High Performance Python (2nd Edition)

This book is the Chinese translation of the second edition of O'Reilly's classic High Performance Python, written for Python 3, with 12 chapters in total. It does not teach introductory Python syntax; rather, it teaches you how to scientifically make your code "run fast" after it already "runs": first use profiling methods to find the real bottlenecks, then apply targeted remedies for three types of problems—CPU-bound, I/O-bound, and memory-bound. The book begins with computer architecture, explaining how computing units, storage hierarchies, and communication overhead determine the actual cost of Python code, then introduces profiling tools such as cProfile, SnakeViz, line_profiler, memory_profiler, PySpy, and bytecode disassembly, using the Julia set computation as a running case to gradually speed things up. The middle section discusses choosing between lists and tuples, dictionaries and sets, iterators and generators, NumPy matrix and vectorized computation, and using Cython to compile critical code into C extensions. The second half turns to concurrency and distribution: asyncio and asynchronous I/O, the multiprocessing module, clusters and job queues (including Docker deployment ideas), and reducing memory usage through means such as __slots__, chunking, array, and NumPy. The final chapter summarizes practical experience and lessons. This edition updates the toolchain and cases in the Python 3 ecosystem and is suitable as a systematic performance-tuning reference for readers with some foundation.