Micha Gorelick - High Performance Python

• Background on the machinery of a computer so you know whats happening behind the scenes
• Lists and tuplesthe subtle semantic and speed differences in these fundamental data structures
• Dictionaries and setsmemory allocation strategies and access algorithms in these important data structures
• Iteratorshow to write in a more Pythonic way and open the door to infinite data streams using iteration
• Pure Python approacheshow to use Python and its modules effectively
• Matrices with numpyhow to use the beloved numpy library like a beast
• Compilation and just-in-time computingprocessing faster by compiling down to machine code, making sure youre guided by the results of profiling
• Concurrencyways to move data efficiently
• multiprocessingthe various ways to use the built-in multiprocessing library for parallel computing, efficiently share numpy matrices, and some costs and benefits of interprocess communication (IPC)
• Cluster computingconvert your multiprocessing code to run on a local or remote cluster for both research and production systems
• Using less RAMapproaches to solving large problems without buying a humungous computer
• Lessons from the fieldlessons encoded in war stories from those who took the blows so you dont have to

Python 3 is the future of Python, and everyone is moving toward it. Python 2.7 will nonetheless be around for many years to come (some installations still use Python 2.4 from 2004); its retirement date has been set at 2020.

The shift to Python 3.3+ has caused enough headaches for library developers that people havebeen slow to port their code (with good reason), and therefore people have been slowto adopt Python 3. This is mainly due to the complexities of switching from a mix of string and Unicode datatypes in complicated applications to the Unicode and byte implementation in Python 3.

Typically, when you want reproducible results based on a set of trustedlibraries, you dont want to be at the bleeding edge. High performance Pythondevelopers are likely to be using and trusting Python 2.7 for years to come.

Most of the code in this book will run with little alteration for Python 3.3+(the most significant change will be with print turning from a statement intoa function). In a few places we specifically look at improvements that Python3.3+ provides. One item that might catch you out is the fact that / means integer division in Python 2.7, but it becomes float division in Python 3. Of coursebeing a good developer, your well-constructed unit test suite will already be testing your important code paths, so youll be alerted by your unit tests if this needs to be addressed in your code.

scipy and numpy have been Python 3compatible since late 2010. matplotlib was compatible from 2012, scikit-learn was compatible in 2013, and NLTK is expected to be compatible in 2014. Django has been compatible since 2013. The transition notes for each are available in their repositories and newsgroups; it is worth reviewing the processes they used if youre going to migrate older code to Python 3.