%%HTML
<link rel="stylesheet" type="text/css" href="https://raw.githubusercontent.com/malkaguillot/Foundations-in-Data-Science-and-Machine-Learning/refs/heads/main/docs/utils/custom.css">
%%HTML
<link rel="stylesheet" type="text/css" href="../utils/custom.css">

%%HTML

Foundations in Data Science and Machine Learning

Module 4: More Python

Auguste Debroise

for loops (Concept 1)

Don’t repeat yourself

names = ["Guy", "Ray", "Tim"]
lower_names = [
 names[0].lower(),
 names[1].lower(),
 names[2].lower(),
]
lower_names

['guy', 'ray', 'tim']

• This code repetition is problematic

  • If we have a typo, we need to fix it multiple times
  • Cumbersome if list becomes longer

• In many situations we want to do similar things multiple times

  • Cleaning several similar variables
  • Fitting several models

A simple for loop

• for loops let us do things repeatedly
• First line ends with a :
• In each iteration, the running variable is bound to a new value
• Loop body with one or several lines is indented by 4 spaces

for i in range(5):
    print(i ** 2)

0
1
4
9
16

Looping over lists and tuples

• Looping over lists and tuples works in the same way
• Running variable is iteratively bound to the iterable’s elements
• Try to choose a good name for the running variable!

names = ["Guy", "Ray", "Tim"]
for name in names:
    print(name.lower())

guy
ray
tim

Looping over dictionaries

• By default you loop over dictionary keys

let_to_pos = {
    "a": 0,
    "b": 1,
    "c": 2,
}
for let in let_to_pos:
    print(let)

a
b
c

• Use .items() for looping over key/value pairs

for let, pos in let_to_pos.items():
    print(let, pos)

a 0
b 1
c 2

if statements (Concept 2)

Motivation

• So far, all of our instructions in Python were very explicit
• There was no way of reacting to different situations:
  • Collecting elements of a list that fulfil a condition
  • Doing different things for different types of variables
  • …
• This is what if conditions are for

Example: clipping a number

• if ,elif , and else are special keywords
• End each condition with a :
• What happens if that condition is True needs to be indented by 4 spaces and can span one or multiple lines
• Code following False conditions is skipped
• elif x: is the same as
• else: + nested if x:

number = -3.1

if number < -3:
    clipped = -3.0
elif number > 3:
    clipped = 3.0
else:
    clipped = number

clipped

-3.0

More on Booleans

• That is not a Boolean can be converted to a Boolean
• This conversion happens implicitly after if and elif
• Can be useful and elegant but might compromise readability
• Rules of thumb:
  • 0 is False -ish

bool(0)

False

• Other numbers are True -ish

bool(1)

True

• Len-0 collections are False -ish

bool([])

False

• Len>0 collections are True -ish

bool([1, 3])

True

More complex conditions

• Remember operators from "Assignments and Scalar Types":
  • and
  • or (inclusive)
  • not
• Example:

a = 3 
b = 2
some_cutoff = 1
if a > b and b > some_cutoff:
    print("do_something()")
else:
    print("do_something_else()")

do_something()

Filtering loops

• Can filter lists based on properties of items
• Can filter dictionaries based on properties of keys and/or values
• Example usecases:
  • Find elements above a cutoff
  • Extract female names
  • Exclude invalid data

names = ["Guy", "Ray", "Tim"]
names_with_i = []
for n in names:
    if "i" in n:
        names_with_i.append(n)
names_with_i

['Tim']

Defining Functions (Concept 3)

Anatomy of Python functions

• Start with the def keyword
• Name is lowercase_with_underscores
• There can be one or several parameters (a.k.a. arguments)
• You can assign default values for arguments
• Function body is indented by 4 spaces and can have one or several lines
• Inside the body you can do everything you have seen so far!

)

Example

• Function calls work with positional and keyword arguments
• Pass keyword arguments for any function with more than one argument!

def utility_crra(c, y=1.5):
    return c ** (1 - y) / (1 - y)

utility_crra(1.0)

-2.0

utility_crra(1.0,  y=0)

1.0

Principles for Good Functions

Why functions are important

• Help to re-use code and avoid duplication
• Help to structure code and reduce cognitive load
• Make individual code snippets testable
• Help to make your projects more reproducible
• Unlock the power of functional programming concepts
• Are also the basis for good object oriented code

Pass all variables you want to use inside

• Inside a function you have access to variables in the enclosing scope
• This is dangerous because the behaviour of the function now depends on global variables
• Do not use this in your code!

# bad example
global_msg = "Hello {}!"
def greet_with_global(name):
    print(global_msg.format(name))
greet_with_global("Guido")

Hello Guido!

# solution 1: define inside function
def greet(name):
    msg = "Hello {}!"
    print(msg.format(name))
greet("Guido")

Hello Guido!

# solution 2: pass as argument
def greet_explicit(name, msg):
    print(msg.format(name))
greet_explicit("Guido", "Hello {}!")

Hello Guido!

Do not modify mutable arguments

• Arguments are passed by reference, i.e. without making a copy
• Make sure that functions do not modify mutable arguments!
  • Make copies
  • Avoid changing objects in the first place

def append_4(some_list):
    some_list.append(4)
    return some_list
my_list = [1, 2, 3]
append_4(my_list)   
my_list

[1, 2, 3, 4]

# better solution
def append_4(some_list):
    out = some_list.copy()
    out.append(4)
    return out

Dealing with files and path using pathlib

Why do we need pathlib?

• There are many ways to work with file paths in Python
• Some are not portable
• We want to give you one way that is guaranteed to work!

What you should not do

import pandas as pd
path = "C:\Users\xyz\Documents\python\lectures\03-more-python\data\iris.csv"
data = pd.read_csv(path)

• This only works on one Computer: it is an absolute path
• Backslashes ( \ ) only work on Windows
• Warning: This is what you get when you copy a path from your file explorer

What you should do

• Start paths relative to the root folder of the project
• Only make assumptions about directory structure inside the project
• Define the path in a way that is portable across operating systems

Get a path to the project root

Path(".") gives a relative path to current directory

from pathlib import Path
# get a path to the current directory
this_dir = Path(".")
print(this_dir)

.

.resolve() makes it absolute for readability

this_dir = this_dir.resolve()
print(this_dir)

/Users/malka/Dropbox/teaching-uliege/Foundations-in-Data-Science-and-Machine-Learning/docs/m3

.parent moves up one file/directory

# move up to the parent directory
root = this_dir.parent
print(root)

/Users/malka/Dropbox/teaching-uliege/Foundations-in-Data-Science-and-Machine-Learning/docs

• The output differs on every computer!
• No assumptions made on usernames or folders outside the project

Get a path to the project root

In a Python script, you can use the following code to get a path to the project root:

from pathlib import Path
# get a path to the current file
this_file = Path(__file__)
print("this_file", this_file)
# move up several times (here twice) to the project root
root = this_file.parent.parent
print("root", root)

• In a .py and a notebook file Path() would lead us to the current directory of the shell from which the file was executed
• The file variable is a magic variable with the path to the current file

# IN my notebook, I have a directory structure like this:
root = this_dir.parent.parent
print(root)

/Users/malka/Dropbox/teaching-uliege/Foundations-in-Data-Science-and-Machine-Learning

From the project root to the data

• Once root is defined, the rest works the same in notebooks and .py files
• Concatenate different path snippets with /
• Resulting path works on all platforms!

data_path = root / "data" / "iris.csv"
print("data", data_path)
print(data_path.exists())

data /Users/malka/Dropbox/teaching-uliege/Foundations-in-Data-Science-and-Machine-Learning/data/iris.csv
True

File path rules

1. Always use pathlib Path objects instead of strings
2. Do not hardcode any parts of a path outside of the project’s directory
3. Always concatenate paths with /

Remember:

If you copy paste a path from your Windows File Explorer, all three rules are violated!