Lecture 5 – Functions and Control

Data 6, Summer 2021

Motivation

We've seen a few in-built Python functions so far.

int('-14')              # Evaluates to -14

-14

abs(-14)                # Evaluates to 14

14

max(-14, 15)            # Evaluates to 15

15

print('zoology')        # Prints zoology, evaluates to None

zoology

We don't currently have a good way to prevent our code from getting repetitive. For example, if we want to determine whether or not different students are ready to graduate:

units_1 = 104
year_1 = 'sophomore'
ready_to_graduate_1 = (year_1 == 'senior') and (units_1 >= 120)
ready_to_graduate_1

False

units_2 = 121
year_2 = 'senior'
ready_to_graduate_2 = (year_2 == 'senior') and (units_2 >= 120)
ready_to_graduate_2

True

units_3 = 125
year_3 = 'junior'
ready_to_graduate_3 = (year_3 == 'senior') and (units_3 >= 120)
ready_to_graduate_3

False

Functions

Here's a better solution:

def ready_to_graduate(year, units):
    return (year == 'senior') and (units >= 120)

ready_to_graduate(year_1, units_1)

False

ready_to_graduate(year_2, units_2)

True

ready_to_graduate(year_3, units_3)

False

By using a function, we only had to write out the logic once, and could easily call it any number of times.

Other function examples:

# This function has one parameter, x.
# When we call the function, the value we pass in
# as an argument will replace x in the computation.

def triple(x):
    return x*3

triple(15)

45

triple(-1.0)

-3.0

# Functions can have zero parameters!
def always_true():
    return True

# The body of a function can be
# longer than one line.
def pythagorean(a, b):
    c_squared = a**2 + b**2
    return c_squared**0.5

always_true()

True

# Good
def square(x):
    return x**2

# Bad
def square(x):
return x**2

  File "/tmp/ipykernel_142/4284077287.py", line 3
    return x**2
    ^
IndentationError: expected an indented block

Quick Check 4

def mystery(t):
    return t + '0'

alpha = mystery('19')
beta = mystery(19)
charlie = mystery('1' + '9')
(alpha, beta, charlie)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
/tmp/ipykernel_142/809340847.py in <module>
      1 alpha = mystery('19')
----> 2 beta = mystery(19)
      3 charlie = mystery('1' + '9')
      4 (alpha, beta, charlie)

/tmp/ipykernel_142/954243894.py in mystery(t)
      1 def mystery(t):
----> 2     return t + '0'

TypeError: unsupported operand type(s) for +: 'int' and 'str'

Parameters and return values

Scoping

def eat(zebra):
    return 'ate ' + zebra

eat('lionel')

'ate lionel'

zebra

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
/tmp/ipykernel_142/1378667599.py in <module>
----> 1 zebra

NameError: name 'zebra' is not defined

N = 15
def half(N):
    return N/2

half(0)

0.0

half(12)

6.0

half(N)

7.5

N = 15
def addN(x):
    return x + N

addN(0)

15

addN(3)

18

triple(15)

45

triple(1/0)

---------------------------------------------------------------------------
ZeroDivisionError                         Traceback (most recent call last)
/tmp/ipykernel_142/2073793153.py in <module>
----> 1 triple(1/0)

ZeroDivisionError: division by zero

triple(3, 4)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
/tmp/ipykernel_142/3999475957.py in <module>
----> 1 triple(3, 4)

TypeError: triple() takes 1 positional argument but 2 were given

print('my', 'name', 'is', 300)

my name is 300

Returning

def add_and_print(a, b):
    total = a + b
    print(total)

total = add_and_print(3, 4)

7

total

print(total)

None

Nothing after the return keyword is run.

def odd(n):
    return n % 2 == 1
    print('this will never be printed!')

odd(15)

True

odd(2)

False

Quick Check 5

total = 3
def square_and_cube(a, b):
    return a**2 + total**b
total = square_and_cube(1, 2)
total

10

total = square_and_cube(1, 2)
total

101

String methods

'ian'.upper()

'IAN'

s = 'JuNiOR12'
s.upper()

'JUNIOR12'

s.lower()

'junior12'

s.replace('i', 'iii')

'JuNiiiOR12'

Demo

Let's load in the same Wikipedia countries data from this week's earlier lectures. But this time, we will write some of the data cleaning functions ourself.

from datascience import *
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
import numpy as np

data = Table.read_table('data/countries.csv')
data = data.take(np.arange(0, data.num_rows - 1))
data = data.relabeled('Country(or dependent territory)', 'Country') \
           .relabeled('% of world', '%') \
           .relabeled('Source(official or UN)', 'Source')
data = data.with_columns(
    'Country', data.apply(lambda s: s[:s.index('[')].lower() if '[' in s else s.lower(), 'Country'))

def first_letter(s):
    return s[0]

def last_letter(s):
    return s[-1]

data

Rank	Country	Population	%	Date	Source
1	china	1,405,936,040	17.9%	27 Dec 2020	National population clock[3]
2	india	1,371,366,679	17.5%	27 Dec 2020	National population clock[4]
3	united states	330,888,778	4.22%	27 Dec 2020	National population clock[5]
4	indonesia	269,603,400	3.44%	1 Jul 2020	National annual projection[6]
5	pakistan	220,892,331	2.82%	1 Jul 2020	UN Projection[2]
6	brazil	212,523,810	2.71%	27 Dec 2020	National population clock[7]
7	nigeria	206,139,587	2.63%	1 Jul 2020	UN Projection[2]
8	bangladesh	169,885,314	2.17%	27 Dec 2020	National population clock[8]
9	russia	146,748,590	1.87%	1 Jan 2020	National annual estimate[9]
10	mexico	127,792,286	1.63%	1 Jul 2020	National annual projection[10]

... (231 rows omitted)

Let's look at the 'Population' column.

# ignore
china_pop = data.column('Population').take(0)

china_pop

'1,405,936,040'

We want these numbers to be integers, so that we can do arithmetic with them or plot them. However, right now they are not.

Let's write a function that takes in a string with that format, and returns the corresponding integer. But first, proof that the int function doesn't work here (it doesn't like the commas):

int(china_pop)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_142/2345823718.py in <module>
----> 1 int(china_pop)

ValueError: invalid literal for int() with base 10: '1,405,936,040'

china_pop

'1,405,936,040'

def clean_population_string(pop):
    no_comma = pop.replace(',', '')
    return int(no_comma)

china_pop_clean = clean_population_string(china_pop)
china_pop_clean

1405936040

Cool!

Using techniques we haven't yet learned, we can apply this function to every element of the 'Population' column, so that when we visualize it, things work.

# ignore
data = data.with_columns('Population', data.apply(clean_population_string, 'Population'))

data

Rank	Country	Population	%	Date	Source
1	china	1405936040	17.9%	27 Dec 2020	National population clock[3]
2	india	1371366679	17.5%	27 Dec 2020	National population clock[4]
3	united states	330888778	4.22%	27 Dec 2020	National population clock[5]
4	indonesia	269603400	3.44%	1 Jul 2020	National annual projection[6]
5	pakistan	220892331	2.82%	1 Jul 2020	UN Projection[2]
6	brazil	212523810	2.71%	27 Dec 2020	National population clock[7]
7	nigeria	206139587	2.63%	1 Jul 2020	UN Projection[2]
8	bangladesh	169885314	2.17%	27 Dec 2020	National population clock[8]
9	russia	146748590	1.87%	1 Jan 2020	National annual estimate[9]
10	mexico	127792286	1.63%	1 Jul 2020	National annual projection[10]

... (231 rows omitted)

The '%' column is also a little fishy.

china_pct = data.column('%').take(0)
china_pct

'17.9%'

Percentages should be floats, but here they're strings.

Let's suppose we want to have the proportion of the total global population that lives in a given country as a column in our table. Proportions are decimals/fractions between 0 and 1. We can do this two ways:

• write a function, similar to clean_population_string, that correctly extracts the proportion we need
• calculate this by hand using all of the values in 'Population'

Let's do... both!

def clean_pct_string(pct):
    no_symbol = pct.replace('%', '')
    prop = float(no_symbol) / 100
    return prop

clean_pct_string(china_pct)

0.179

Nice! The other way requires adding together all of the values in the 'Population' column. We haven't covered how to do that just yet, so ignore the code for it and assume it does what it should.

total_population = data.column('Population').sum()
total_population

7710658195

Assume this is the total population of the world. How would you calculate the proportion of people living in one country?

def compute_proportion(population):
    return population / total_population

china_pop_clean

1405936040

compute_proportion(china_pop_clean)

0.18233670906482194

Pretty close to clean_pct_string(china_pct). The difference is likely due to some countries not being included in one column or the other.

Hopefully this gives you a glimpse of the power of functions!