Blog

Changing one line from CSV to Parquet makes reading the fares of ten million New York taxi trips 27 times faster, and the file is a sixth of the size. In this chapter, we measure why with Polars, DuckDB, PyArrow, and pandas: columnar storage, types in the file, the footer, encodings, pages, compression, predicate pushdown, and row group size. We also see where row storage still wins.

An array stores items of one type in equal slots next to each other in memory, but a Python list can grow and hold items of any type. In this chapter, we start from memory as one long row of bytes and build up arrays, the stack and the heap, pointers, dynamic arrays like Java’s ArrayList, CPython’s list, and linked lists. Then we reason through what each operation costs in an array, a list, and a linked list, and how those costs add up in a coding interview.

A function with one loop removes duplicates from a thousand items in 2.5 ms and from a hundred thousand in 27 seconds. In this chapter, we count steps instead of seconds to find out why: the growth rates from constant to factorial, the time and space complexity of every line, Big-O, Big Omega, and Big Theta with best, worst, and average cases, and amortized cost. We also turn a coding problem’s input size into a time budget.

A short story about what Pythonic might mean, an idea that involves a fair amount of interpretation and subjectivity.

In the age of AI, syntax by itself is the least important part of object-oriented programming to focus on. In this chapter, we cover classes, instance and class attributes, static and class methods, properties, encapsulation, inheritance, composition, and polymorphism against a problem from production code, a rate limiter for alerts. Along the way, we see what the SOLID principles warn about, what inheritance costs, and why composition is the more flexible approach.

Concurrency is arguably the hardest concept for Python developers, because the important ideas already assume you understand the operating system underneath. In this chapter, we start from the kernel and system calls, build up processes, threads, race conditions, and the GIL, then work out when to reach for asyncio, threading, or multiprocessing, why a GPU beats a CPU on the same arithmetic, and how to get better at concurrent programming.

Scarf ran Haskell in production for seven years and is now moving to Python.

From kindergarten to PhD, this may be the most comprehensive chapter on Python functions. We start with the basics of defining and calling functions, learn how to write them well, then learn closures, decorators, recursion, and generators, all the way to theory of computation and how CPython executes functions under the hood.

Every if, loop, break, and continue in Python compiles down to the same tricks. We visualize the Python bytecode behind the control flow to understand how Python really works under the hood.

Picking the right data type in Python starts with the basics. In Part 1 of Python’s data types, we cover NoneType, bool, int, float, and complex.