Scientific Python Tutorial

Scientific Python
Yann Tambouret
Scientific Computing and Visualization
Information Services & Technology
Boston University
111 Cummington St.
[email protected]

October, 2012

Yann - [email protected] (SCV)

Scientific Python

October 2012

1 / 76

This Tutorial
This tutorial is for someone with basic python
experience.
First:
1
2
3
4

matplotlib - plotting library, like Matlab

numpy - fast arrays manipulation
scipy - math methods galore.
ipython - better interactive python

Then I will cover a few intermediate details about

python programming in general

Yann - [email protected] (SCV)

Scientific Python

October 2012

2 / 76

Plotting in scripts
is a package that has many
modules, pyplot is the main driver.
matplotlib is designed for both
interactive and script-based use.
matplotlib

Yann - [email protected] (SCV)

Scientific Python

October 2012

3 / 76

Plotting in scripts
is a package that has many
modules, pyplot is the main driver.
matplotlib is designed for both
interactive and script-based use.
a Figure can contain many Axes ,
which contain many plots.
pyplot creates a default Figure and
Axes if you just want get to plotting
things quickly.
matplotlib

Yann - [email protected] (SCV)

Scientific Python

October 2012

3 / 76

Basic Example

1
2
3
4
5
6
7
8
9

from matplotlib import pyplot

pyplot . plot ([0 , 2 , 4 , 8 , 16 , 32] , " o " )
pyplot . ylabel ( " Value " )
pyplot . xlabel ( " Time " )
pyplot . title ( " Test plot " )
pyplot . show ()

Yann - [email protected] (SCV)

Scientific Python

October 2012

4 / 76

Basic Plot - Results

Yann - [email protected] (SCV)

Scientific Python

October 2012

5 / 76

World Population - Practice Part 1

Explaining whats there

practice/world_population.py
3 def get_data ():
4
data = file ( " w o r l d _ p o p u l a t i o n . txt " , " r " ). readlines ()
5
dates = []
6
populations = []
7
for point in data :
8
date , population = point . split ()
9
dates . append ( date )
10
populations . append ( population )
11
return dates , populations

We read in the data

Yann - [email protected] (SCV)

Scientific Python

October 2012

6 / 76

World Population - Practice Part 1

Explaining whats there

practice/world_population.py
3 def get_data ():
4
data = file ( " w o r l d _ p o p u l a t i o n . txt " , " r " ). readlines ()
5
dates = []
6
populations = []
7
for point in data :
8
date , population = point . split ()
9
dates . append ( date )
10
populations . append ( population )
11
return dates , populations

1
2

We read in the data

For each line (point) we need to separate the date
from population

Yann - [email protected] (SCV)

Scientific Python

October 2012

6 / 76

World Population - Practice Part 1

Explaining whats there

practice/world_population.py
3 def get_data ():
4
data = file ( " w o r l d _ p o p u l a t i o n . txt " , " r " ). readlines ()
5
dates = []
6
populations = []
7
for point in data :
8
date , population = point . split ()
9
dates . append ( date )
10
populations . append ( population )
11
return dates , populations

1
2

We read in the data

For each line (point) we need to separate the date
from population
split() splits text at any whitespace, by default.

Yann - [email protected] (SCV)

Scientific Python

October 2012

6 / 76

World Population - Getting Help

How to read online docs
1
2
3

Google matplotlib and pick first choice

Quick Search (on right side) for pyplot.plot
Select result entitled matplotlib.pyplot.plot

Yann - [email protected] (SCV)

Scientific Python

October 2012

7 / 76

World Population - Practice Part 2

Your Assignment
13
14
15
16
17
18

def plot_world_pop ( dates , populations ):

pass
dates , populations = get_data ()
plot_world_pop ( dates , populations )
pyplot . show ()
1
2
3
4
5

Make a plot of population (y) vs dates (x)

Title: World population over time
Y-axis: World population in millions
X-axis: Year
Set the plot type to a
Magenta, downward-triangle

Yann - [email protected] (SCV)

Scientific Python

October 2012

8 / 76

World Population - Final

Yann - [email protected] (SCV)

Scientific Python

October 2012

9 / 76

Life Expectancy - Practice Part 1

practice/life_expectacies_usa.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

from matplotlib import pyplot

def get_data ():
# each line : year , men , women
data = file ( " l i f e _ e x p e c t a n c i e s _ u s a . txt " , " r " ). readlines ()
dates = []
men = []
women = []
# finish me !
return dates , men , women
def p l o t _ e x p e c t a n c i e s ( dates , men , women ):
pass
dates , men , women = get_data ()
p l o t _ e x p e c t a n c i e s ( dates , men , women )
pyplot . show ()

Yann - [email protected] (SCV)

Scientific Python

October 2012

10 / 76

Life Expectancy - Practice Part 2

1
2
3
4
5

Use split(,) to split strings at commas

Add a label to each plot (look at documentation)
Label Axes and give a title.
Call plot 2x to plot two lines
Add a legend: pyplot.legend
I
I

Quick Search again

search for pyplot.legend

Yann - [email protected] (SCV)

Scientific Python

October 2012

11 / 76

Life Expectancy - Results

Yann - [email protected] (SCV)

Scientific Python

October 2012

12 / 76

Plotting Interactively
1

Yann - [email protected] (SCV)

Home - Backward Foward - Control edit history

Scientific Python

October 2012

13 / 76

Plotting Interactively
1

Yann - [email protected] (SCV)

Home - Backward Foward - Control edit history

Pan - Zoom - Left click +
drag shifts center, right click
+ drag changes zoom

Scientific Python

October 2012

13 / 76

Plotting Interactively
1

Yann - [email protected] (SCV)

Home - Backward Foward - Control edit history

Pan - Zoom - Left click +
drag shifts center, right click
+ drag changes zoom
Zoom - Select zoom region

Scientific Python

October 2012

13 / 76

Plotting Interactively
1

3
4

Yann - [email protected] (SCV)

Home - Backward Foward - Control edit history

Pan - Zoom - Left click +
drag shifts center, right click
+ drag changes zoom
Zoom - Select zoom region
Save

Scientific Python

October 2012

13 / 76

Plotting Interactively
1

3
4

Home - Backward Foward - Control edit history

Pan - Zoom - Left click +
drag shifts center, right click
+ drag changes zoom
Zoom - Select zoom region
Save

is an alternative to pyplot.show()
when you are using pyplot non-interactively.

pyplot.savefig(filename)

Yann - [email protected] (SCV)

Scientific Python

October 2012

13 / 76

Matplotlib Resources
Possibilities:
matplotlib.org/gallery.html
http://matplotlib.org/basemap/users/examples.html

Guide/Tutorial:
matplotlib.org/users/index.html
Questions:
stackoverflow.com
Or contact us: [email protected]
Source for tutorial:
openhatch.org/wiki/Matplotlib
Yann - [email protected] (SCV)

Scientific Python

October 2012

14 / 76

Array Creation

By convention:

import numpy as np

x = np.array([1,2,3], int);

x is a Numpy array
x is like a list, but certain operations are much faster

Yann - [email protected] (SCV)

Scientific Python

October 2012

15 / 76

Array Creation

A 2D array:
1 A = np . array (((1 ,0 ,0) ,
2
(0 , 0 , -1) ,
3
(0 , 1 , 0)))

Yann - [email protected] (SCV)

Scientific Python

October 2012

16 / 76

Array Creation

makes array([ 1., 1.,

same thing for zeroes

np.ones(5)
np.zeros

Yann - [email protected] (SCV)

Scientific Python

1., 1., 1.])

October 2012

17 / 76

Array Creation

makes array([ 1., 1., 1., 1., 1.])

np.zeros same thing for zeroes
np.zeros((2,2)) makes array([[0., 0.], [0., 0.]])
that is a 2D, 2x2 array of zeros
np.ones(5)

Yann - [email protected] (SCV)

Scientific Python

October 2012

17 / 76

Array Creation

np.arange,

numpys version of built-in range method

np.linspace is similar to np.arange but you pass the
number of elements, not step size

Yann - [email protected] (SCV)

Scientific Python

October 2012

18 / 76

Array Basics - Results

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

>>> import numpy as np

>>> x = np . array (((1 , 2 , 3) , (4 , 5 , 6)))
>>> x . size # total number of elements
6
>>> x . ndim # number of dimensions
2
>>> x . shape # number of elements in each dimension
(2 L , 3 L )
>>> x [1 ,2] # first index is the rows , then the column
6
>>> x [1] # give me 1 row
array ([4 , 5 , 6])
>>> x [1][2]
6
>>> x . dtype
dtype ( int32 )

Yann - [email protected] (SCV)

Scientific Python

October 2012

19 / 76

Basic Operations

14 x = np . array ([0 , np . pi /4 , np . pi /2])

15 np . sin ( x )
16 np . dot ([2 , 2 , 2] , [2 , 2 , 2])

Yann - [email protected] (SCV)

Scientific Python

October 2012

20 / 76

Basic Operations

14 x = np . array ([0 , np . pi /4 , np . pi /2])

15 np . sin ( x )
16 np . dot ([2 , 2 , 2] , [2 , 2 , 2])
22
23
24
25
26

>>> x = np . array ([0 , np . pi /4 , np . pi /2])

>>> np . sin ( x )
array ([ 0.
, 0.70710678 , 1.
>>> np . dot ([2 , 2 , 2] , [2 , 2 , 2])
12

Yann - [email protected] (SCV)

Scientific Python

])

October 2012

20 / 76

Its a Simple Game

Yann - [email protected] (SCV)

Scientific Python

October 2012

21 / 76

Calculating Neighbors

Yann - [email protected] (SCV)

Scientific Python

October 2012

22 / 76

Use the Neighbors to Update

Yann - [email protected] (SCV)

Scientific Python

October 2012

23 / 76

A Practice Problem
In the
1

practice

conway.py

directory:
is a program; try it out.

Yann - [email protected] (SCV)

Scientific Python

October 2012

24 / 76

A Practice Problem
directory:
1
conway.py is a program; try it out.
2
conway work.py is an assignment; try to do it.
3
test conway.py is a program to test your work.
Remember, the shape attribute:
In the

practice

1 a = np . array (((1 ,2) ,(3 ,4) ,(5 ,6))) # make a 3 x2

2 print a . shape [0] # prints 3
3 print a . shape [1] # prints 2

Yann - [email protected] (SCV)

Scientific Python

October 2012

24 / 76

Conways Game of Life

np.roll(a, 1, axis=0)

returns a new array that cycles the

values.
1
2
3
4
5
6
7
8

>>> a
array ([[0 , 1 , 2] ,
[3 , 4 , 5] ,
[6 , 7 , 8]])
>>> np . roll (a , 1 , axis =0)
array ([[6 , 7 , 8] ,
[0 , 1 , 2] ,
[3 , 4 , 5]])

This can be used to make the

Yann - [email protected] (SCV)

neighbors

Scientific Python

function.

October 2012

25 / 76

Conways Game of Life

np.where(b ==0)
np.where(b ==0,
1
2
3
4
5
6
7
8
9
10

returns a series of indicies where its true.

-1, b) to replace 0s with -1s.

>>> b
array ([[ 0. , 0. , 0.] ,
[ 2. , 2. , 2.] ,
[ 0. , 3. , 0.]])
>>> np . where ( b == 0)
( array ([0 , 0 , 0 , 2 , 2]) , array ([0 , 1 , 2 , 0 , 2]))
>>> np . where ( b == 0 , -1 , b )
array ([[ -1. , -1. , -1.] ,
[ 2. , 2. , 2.] ,
[ -1. , 3. , -1.]])

Look up the np.logical or method; this and

be used to redefine the update function.
Yann - [email protected] (SCV)

Scientific Python

np.where

October 2012

can
26 / 76

Matplotlib Imshow

0
1
2
3
4
5
6
7

import matplotlib . pyplot as plt

import numpy as np
a = np . zeros ((3 ,3))
a [1 ,1] = 1
plt . imshow (a ,
interpolation = nearest ,
cmap = plt . winter () ,
origin = lower )

Yann - [email protected] (SCV)

Scientific Python

October 2012

27 / 76

Nearest Neighbors

Yann - [email protected] (SCV)

Scientific Python

October 2012

28 / 76

Matplotlib Imshow

11 plt . imshow (a ,
12 #
interpolation = nearest ,
13
cmap = plt . winter () ,
14
origin = lower )
15 plt . show ()

Yann - [email protected] (SCV)

Scientific Python

October 2012

29 / 76

Default: usually bilinear

Yann - [email protected] (SCV)

Scientific Python

October 2012

30 / 76

Plenty More

www.scipy.org/Tentative_NumPy_Tutorial
Universal Functions section
PyTrieste Numpy Tutorial

Yann - [email protected] (SCV)

Scientific Python

October 2012

31 / 76

scipy: Many Useful Scientific Tools

http://docs.scipy.org/doc/scipy/reference/
tutorial/index.html
Numpy
Integration
Optimization
Interpolation
FFT
Linear Algebra
Statistics
And much more....
Yann - [email protected] (SCV)

Scientific Python

October 2012

32 / 76

Constants
4
5
6
7
8
9
10
11
12
13
14
15
16
17

>>> from scipy import constants

>>> constants . c # speed of light
299792458.0
>>> # physical constants : value , units , uncertainty
>>> constants . p h y s i c a l _ c o n s t a n t s [ " electron mass " ]
(9.10938291 e -31 , kg , 4e -38)
>>> # look - up constants , only first 3 !
>>> masses = constants . find ( " mass " )[:3]
>>> for val in masses :
...
print " {} is available " . format ( val )
...
Planck mass is available
Planck mass energy equivalent in GeV is available
alpha particle mass is available

Yann - [email protected] (SCV)

Scientific Python

October 2012

33 / 76

Zombie Apocalypse - ODEINT

http://www.scipy.org/Cookbook/Zombie_
Apocalypse_ODEINT

Yann - [email protected] (SCV)

Scientific Python

October 2012

34 / 76

Zombie Apocalypse - ODEINT

Yann - [email protected] (SCV)

Scientific Python

October 2012

35 / 76

Zombie Apocalypse - ODEINT

Look at

examples\zombie.py

5 def calc_rate ( P =0 , d =0.0001 , B =0.0095 , G =0.0001 , A =0.0001 , S0 =500):

6
def f (y , t ):
7
Si = y [0]
8
Zi = y [1]
9
Ri = y [2]
10
# the model equations ( see Munz et al . 2009)
11
f0 = P - B * Si * Zi - d * Si
12
f1 = B * Si * Zi + G * Ri - A * Si * Zi
13
f2 = d * Si + A * Si * Zi - G * Ri
14
return [ f0 , f1 , f2 ]
15
Z0 = 0
# initial zombie population
16
R0 = 0
# initial death population
17
y0 = [ S0 , Z0 , R0 ]
# initial condition vector
18
t = np . linspace (0 , 5. , 1000)
# time grid
19
# solve the DEs
20
soln = odeint (f , y0 , t )
21
S = soln [: , 0]
22
Z = soln [: , 1]
23
R = soln [: , 2]
24
return t , S , Z

Yann - [email protected] (SCV)

Scientific Python

October 2012

36 / 76

ipython - plotting

Amazing interactive python.

Using the -pylab flag, you can get plotting for free
1 % ipython - pylab
2 ...
3 In [1]: plot ( range (10))

Yann - [email protected] (SCV)

Scientific Python

October 2012

37 / 76

ipython - Magic Functions

Amazing interactive python.

It provids Magic functions:
cd - like unix change directory
ls - list directory
timeit - time execution of statement
and so on ...

Yann - [email protected] (SCV)

Scientific Python

October 2012

38 / 76

ipython - Logging
Amazing interactive python.
You can log an interactive session:
1 In [1]: logstart mylogfile . py

Everything you type will be recorded to mylogfile.py

logon and logoff turn it on and off
logstop turns it off and saves file
To re-run it in ipython, us the run command:
1 In [1]: run -i mylogfile . py

Yann - [email protected] (SCV)

Scientific Python

October 2012

39 / 76

ipython - Misc.
Amazing interactive python.
You can use ? instead of help().
In [1]: len? and Enter will print help
In fact you get even more information than the
standard help()
Tab-completion
Start typing a method, and pop-up help is shown
You need a newer version than Katanas, and
you need to type ipython qtconsole

Yann - [email protected] (SCV)

Scientific Python

October 2012

40 / 76

Simple Parallelized Python

is a module that allows you to easily start

other processes. They do not share memory, but
multiprocess helps them communicate info/data.
multiprocessing

Yann - [email protected] (SCV)

Scientific Python

October 2012

41 / 76

Pool
creates a pool of processes that you
give tasks to perform. See in
multiprocessing.Pool

examples/mp_vowels_pool.py
16
17
18
19
20
21

nproc = multiprocessing . cpu_count ()

task_pool = multiprocessing . Pool (
processes = nproc ,
initializer = start_proc )
words = " Mary had a little lamb " . split ()

Yann - [email protected] (SCV)

Scientific Python

October 2012

42 / 76

map

takes a function and a list of arguments,

and applies that function to each argument, in parallel.
task pool.map

24

map_results = task_pool . map ( count_vowels , words )

The map method is blocking. The code progresses when

all tasks are complete.

Yann - [email protected] (SCV)

Scientific Python

October 2012

43 / 76

apply async
takes a function and a sequence of
arguments and applies that function to that argument.
task pool.apply async

27
28
29
30
31
32
33
34

async_results = []
for word in words :
async_results . append (
task_pool . apply_async ( count_vowels ,
( word ,)))
task_pool . close () # no more tasks
task_pool . join () # wait here until all

This is not blocking

Yann - [email protected] (SCV)

Scientific Python

October 2012

44 / 76

apply async, getting results

task pool.map
39
40
41

just returns a list of results

for word , nvowels in zip ( words , map_results ):

print " { word : s } has { nvowels : d } vowels " . format ( word = word ,
nvowels = nvowels )

But task pool.apply async returns an object from which

you can get results.
45
46
47

for word , result in zip ( words , async_results ):

print " { word : s } has { nvowels : d } vowels " . format ( word = word ,
nvowels = result . get ())

Yann - [email protected] (SCV)

Scientific Python

October 2012

45 / 76

Outline for section 2

1

Scientific Python

Beyond the Basics

Super Scripting
More on Functions
String Formatting
Files
External Programs
Command line arguments

Solving Laplaces Equation

Yann - [email protected] (SCV)

Scientific Python

October 2012

46 / 76

pass

is pythons way of saying

Keep moving, nothing to see here....

pass

Yann - [email protected] (SCV)

Scientific Python

October 2012

47 / 76

pass

is pythons way of saying

Keep moving, nothing to see here....
Its used for yet unwritten function.
If you try to call a function that doesnt exist, you
get an error.

pass

Yann - [email protected] (SCV)

Scientific Python

October 2012

47 / 76

pass

is pythons way of saying

Keep moving, nothing to see here....
Its used for yet unwritten function.
If you try to call a function that doesnt exist, you
get an error.
pass just creates a function that does nothing.
Great for planning work!

pass

1 def step1 ():

2
pass
3 step1 () # no error !
4 step2 () # error , darn !

Yann - [email protected] (SCV)

Scientific Python

October 2012

47 / 76

None

is Pythons formal value for nothing

Use this as a default value for a variable,
or as a return value when things dont work,
and you dont want a catastrophic error.
None

Yann - [email protected] (SCV)

Scientific Python

October 2012

48 / 76

None

is Pythons formal value for nothing

Use this as a default value for a variable,
or as a return value when things dont work,
and you dont want a catastrophic error.
Test if something is None, to see if you need to handle
these special cases
None

1 name = None
2 if name is None :
3
name = Johnny

Yann - [email protected] (SCV)

Scientific Python

October 2012

48 / 76

"__main__"

Every module has a name, which is stored in __name__

The script/console that you are running is called
"__main__".

Yann - [email protected] (SCV)

Scientific Python

October 2012

49 / 76

"__main__"

Every module has a name, which is stored in __name__

The script/console that you are running is called
"__main__".
Use this to make a file both a module and a script.
1
2
3
4
5
6
7

# module greeting . py
def hey_there ( name ):
print " Hi ! " , name , " ... How s it going ? "
hey_there ( Joey )
if __name__ == " __main__ " :
hey_there ( Timmy )

python greeting.py
Yann - [email protected] (SCV)

Hi! Timmy ... Hows it going?

Scientific Python

October 2012

49 / 76

The Docstring
This is an un-assigned string that is used for
documentation.
It can be at the top of a file, documenting a script
or module
or the first thing inside a function, documenting it.

Yann - [email protected] (SCV)

Scientific Python

October 2012

50 / 76

The Docstring
This is an un-assigned string that is used for
documentation.
It can be at the top of a file, documenting a script
or module
or the first thing inside a function, documenting it.
This is what help() uses...
1 def dot_product ( v1 , v2 ):
2
""" Perform the dot product of v1 and v2 .
3
4
v1 is a three element vector .
5
v2 is a three element vector .
6
"""
7
sum = 0 # ....
Yann - [email protected] (SCV)

Scientific Python

October 2012

50 / 76

Keyword Arguments

Arguments that have default values.

In the function signature, keyword=default_value
None is a good default if you want to make a decision
1 def hello ( name = Joe , repeat = None ):
2
if repeat is None :
3
repeat = len ( name )
4
print Hi ! , name * repeat

Yann - [email protected] (SCV)

Scientific Python

October 2012

51 / 76

Returning Values
A function can return multiple values.
Formally this creates a Tuple - an immutable list

Yann - [email protected] (SCV)

Scientific Python

October 2012

52 / 76

Returning Values
A function can return multiple values.
Formally this creates a Tuple - an immutable list
You can collect the returned values as a Tuple, or as
individual values.
1
2
3
4
5

def pows ( val ):

return val , val * val , val * val * val
two , four , eight = pows (2)
ones = pows (1)
print ones [0] , ones [1] , ones [2] # 1 1 1

Yann - [email protected] (SCV)

Scientific Python

October 2012

52 / 76

string.format()

Strings can be paired with values to control printing.

4
5
6
7
8
9

>>> " Hi there {}! " . format ( Yann )

Hi there Yann !
>>> " Coords : {} , {}; " . format ( -1 , 2)
Coords : -1 , 2;
>>> " {1} , the cost is {0:.2 f } " . format (1125.747025 , Y
Yann , the cost is 1125.75

Yann - [email protected] (SCV)

Scientific Python

October 2012

53 / 76

Keyword arguments to format

Its sometimes tricky to format many values
You can name some of the format targets
9
10
11
12
13
14
15
16

email = """
Subject : { subject }
Date : { mon :2 d }/{ day :2 d }/{ year :2 d }
Message : { msg }
"""
print email . format ( mon =10 , year =12 , day =31 ,
subject = Happy Halloween ,
msg = Booh )

Yann - [email protected] (SCV)

Scientific Python

October 2012

54 / 76

Keyword arguments to format result

13
14
15
16
17
18
19
20
21
22
23
24

>>>
...
...
...
...
>>>
...
...

email = " " "

Subject : { subject }
Date : { mon :2 d }/{ day :2 d }/{ year :2 d }
Message : { msg }
"""
print email . format ( mon =10 , year =12 , day =31 ,
subject = Happy Halloween ,
msg = Booh )

Subject : Happy Halloween

Date : 10/31/12
Message : Booh

More features at:

http://docs.python.org/library/string.html
Yann - [email protected] (SCV)

Scientific Python

October 2012

55 / 76

csv reader and writer

For reading and writing comma-separated-values
csv.reader for reading, csv.writer for writing
Dialects option correspond to predefined formats
excel for excel output without needing to know the
separator and quote characters
1
2
3
4
5

reader = csv . reader ( file )

for row in reader :
# row is a list
writer = csv . writer ( file )
writer . writerow ([1 ,2 ,3])

Yann - [email protected] (SCV)

Scientific Python

October 2012

56 / 76

subprocess

running external programs

1 import subprocess
2 output = subprocess . call ([ ls , -1 ])
3 print " output = " , output

provides many tools

The most basic is the call function.
It takes a list that is joined and executed.
output just holds the exit code (0 if successful)
check_output is like call but 1) returns output and 2)
causes an error if program fails
subprocess

Yann - [email protected] (SCV)

Scientific Python

October 2012

57 / 76

argparse

easy command line arguments

module is the easiest way.

You first create a ArgumentParser object
You define the allowable arguments with the
add_argument function
You can add required or optional arguments
sys.argv is a list of arguments passed to the
program/script.
Pass this list to parse_args function to process and
get an object with your parameters defined.
argparse

Yann - [email protected] (SCV)

Scientific Python

October 2012

58 / 76

argparse

example

Look over
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

examples\argparse_ex.py

import sys
import argparse
parser = argparse . ArgumentParser ( description = Plot zombie statistics )
parser . add_argument ( prefix , help = the prefix for the plot filename )
parser . add_argument ( -p , dest = pop , default =500 , type = int ,
help = Set the startring population )
parser . add_argument ( -s , dest = show ,
help = Show the figure , action = store_true )
parser . add_argument ( city , help = Plot information for a specific city ,
nargs = ? , default = None )
args = sys . argv [1:]
params = parser . parse_args ( args )
print " prefix = " , params . prefix
print " pop = " , params . pop
if params . city is not None :
print " city = " , params . city
print " show ? "
if params . show :
print " yes ! "
else :
print " no : -( "

Yann - [email protected] (SCV)

Scientific Python

October 2012

59 / 76

Solve Laplaces Equation 1a

Solve 2 u = 0

Yann - [email protected] (SCV)

Scientific Python

October 2012

60 / 76

Solve Laplaces Equation 1a

Solve 2 u = 0
Solve iteratively, each time changing u with
following equation:
n+1
n+1
n+1
n
n
uj,l
= 1/4(uj+1,l
+ uj1,l
+ uj,l+1
+ uj,l1
)
Just an averages of the neighboring points.

Yann - [email protected] (SCV)

Scientific Python

October 2012

60 / 76

Solve Laplaces Equation 1a

Solve 2 u = 0
Solve iteratively, each time changing u with
following equation:
n+1
n+1
n+1
n
n
uj,l
= 1/4(uj+1,l
+ uj1,l
+ uj,l+1
+ uj,l1
)
Just an averages of the neighboring points.

Yann - [email protected] (SCV)

Scientific Python

October 2012

60 / 76

Solve Laplaces Equation 1b

Solve 2 u = 0
Solve iteratively, each time changing u with
following equation:
n+1
n+1
n+1
n
n
uj,l
= 1/4(uj+1,l
+ uj1,l
+ uj,l+1
+ uj,l1
)
Just an averages of the neighboring points.

Yann - [email protected] (SCV)

Scientific Python

October 2012

61 / 76

Solve Laplaces Equation 1c

Solve 2 u = 0
Solve iteratively, each time changing u with
following equation:
n+1
n+1
n+1
n
n
uj,l
= 1/4(uj+1,l
+ uj1,l
+ uj,l+1
+ uj,l1
)
Just an averages of the neighboring points.

Yann - [email protected] (SCV)

Scientific Python

October 2012

62 / 76

Solve Laplaces Equation 1d

Solve 2 u = 0
Solve iteratively, each time changing u with
following equation:
n+1
n+1
n+1
n
n
uj,l
= 1/4(uj+1,l
+ uj1,l
+ uj,l+1
+ uj,l1
)
Just an average of the neighboring points.
Repeat this calculation until rms(u n+1 u n ) < ,
some threshold
Set some limit on total number of iterations
practice/laplace.py

Yann - [email protected] (SCV)

Scientific Python

October 2012

63 / 76

Solve Laplaces Equation 2

First task:

19 def pu r e _ p y t h o n _ s t e p ( u ):
20
Pure python imple mentati on of Gauss - Siedel method . Performs
21
iteration .
22
rms_err = 0
23
# use for loop to loop through each element in u
24
#
temporarily store old value to later calculate rms_err
25
#
update current u using 4 point averaging
26
#
update running value of rms_err
27
# when done looping complete rms_err calculation and return it
28
return rms_err

Yann - [email protected] (SCV)

Scientific Python

October 2012

64 / 76

Slicing - 1

19
20
21
22
23
24
25

x = np . array ((1 , 2 , 3 , 4 , 5 , 6))

x [2]
x [2:5]
x [2: -1]
x [:5]
x [:5:2] = 10
x

Yann - [email protected] (SCV)

Scientific Python

October 2012

65 / 76

Slicing - 2
29
30
31
32
33
34
35
36
37
38
39
40

>>> x = np . array ((1 , 2 , 3 , 4 , 5 , 6))

>>> x [2]
3
>>> x [2:5]
array ([3 , 4 , 5])
>>> x [2: -1]
array ([3 , 4 , 5])
>>> x [:5]
array ([1 , 2 , 3 , 4 , 5])
>>> x [:5:2] = 10
>>> x
array ([10 , 2 , 10 , 4 , 10 , 6])

Yann - [email protected] (SCV)

Scientific Python

October 2012

66 / 76

Array Copying
43
44
45
46
47
48
49
50
51
52
53
54
55

>>> a = np . array ((1 ,2 ,3 ,4))

>>> b = a
>>> b is a
True
>>> c = a . view ()
>>> c . shape = 2 ,2
>>> c [1 ,1]=10
>>> c
array ([[ 1 , 2] ,
[ 3 , 10]])
>>> a
array ([ 1 , 2 , 3 , 10])
>>> d = a . copy ()

Yann - [email protected] (SCV)

Scientific Python

October 2012

67 / 76

Solve Laplaces Equation 3a

Yann - [email protected] (SCV)

Scientific Python

October 2012

68 / 76

Solve Laplaces Equation 3b

Yann - [email protected] (SCV)

Scientific Python

October 2012

69 / 76

Solve Laplaces Equation 3c

Yann - [email protected] (SCV)

Scientific Python

October 2012

70 / 76

Solve Laplaces Equation 3d

Second task:

35 def numpy_step ( u ):
36
Numpy based Jacobi s method . Performs one iteration .
37
# make a copy so that you can calculate the error
38
u_old = u . copy ()
39
# use slicing to shift array
40
# utmp = u [1: -1 , 1: -1] makes a new array , so that utmp [0 ,0] is
41
# as u [1 ,1]
42
# then
43
# utmp = u [0: -2 , 1: -1] makes a new array that leads to a shift
44
# because utmp [0 ,0] is the same as u [0 , 1]
45
# use this concept to solve this equation in on line
46
# u = 1/4*( u_ {j -1 , i } + u_ { j +1 , i } + u_ {j , i -1} + u_ {j , i +1})
47
return calc_err (u , u_old )

Yann - [email protected] (SCV)

Scientific Python

October 2012

71 / 76

Solve Laplaces Equation 4a

Third task:

4 def laplace _driver (u , stepper , maxit =100 , err =1 e3 ):

5
Repeatedly call stepper ( u ) to solve nabla ^2 u = 0 until
6
rms error < err or maxit number of iterations is reached .
7
8
u - a numpy array
9
stepper - a function whose sole argument is u
10

11
rms_err = 0
12
# take one step with stepper , to define initial rms_err
13
# loop until rms_err < err
14
#
check to see that number of iterations is less than maxi
15
#
perform single iteration using stepper method
16
# return rms_error
17
return rms_err

Yann - [email protected] (SCV)

Scientific Python

October 2012

72 / 76

Solve Laplaces Equation 4b

59 def time_method ( stepper ):

60
Time how long a particular stepper takes to solve an ideal
61
u = set_bc ( np . zeros ((100 ,100)))
62
start = time . time ()
63
err = la place_dr iver (u , stepper )
64
return time . time () - start
65
66 if __name__ == " __main__ " :
67
p u r e _ p y t h o n _ t i m e = time_method ( p u r e _ p y t h o n _ s t e p )
68
numpy_time = time_method ( numpy_step )
69
print " Pure python method takes {:.3 f } seconds " . format ( p u r e _ p y
70
print " Numpy method takes {:.3 f } seconds " . format ( numpy_time )

Yann - [email protected] (SCV)

Scientific Python

October 2012

73 / 76

Solve Laplaces Results

1 Pure python method takes 3.624 seconds

2 Numpy method takes 0.016 seconds

Yann - [email protected] (SCV)

Scientific Python

October 2012

74 / 76

Resources

SCV help site

Numpy Tutorial
Scipy tutorial
Scientific Python Tutorial
Another Scientific Python Tutorial
iPython Tutorial

Yann - [email protected] (SCV)

Scientific Python

October 2012

75 / 76

What I Didnt Cover

Exception Handling
Duck Typing
Virtualenv
Pandas
mpi4py
And many other useful things

Yann - [email protected] (SCV)

Scientific Python

October 2012

76 / 76