Why Python became the language of data

Python is small, and the ecosystem is large

Python itself doesn't know how to train a neural network, draw a graph, or read a spreadsheet. None of this is built-in. What Python has is an easy-to-read syntax and a huge ease of calling code written in other languages. This is what attracted those working with numerical computation: it was possible to write the heavy core in C or Fortran and use it through a simple interface in Python.

NumPy was born in the mid-2000s from the unification of two previous projects (Numeric and Numarray). The rest of the data stack was built on top of it:

None of these libraries are part of Python. Each one was written by people, published for free, and maintained by a community. You can check this in the interpreter itself:

import importlib.util
import sys

import numpy as np

print("numpy faz parte da biblioteca padrão?", "numpy" in sys.stdlib_module_names)
print("numpy está instalado aqui?", importlib.util.find_spec("numpy") is not None)
print(type(np.add).__name__, type(np.ndarray).__name__)

numpy faz parte da biblioteca padrão? False
numpy está instalado aqui? True
ufunc type

The np.add is an object of type ufunc, a compiled function that operates on integer arrays. The heavy lifting happens in this compiled code, and Python just orchestrates. This is the healthy way to think about the whole stack: Python is the glue, and the speed comes from the lower layers.

How much Python do I need to know beforehand?

Less than you think. Variables, lists, loops, functions, and import are enough to get started, and the rest you learn by using. No one memorizes an entire library. Those who work with data consult the documentation all the time, and what matters is knowing what is possible, to know what to look for.