to ones() except fills >>> a = zeors(20)
zeros()
array with zeros a 20 element 1D array
Create a new array of zeros or
ones_like() >>> a = ones(10)
ones that replicates the shape
zeros_like() >>> b = zeros_like(a)
of the input argument
Special
1D
Array
Creation
Methods
Create an array of values
arange([start, ] running from startto stop- >>> x = arange(0,5,0.5)
stop[,step] step, where stepis the step creates an array of floats [0,0.5,1.0,…,4.5]
size
Create an array of linearly
linspace(start,s >>> x linspace(1,2,6)
spaced of numvalues running
top,num=50) creates the array [1.0,1.2,1.4,1.6,1.8,2.0]
from startto stop
Create an array of log spaced of >>> x = logspace(0,1,10)
logspace(start, numvalues running from creates the array [1. , 1.291, 1.668, …,
stop, num=50)
10startto 10stop 5.995, 7.743, 10.]

Note:I fr equently usearange()to cr eate index vectors and initialize arrays usingzeros()and/or
zeros_like().
Tip: If you add stepto stopin arange()the fi nal value will bestop.

Working With Arrays

Working with arrays is where it’s at! You want to solve problems using a technical computing and
visualization environment. Working with arrays is how you get your analysis and simulation
work done. There are many core functions/methods for this. In the following four tables below I
provide some important example methods. Obviously there are many more, and with SciPy and
many code modules written by people all over the world, the list goes on and on and on.
A good Web site to go to is PyPI. Not all packages are listed here (mine included at present),
but many are. Web searches often end up at this site.

In [29]: SVG('Python_Basics_files/ndarray_methods1_table.svg')

Out[29]:

36
 Popular methods for working withndarrays
Function Description Example
Logical
all() True if all elements are nonzero >>> x =
array([0,1,2,0,4,5])
all(x) = False
any() True if any (at least one) elements ar
e nonzero >>> x =
array([0,1,2,0,4,5])
any(x) = True
find() Return the indices where ravel(condition) is >>> x =
true array([0,1,2,2,1,7])
find(x >= 3) =
array([5])

Slicing
1D
arrays
(a
few
cases)
x[n:m] The 1D subarray from n to m-1 >>> x =
x[:m] The 1D subarray from 0 to m-1 array([0,1,2,3,4,5])
x[n:] The 1D subarray from n to the end x[:2] = array([0,1])
x[n:-1] The 1D subarray from n to end-1 x[::3] = array([0,3])
x[n:-2] The 1D subarray from n to end-2 x[1:-2] = array([1,2])
x[n:m:k] The 1D subarray from n to m-1 with k index
striding
Slicing
2D
arrays
(a
few
cases)
x[n:m,j:k] The 2D subarray from n to m-1, j to k-1 >>> x =
x[n:m,:] The 2D subarray from 0 to m-1, all columns array([[0,1,2],
x[:,j:k] The 2D subarray all rows, columns j to k-1 [3,4,5]])
x[n:m:o,j:k:l] The 2D subarray with striding by o and l in x[:2,:2] =
rows and columns respectively array([[0,1],[3,4])
x[-1,-1] =
x[n:-1,:] The 2D subarray from n to end-1, all columns array([[5]])
x[:,j:-2] The 2D subarray all rows, columns j to k-2
x[3,:] The 2D subarray row 3, all columns
x[:,0] The 2D subarray all rows, column 0

In [30]: SVG('Python_Basics_files/ndarray_methods2_table.svg')

Out[30]:

37
 Popular methods for working withndarrays(cont.)
Function Description Example
Shape
&
Concatenation
reshape() Reshape a 1D or 2D array to a new shape; the >>> x = arange(0,5)
new shape must be consistent. 1D 6 elements
y = reshape(x,(2,3))
2D 2x3 elements
concatenate() Join a sequence of arrays together . The arrays >>> x =
must have the same shape except in the axis array([[0,1,2,3,4,5]])
used for combining. axis=0 is or ws, axis=1 is 2D 1x6 elements
columns. concatenate((x,x),
axis=0)
2D 1x6 elements
concatenate((x,x)),
axis=1)
2D 1x12 elements
hstack() Stack arrays horizontally. A subset of >>> x =
concatenate() array([[0,1,2,3,4,5]])
2D 1x6 elements
x = x.T #transpose
y=hstack((x,x))
2D 6x2 columns
vstack() Stack arrays vertically. A subset of >>> x =
concatenate() array([[0,1,2,3,4,5]])
2D 1x6 elements
y=vstack((x,x))
2D 2x6 columns
flatten() Values of the argument array become a 1D >>> x =
array. May be done in-place with x.flatten() array([[0,1,2,3,4,5]])
x.flatten()
1D 6 element
transpose() or Like matrix transpose for 2D arrays. In-place >>> x =
array.T via x.T. array([[0,1,2,3,4,5]])
x.T
2D 6x1 array

In [31]: SVG('Python_Basics_files/ndarray_methods3_table.svg')

Out[31]:

38
 Popular methods for working withndarrays(cont.)
Function Description Example
Many
other
standard
functions,
e.g.,
trig,
Math
are
also
available
for
array
operations
mean(x) The sample mean of the values contained in >>> x =
array x. array([0,1,2,3,4,5])
mean(x) = 2.5
var(x) The sample variance of the values contained >>> x =
in array x. array([0,1,2,3,4,5])
var(x) = 2.9167
std(x) The sample standard deviation of the values >>> x =
contained in array x. array([0,1,2,3,4,5])
std(x) = 1.7078
sum(x) The sum of the values contained in array x. >>> x =
array([0,1,2,3,4,5])
sum(x) = 15
prod(x) The sample mean of the values contained in >>> x =
array x. array([0,1,2,3,4,5])
prod(x) = 0
cumsum(x) The sample mean of the values contained in >>> x =
array x. array([0,1,2,3,4,5])
cumsum(x) = array([0,
1,3,6,10,15])
cumprod(x) The sample mean of the values contained in >>> x =
array x. array([1,1,2,3,4,5])
cumprod(x) = array([1,
1,2,6,24,120])
min(x) The sample mean of the values contained in >>> x =
array x. array([0,1,2,3,4,5])
min(x) = 0
max(x) The sample mean of the values contained in >>> x =
array x. array([0,1,2,3,4,5])
max(x) = 5
conj(x) The sample mean of the values contained in >>> x = array([2+5j])
array x. conj(x) =
array([2+5j])
x.real & The real or imaginary part of the values >>> x = array([2+5j])
x.imag contained in array x. Also er al(x), imag(x) x.real = array([ 2.])
imag(x) = array([ 5.])

In [32]: SVG('Python_Basics_files/ndarray_methods4_table.svg')

Out[32]:

39
 Popular methods for working withndarrays(cont.)
Function Description Example
To
and
From
Files
and
Conversion
x.tofile(fname) Writes an array to a binary file. Assume a >>> x =
float type. This a quick means to save array([0,1,2,3,4,5],dty
data in binary form, but not very obust.
r pe-float64)
x.tofile(‘x_arr.bin’)
Reads an array from a binary file. >>> x =
Assumes a l=float type by default. fromfile(x_arr.bin)
x =
Undoes the operation oftofile(see returns
fromfile(fname)
above) array([1.,2.,3.,4.,5.])

tolist(x) Converts an array to a standard Python >>> x =

list. Leave the comforts ofndarrays. array([0,1,2,3,4,5])
x.tolist()
returns [0,1,2,3,4,5]
savetxt(x) Save an array to a text file with rows and >>> x =
columns matching x. Columns space array([[0,1,2,3,4,5],
separated. [6,7,8,9,10])
savetxt(x)
2 rows and 6 columns of text
x = Load a text file into one or more arrays. A >>> x,y =
loadtxt(fname) very flexible means of reading data sets. loadtxt(fname,
usecols=(0,2),
unpack=True)
Take out columns 0 & 2

Graphics and More with Matplotlib

Being able to integrate visualization with engineering calculations is extremely important. In
Python this is done using matplotlib. When you import pylab, see the first few cells of this docu-
ment/notebook, matplotlib is brought into your workspace.

Signals and Systems Tools and Examples

The Scipy Module scipy.signal
The full on-line help is here. The function name listing is given below:

dir(signal)
Out[31]:
[ 'abcd_normalize', 'absolute_import', 'argrelextrema', 'argrelmax',
'argrelmin', 'band_dict', 'band_stop_obj', 'barthann', 'bartlett',
'bench', 'bessel', 'besselap', 'bilinear', 'blackman',
'blackmanharris', 'bode', 'bohman', 'boxcar', 'bspline', 'bsplines',
'buttap', 'butter', 'buttord', 'cascade', 'cheb1ap', 'cheb1ord',
'cheb2ap', 'cheb2ord', 'chebwin', 'cheby1', 'cheby2', 'chirp',

40
 'cmplx_sort', 'cont2discrete', 'convolve', 'convolve2d',
'correlate', 'correlate2d', 'cosine', 'cspline1d', 'cspline1d_eval',
'cspline2d', 'cubic', 'cwt', 'daub', 'decimate', 'deconvolve',
'detrend', 'dimpulse', 'division', 'dlsim', 'dltisys', 'dstep',
'ellip', 'ellipap', 'ellipord', 'fftconvolve', 'filter_design',
'filter_dict', 'filtfilt', 'find_peaks_cwt', 'findfreqs',
'fir_filter_design', 'firwin', 'firwin2', 'flattop', 'freqresp',
'freqs', 'freqz', 'gauss_spline', 'gaussian', 'gausspulse',
'general_gaussian', 'get_window', 'hamming', 'hann', 'hanning',
'hilbert', 'hilbert2', 'iirdesign', 'iirfilter', 'impulse',
'impulse2', 'invres', 'invresz', 'kaiser', 'kaiser_atten',
'kaiser_beta', 'kaiserord', 'lfilter', 'lfilter_zi', 'lfiltic',
'lombscargle', 'lp2bp', 'lp2bs', 'lp2hp', 'lp2lp', 'lsim',
'lsim2', 'lti', 'ltisys', 'medfilt', 'medfilt2d', 'morlet',
'normalize', 'np', 'nuttall', 'order_filter', 'parzen',
'periodogram', 'print_function', 'qmf', 'qspline1d',
'qspline1d_eval', 'qspline2d', 'quadratic', 'remez', 'resample',
'residue', 'residuez', 'ricker', 's', 'savgol_coeffs',
'savgol_filter', 'sawtooth', 'scoreatpercentile', 'sepfir2d',
'signaltools', 'sigtools', 'slepian', 'spectral', 'spline',
'spline_filter', 'square', 'ss2tf', 'ss2zpk', 'step', 'step2',
'sweep_poly', 'symiirorder1', 'symiirorder2', 'test', 'tf2ss',
'tf2zpk', 'triang', 'unique_roots', 'vectorstrength',
'waveforms', 'wavelets', 'welch', 'wiener', 'windows', 'xrange',
'zpk2ss', 'zpk2tf']

Using scikit-dsp-comm
Folloiw the iunstruction of the README at the above scikit-dsp-comm link to clone and install the
repository and then begin using it. The support docs for this package are located at read the docs.
For the complete index see: index.
In particular the module sk_dsp_comm.siggsys, imported at the top of this notebook as

import sk_dsp_comm.sigsys as ss

was originally written for the book Signals and Systems for Dummies. The contents for this
module can be found using dir(ss)

dir(ssd)
Out[30]:
['BPSK_tx', 'CIC', 'NRZ_bits', 'NRZ_bits2', 'OA_filter',
'OS_filter', 'PN_gen', 'am_rx', 'am_rx_BPF', 'am_tx',
'biquad2', 'bit_errors', 'cascade_filters', 'conv_integral',
'conv_sum', 'cpx_AWGN', 'cruise_control', 'deci24',
'delta_eps', 'dimpulse', 'downsample', 'drect', 'dstep',
'env_det', 'ex6_2', 'eye_plot', 'fft', 'fir_iir_notch',
'from_wav', 'fs_approx', 'fs_coeff', 'ft_approx',
'interp24', 'line_spectra', 'lms_ic', 'lp_samp',
'lp_tri', 'm_seq', 'mlab', 'my_psd', 'np', 'peaking',

41
 'plot_na', 'plt', 'position_CD', 'prin_alias', 'pylab',
'rc_imp', 'rect', 'rect_conv', 'scatter', 'signal',
'simpleQuant', 'simple_SA', 'sinusoidAWGN', 'soi_snoi_gen',
'splane', 'sqrt_rc_imp', 'step', 'ten_band_eq_filt',
'ten_band_eq_resp', 'to_wav', 'tri', 'unique_cpx_roots',
'upsample', 'wavfile', 'zplane']

More Modules
The are are many more modules in the scikit-dsp-comm package. Visit the README1 to get the
details. There is also the GitHub repo for the Scipy 2017 tutorial that uses this package: SciPy 2017
tutorial.

A Simple DSP Class Case Study

Filters are used frequently in DSP. Filters have characteristics, such as impulse response, frquency
response, pole-zero plot. Filters are also used to operate on signals (sequences). You may want
to use a filter operate on contiguous blocks/frames of data. When this is done the filter has to
maintain state from use-to-use. Lowpass filters are used in decimators and interpolators,

The class Code Base

A filter object would be nice for keeping all of the above information organized. A preliminary
version of the class is implemented below:

In [52]: from __future__ import division #provides float div as x/y and int div as x//y
import numpy as np
import scipy.signal as signal
import ssd
# Create an FIR filter object around the signal.firwin method
class FIR_filter(object):
"""
An FIR filter class that implements LPF, HPF, BPF, and BSF designs using
the function signal.firwin.

Mark Wickert October/November 2014

"""
def __init__(self,order=20,f_type='lpf',cutoff=(0.1,),fsamp = 1.0,
window_type='hamming'):
"""
Create/instantiate a filter object:
fir_object = FIR_filter(order,f_type,cutoff=(0.1,),fsamp=1.0,
window_type='hamming')

order = the filter polynomial order; the number of taps is 1 + order

f_type = the filter type: 'LPF' (lowpass), 'HPF' (highpass),
'BPF' (bandpass), or 'BSF' (bandstop)
cutoff = the cutoff frequency/frequencies in Hz input as a tuple.

42
 a pair of cutoff frequencies is needed for BPF and BSF designs
fsamp = sampling rate in Hz
window_type = the default is hamming, but others can be found in
signal.windows, e.g., hanning (or hann)

"""
self.N = order # The number of filter taps is N+1
self.f_type = f_type # 'lpf', 'hpf', 'bpf', 'bsf'
self.fc = array(cutoff) # The cutoff freq in Hz; two cutoffs for bpf & bsf
self.fs = fsamp # In Hz
# Choose a window from from the type in the signal catalog
self.window = window_type
# Design the filter
# Note under some circumstances the end coeffients may be almost zero
# or zero. In these cases trim the filter length and report that the
# requested filter order was not not achieved. The threshold for removing
# coefficients is b_eps
b_eps = 1e-10
if f_type.lower() == 'lpf':
if len(self.fc) == 1:
self.b = signal.firwin(self.N+1,2*self.fc/self.fs,
window=window_type,pass_zero=True)
else:
print('For LPF only one cutoff frequency required')
elif f_type.lower() == 'hpf':
if len(self.fc) == 1:
self.b = signal.firwin(self.N+1,2*self.fc/self.fs,
window=window_type,pass_zero=False)
else:
print('For HPF only one cutoff frequency required')
elif f_type.lower() == 'bpf':
if len(self.fc) == 2:
self.b = signal.firwin(self.N+1,2*self.fc/self.fs,
window=window_type,pass_zero=False)
else:
print('For BPF two cutoff frequencies required')
elif f_type.lower() == 'bsf':
if len(self.fc) == 2:
self.b = signal.firwin(self.N+1,2*self.fc/self.fs,
window=window_type,pass_zero=True)
else:
print('For BSF two cutoff frequencies required')
else:
print('Filter type must be LPF, HPF, BPF, or BSF')
#Remove small or zero coefficients from the end of the filter
if self.b[0] < b_eps and self.b[-1] < b_eps:
self.b = self.b[1:-1]
print('Effective/realized filter order = %d' % (len(self.b)-1))

43
 """
WRITE ANY ADDITIONAL INITIALIZATION CODE HERE
"""

def freq_resp(self,mode = 'dB',Npts = 1024):

"""
A method for displaying the filter frequency response magnitude
or phase. A plot is produced using matplotlib

freq_resp(self,mode = 'dB',Npts = 1024)

mode = display mode: dB magnitude or phase in radians, both versus

frequency in Hz
"""
f = np.arange(0,Npts)/(2.0*Npts)
w,H = signal.freqz(self.b,[1],2*np.pi*f)
if mode.lower() == 'db':
plot(f*self.fs,20*np.log10(np.abs(H)))
xlabel('Frequency (Hz)')
ylabel('Gain (dB)')
title('Frequency Response - Magnitude')
elif mode.lower() == 'linear':
"""
Write code here
"""
pass

elif mode.lower() == 'phase':

plot(f,np.angle(H))
xlabel('Frequency (Hz)')
ylabel('Phase (rad)')
title('Frequency Response - Phase')
elif mode.lower() == 'degrees':
"""
Write code here
"""
pass

elif mode.lower() == 'groupdelay':

"""
Notes
-----

Since this calculation involves finding the derivative of the

phase response, care must be taken at phase wrapping points
and when the phase jumps by +/-pi, which occurs when the
amplitude response changes sign. Since the amplitude response
is zero the sign changes, the jumps do not alter the group

44
 delay results.

Mark Wickert November 2014

"""
theta = np.unwrap(np.angle(H))
# Since theta for an FIR filter is likely to have many pi phase
# jumps too, we unwrap a second time 2*theta and divide by 2
theta2 = np.unwrap(2*theta)/2.
theta_dif = np.diff(theta2)
f_diff = np.diff(f)
Tg = -np.diff(theta2)/np.diff(w)
plot(f[:-1],Tg)
min_Tg = np.min(Tg)
max_Tg = np.max(Tg)
ylim([np.floor(np.min([min_Tg,0])),1.2*np.ceil(max_Tg)])
xlabel('Frequency (Hz)')
ylabel('Group Delay (samples)')
title('Frequency Response - Group Delay')
else:
print('Error, mode must be "dB" or "phase"')

def pz_plot(self,auto_scale=True,size=1.5):
"""
Write doc string
"""
"""
Write code here
"""
pass

def impulse_resp(self):
"""
Write doc string
"""
"""
Write code here
"""
pass

def step_resp(self):
"""
Write doc string
"""
"""
Write code here
"""
pass

45
 def firfilt(self,x,reset=False):
"""
Write doc string
"""
"""
Write code here
"""
pass

def decimate(self,x,M,reset=False):
"""
Assuming the filter design is lowpass of the appropriate bandwidth,
follow LPF filtering with downsampling by M.
"""
"""
Write code here
"""
pass

def interpolate(self,x,L,reset=False):
"""
Assuming the filter design is lowpass of the appropriate bandwidth,
upsample by L then LPF filter. A gain scale of L is also included.
"""
"""
Write code here
"""
pass

The key features of the class at present is that it can design lowpass, highpass, bandpass, and
bandstop FIR filters using the window method. Once a filter object is created using say

fir = FIR_filter(31,'LPF',(100,),1000)

you can then use methods to plot the frequency response magnitude in dB and the frequency
response phase in radians.
Notice that code place holders are present for adding more methods to the class:

1. Not shown impulse response plotting.

2. Not shown step response plotting.
3. Frequency response magnitude linear scale.
4. Frequency response phase in degrees.
5. Pole-zero plot using the function ssd.zplane.
6. Filtering of an input sequence x [n] to produce output y[n], with initial conditions maintained
should more than one frame of data be processed.
7. Decimation of x [n] by the factor M should the filter be an appropriately chosen lowpass filter.
The implementation of state maintenance is intended, so again seamless frames processing
is possible.

46
 8. Interpolation of x [n] by the factor L should the filter be an appropriately chosen lowpass
filter. The implementation of state maintenance is intended, so again seamless frames pro-
cessing is possible.
9. Not shown is rational number rate changing.
10. Not shown is a means to choose alternate FIR types such as equal-ripple (remez) and fre-
quency domain sampling (fir2).

Making a Standalone Module The code has imports listed at the top should you desire to place
it in a module by itself. There is one detail missing however. Any of the current commands that
plot, i.e., plot() or stem() will require some rework in a standalone code module. YOu will want
to changes the import section of the module to look something like:

from __future__ import division

from matplotlib import pylab
from matplotlib import mlab
import matplotlib.pyplot as plt
import numpy as np
import scipy.signal as signal
import ssd

All three matplotlib imports are needed, but it is plt that you will directly work with for
doing plotting inside the module. Take a portion of the frequency response plotting method for
example. In the following code listing I have added or augmented five lines:

def freq_resp(self,mode = 'dB',Npts = 1024):

f = np.arange(0,Npts)/(2.0*Npts)
w,H = signal.freqz(self.b,[1],2*np.pi*f)
plt.figure() # create a blank figure using the plt object imported
if mode.lower() == 'db':
plt.plot(f*self.fs,20*np.log10(np.abs(H))) #Draw a plot on the plt object
plt.xlabel('Frequency (Hz)') #Place a label on the plt object
plt.ylabel('Gain (dB)') #Place another label on the plt object
plt.title('Frequency Response - Magnitude') #Place a title on the plt object

The changes need to be made throught the class definition so it can draw plots when methods
are called from FIR_filter objects. This of course assumes you have imported the module into
your IPython notebook or IPython qt console session.

Lowpass and Bandpass Examples

Try out the class with a few quick examples. I first make a lowpass filter and then a bandpass
filter.

In [60]: # Lowpass: N = 31 or 32 Taps, fs = 1000 Hz and fc = 200 Hz

fir1 = FIR_filter(31,'LPF',(200,),1000)

In [75]: # Bandpass: N = 64 or 65 Taps, fs = 1000 Hz and fc1 = 200 Hz, fc2 = 300 Hz
fir2 = FIR_filter(64,'BPF',(200,300),1000)

47
Effective/realized filter order = 62

You may wonder in the above BPF design what the message *effective filter order of 62 is all
about. With the windowed FIR design approach, it is possible for the first and last coefficients to
be very small or even zero. This effectively reduces the filter order by two. In the filter constuctor
I remove these coefficients to reduce the calculation count and reduce the filter delay.

Frequency Response Magnitude Plots Verify that the frequency response magnitude in dB
method does indeed work.

In [76]: fir1.freq_resp()
fir2.freq_resp()
ylim([-80,0])
grid()
legend(((r'FIR1 (LPF)',r'FIR2 (BPF)')),loc='best').get_frame().set_alpha(0.8)

Frequency Response Phase Plots Verify that the frequency response phase in radians method
does indeed work.

In [77]: fir1.freq_resp('phase')
fir2.freq_resp('phase')
grid()
legend(((r'FIR1 (LPF)',r'FIR2 (BPF)')),loc='best').get_frame().set_alpha(0.8)

48
 Note: The neat matplotlib legend feature (.get_frame().set_alpha(0.8)) that allows the
transparency so the plot lines can be seen behind the legend frame. Here the opacity is 80% (100%
or 1.0) means not opaque.
This is a cross-reference link to Section ??, just to verify that it can be done.

References
Python Converage (core language only no NumPy or SciPy)
[1]: Mark Lutz, Python Pocket Reference, 5th edition, O’Reilly, 2014. On Amazon
[2]: Toby Donaldson, Python: Visual QuickStart Guide, Third Edition, Peachpit Press, 2014. On
Amazon
NumPy/SciPy Python Converage
[3]: Shai Vaingast, Beginning Python Visualization Crafting Visual Transformation Scripts, 2nd edi-
tion, Apress, 2014. On Amazon
[4]: Python Scientific Lecture Notes. I suggest you download the one page per side pdf version.
[5]: Hans Petter Langtangen, A Primer on Scientific Programming with Python, 3rd edition,
Springer, 2012. On Amazon
[6]: Ivan Idris, NumPy Beginner’s Guide 2nd Edition, PACKT PUBLISHING, 2013. On Amazon

49