Python

Computational Economics @ Bundesbank

Pablo Winant

Why Python???

Python:

is popular
is free and opensource
has many many libraries
vibrant community
lots of online resources

Why Python? (3)

Historically, python was a glue language used to interoperate many low-level/system languages.

It has been increasingly used for web-development (cf django)

Nowadays it is the lingua franca of machine learning

Most major machine learning / deep learning libraries have python bindings

An example

def say_hello(name):
    """This function prints morning greetings"""

    print(f"Good morning {name}!\n")

    # we can import libraries
    import datetime
    t = datetime.datetime.now()

    # blocks are defined by indentation and colons
    if (t.hour,t.min) <= (9,15):
        print("All good?\n")
    else:
        print("Time to get started?\n")


say_hello("Pablo")

Python is everywhere

Windows
Linux
Web

The Python family

There are several flavours of Python:

Full Syntax
- CPython, PyPy, Pyston
Subset-Syntax
- micropython
- numba, pythran

Superset-Syntax
- mypy (*)
- cython
- mojo
Near-Syntax
- boo

subset-syntax: restrict functionalities (no classes, simpler objects) for easier compilation
superset-syntax: add type/memory information
near-syntax: different language that looks familiar

Examples

mojo:

fn greet2(name: String) -> String:
    return "Hello, " + name + "!"

cython

from libcpp.vector cimport vector

def primes(unsigned int nb_primes):

    cdef int n, i
    cdef vector[int] p
    p.reserve(nb_primes)  # allocate memory for 'nb_primes' elements.

    n = 2
    while p.size() < nb_primes:  # size() for vectors is similar to len()
        for i in p:
            if n % i == 0:
                break

Python is interpreted

Interpreted language

In an interpreted language, instructions are read and translated into processor instructions, one after another.

As consequence, it is:

flexible
- interactive development
- immediate feedback
slooooww ¹

Intepreters

Python
ipython a.k.a. jupyter
- send instructions to a kernel
- receive back MIME objects (with nice html representation)
VSCode
- has its own python kernel implementation
C API python.h
- julia
- your own…

Packages and Environment

Python modules

A file ending with .py is a python module

program.py

key = "low"
def something():
    return "hey"

The content from a module can be imported

from program import something

To import all objects in a module (functions, strings, …)

from program import *

Submodules

A folder containing modules and an __init.py__ is a package.
import a package or a submodule:
- import package
- from package.submodule import something
The content of modules and submodules is evaluated only once.¹
It is actually precompiled.
This is perfect for distributing a package.
Not so much to develop code interactively.

Package managers

Several ways to create / distribute python packages have been developped over the years.

setup.py, pip
setuptools, distutils, …
pipenv, poetry, …
conda

There are essentially two kinds of packages:

pip packages
conda packages

Pip packages

pip files (what are eggs btw)
- pure python
- binary
can be installed with pip install package
no dependency solving ! no proper uninstallation !
pip files a virtual evnironment created with venv
reproducible setup can be described in
- requirements.txt (old)
- pyproject.toml (new)
directory specific environments can be managed with poetry or venv:
- python -m venv directory

Conda environment

conda files
installed in a conda environment
with proper / reversible dependency solving
- very quick using mamba or micromamba
reproducible environment can be described in:
- environment.yml (dependencies)
- manifest (…)
directory specific environments can be managed with pixi

Syntax Review

Let’s setup the environment specified in requirements.txt

Move to python syntax tutorial